ChatGPT

Scaling enterprise AI requires overcoming architectural oversights that often stall pilots before production, a challenge that goes far beyond model selection. While generative AI prototypes are easy to spin up, turning them into reliable business assets involves solving the difficult problems of data engineering and governance. Ahead of AI & Big Data Global 2026 in London, Franny Hsiao, EMEA Leader of AI Architects at Salesforce, discussed why so many initiatives hit a wall and how organisations can architect systems that actually survive the real world. The ‘pristine island’ problem of scaling enterprise AI Most failures stem from the environment in which the AI is built. Pilots frequently begin in controlled settings that create a false sense of security, only to crumble when faced with enterprise scale. “The single most common architectural oversight that prevents AI pilots from scaling is the failure to architect a production-grade data infrastructure with built-in end to end governance from the start,” Hsiao explains. “Understandably, pilots often start on ‘pristine islands’ – using small, curated datasets and simplified workflows. But this ignores the messy reality of enterprise data: the complex integration, normalisation, and transformation required to handle real-world volume and variability.” When companies attempt to scale these island-based pilots without addressing the underlying data mess, the systems break. Hsiao warns that “the resulting data gaps and performance issues like inference latency render the AI systems unusable—and, more importantly, untrustworthy.” Hsiao argues that the companies successfully bridging this gap are those that “bake end-to-end observability and guardrails into the entire lifecycle.” This approach provides “visibility and control into how effective the AI systems are and how users are adopting the new technology.” Engineering for perceived responsiveness As enterprises deploy large reasoning models – like the ‘Atlas Reasoning Engine’ – they face a trade-off between the depth of the model’s “thinking” and the user’s patience. Heavy compute creates latency. Salesforce addresses this by focusing on “perceived responsiveness through Agentforce Streaming,” according to Hsiao. “This allows us to deliver AI-generated responses progressively, even while the reasoning engine performs heavy computation in the background. It’s an incredibly effective approach for reducing perceived latency, which often stalls production AI.” Transparency also plays a functional role in managing user expectations when scaling enterprise AI. Hsiao elaborates on using design as a trust mechanism: “By surfacing progress indicators that show the reasoning steps or the tools being used, as well images like spinners and progress bars to depict loading states, we don’t just keep users engaged; we improve perceived responsiveness and build trust. “This visibility, combined with strategic model selection – like choosing smaller models for fewer computations, meaning faster response times – and explicit length constraints, ensures the system feels deliberate and responsive.” Offline intelligence at the edge For industries with field operations, such as utilities or logistics, reliance on continuous cloud connectivity is a non-starter. “For many of our enterprise customers, the biggest practical driver is offline functionality,” states Hsiao. Hsiao highlights the shift toward on-device intelligence, particularly in field services, where the workflow must continue regardless of signal strength. “A technician can photograph a faulty part, error code, or serial number while offline. Then an on-device LLM can then identify the asset or error, and provide guided troubleshooting steps from a cached knowledge base instantly,” explains Hsiao. Data synchronisation happens automatically once connectivity returns. “Once a connection is restored, the system handles the ‘heavy lifting’ of syncing that data back to the cloud to maintain a single source of truth. This ensures that work gets done, even in the most disconnected environments.” Hsiao expects continued innovation in edge AI due to benefits like “ultra-low latency, enhanced privacy and data security, energy efficiency, and cost savings.” High-stakes gateways Autonomous agents are not set-and-forget tools. When scaling enterprise AI deployments, governance requires defining exactly when a human must verify an action. Hsiao describes this not as dependency, but as “architecting for accountability and continuous learning.” Salesforce mandates a “human-in-the-loop” for specific areas Hsiao calls “high-stakes gateways”: “This includes specific action categories, including any ‘CUD’ (Creating, Uploading, or Deleting) actions, as well as verified contact and customer contact actions,” says Hsiao. “We also default to human confirmation for critical decision-making or any action that could be potentially exploited through prompt manipulation.” This structure creates a feedback loop where “agents learn from human expertise,” creating a system of “collaborative intelligence” rather than unchecked automation. Trusting an agent requires seeing its work. Salesforce has built a “Session Tracing Data Model (STDM)” to provide this visibility. It captures “turn-by-turn logs” that offer granular insight into the agent’s logic. “This gives us granular step-by-step visibility that captures every interaction including user questions, planner steps, tool calls, inputs/outputs, retrieved chunks, responses, timing, and errors,” says Hsiao. This data allows organisations to run ‘Agent Analytics’ for adoption metrics, ‘Agent Optimisation’ to drill down into performance, and ‘Health Monitoring’ for uptime and latency tracking. “Agentforce observability is the single mission control for all your Agentforce agents for unified visibility, monitoring, and optimisation,” Hsiao summarises. Standardising agent communication As businesses deploy agents from different vendors, these systems need a shared protocol to collaborate. “For multi-agent orchestration to work, agents can’t exist in a vacuum; they need common language,” argues Hsiao. Hsiao outlines two layers of standardisation: orchestration and meaning. For orchestration, Salesforce is adopting open-source standards like MCP (Model Context Protocol) and A2A (Agent to Agent Protocol).” “We believe open source standards are non-negotiable; they prevent vendor lock-in, enable interoperability, and accelerate innovation.” However, communication is useless if the agents interpret data differently. To solve for fragmented data, Salesforce co-founded OSI (Open Semantic Interchange) to unify semantics so an agent in one system “truly understands the intent of an agent in another.” The future enterprise AI scaling bottleneck: agent-ready data Looking forward, the challenge will shift from model capability to data accessibility. Many organisations still struggle with legacy, fragmented infrastructure where “searchability and reusability” remain difficult. Hsiao predicts the next major hurdle – and solution – will be making enterprise data “‘agent-ready’ through searchable, context-aware architectures that replace traditional, rigid ETL pipelines.” This shift is necessary to enable “hyper-personalised and transformed user experience because agents can always access the right context.” “Ultimately, the next year isn’t about the race for *******, newer models; it’s about building the orchestration and data infrastructure that allows production-grade agentic systems to thrive,” Hsiao concludes. Salesforce is a key sponsor of this year’s AI & Big Data Global in London and will have a range of speakers, including Franny Hsiao, sharing their insights during the event. Be sure to swing by Salesforce’s booth at stand #163 for more from the company’s experts. See also: Databricks: Enterprise AI adoption shifts to agentic systems Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Franny Hsiao, Salesforce: Scaling enterprise AI appeared first on AI News. View the full article

2026 will see forward-thinking organisations building out their squads of AI agents across roles and functions. But amid the rush, there is another aspect to consider. One of IDC’s enterprise technology predictions for the coming five years, published in October, was fascinating. “By 2030, up to 20% of [global 1000] organisations will have faced lawsuits, substantial fines, and CIO dismissals, due to high-profile disruptions stemming from inadequate controls and governance of AI agents,” the analyst noted. How do you therefore put guardrails in place – and how do you ensure these agents work together and, ultimately, do business together? Patrick Tobler, founder and CEO of blockchain infrastructure platform provider NMKR, is working on a project which aims to solve this – by fusing agentic AI and decentralisation. The Masumi Network, born out of a collaboration between NMKR and Serviceplan Group, launched in late 2024 as a framework-agnostic infrastructure which ‘empowers developers to build autonomous agents that collaborate, monetise services, and maintain verifiable trust.’ “The core thesis of Masumi is that there’s going to be billions of different AI agents from different companies interacting with each other in the future,” explains Tobler. “The difficult part now is – how do you actually have agents from different companies that can interact with each other and send money to each other as well, across these different companies?” Take travel as an example. You want to attend an industry conference, so your hotel booking agent buys a plane ticket from your airline agent. The entire experience and transaction will be seamless – but that implicit trust is required. “Masumi is a decentralised network of agents, so it’s not relying on any centralised payment infrastructure,” says Tobler. “Instead, agents are equipped with wallets and can send stablecoins from one agent to another and, because of that, interacting with each other in a completely safe and trustless manner.” For Tobler, having spent in his words ‘a lot of time’ in crypto, he determined that its benefits were being pointed to the wrong place. “I think there’s a lot of these problems that we have solved in crypto for humans, and then I came to this conclusion that maybe we’ve been solving them or the wrong target audience,” he explains. “Because for humans, using crypto and wallets and blockchains, all that kind of stuff is extremely difficult; the user experience is not great. But for agents, they don’t care if it’s difficult to use. They just use it, and it’s very native to them. “So all these issues that are now arising with agents having to interact with millions, or maybe even billions, of agents in the future – these problems have all already been solved with crypto.” Tobler is attending AI & Big Data Expo Global as part of Discover Cardano; NMKR started on the Cardano blockchain, while Masumi is built completely on Cardano. He says he is looking forward to speaking with businesses that are ‘hearing a lot about AI but aren’t really using it much besides ChatGPT’. “I want to understand from them what they are doing, and then figure out how we can help them,” he says. “That’s most often the thing missing from traditional tech startups. We’re all building for our own bubble, instead of actually talking to the people that would be using it every day.” Discover Cardano is exhibiting at the AI & Big Data Expo Global, in London on February 4-5. Watch the full video interview with NMKR’s Patrick Tobler below: Photo by Google DeepMind The post Masumi Network: How AI-blockchain fusion adds trust to burgeoning agent economy appeared first on AI News. View the full article

A White House paper titled “Artificial Intelligence and the Great Divergence” sets out parallels between the effects of the industrial revolution in the 18th and 19th centuries and the current times, with artificial intelligence positioned as guiding the way the world’s economies will be shaped. Artificial intelligence now sits at the centre of US economic strategy, currently representing a significant portion of the country’s economic activity, as characterised by the building of AI infrastructure, most notably in the form of data centres. The paper says AI investment raised US GDP by 1.3% percent in the first half of 2025, and compares this with the investment in the railway network during the industrial revolution. “Artificial Intelligence and the Great Divergence” says long-term growth depends primarily on gains in productivity, and AI is the tool to achieve those gains. It presents a range of estimates of AI’s impact on GDP, from single-digit increases to 20% productivity growth inside a decade. It also floats some more extreme scenarios, where GDP grows at more than 45% as AI substitutes for human labour in the longer term. Capital deployment in the form of building AI infrastructure, not growing consumption or public spending, is now creating US economic growth. Investment in data processing equipment, buildings, infrastructure, and software grew 28% in early 2025, and AI-related infrastructure represented around a quarter of all US investment in 2025. Training compute capacity used by AI models has increased roughly four-fold per year since 2010, and the length of tasks AI systems can complete has doubled every seven months for six years, the paper states. The cost per token of AI output has fallen by factors ranging from nine to nine hundred per year, depending on task and model. By late 2025, around 78% percent of organisations had reported using AI, up from 55% in 2024, and it’s claimed that 40% of US workers use generative AI in their jobs. Nearly half of US businesses now pay for AI subscriptions. The report poses these figures as evidence that AI has moved from experimentation into routine production. Internationally, the document frames AI as a factor in divergence of economic prosperity, with AI in the US increasing America’s GDP growth faster than in Europe and China. The US leads at the moment in private AI investment, model development, and compute capacity, while the EU’s share of world GDP has fallen since 1980. Additionally, the continent lags in comparable AI metrics – investment, construction, software development, overall capacity, etc. China remains a major player in AI actor, but the report notes that much of its model training relies on US-designed hardware. The White House publication advocates for an integrated national strategy with investment incentives at its core. The One Big Beautiful Bill Act gave significant financial breaks for data centres and IT infrastructure, and created favourable conditions for speedy facility construction, in line with the Act’s aim to lift GDP growth by more than a percentage point per year over the medium term. The report argues that deregulation in the AI industry supports productivity by lowering costs, increasing competition, and speeding innovation. Trade agreements and foreign policy reinforce this approach, with overseas partners committing to large purchases of US-derived AI chips and infrastructure. The paper notes that AI data centres are electricity-intensive, and projects that demand for power by AI infrastructure could reach up to 12% of domestic electricity consumption by 2028. It links the success of AI to energy availability and the ability of the power grid to deliver, positioning the control of energy supply as a prerequisite for international leadership in AI. The report’s conclusion is that the countries that lead in AI investment and adoption will experience higher growth than the mean. The United States is aligning multiple policy rafts to ensure its leading position in the sector. Businesses that build systems in line with its national goals will be part of a dominant economic force shaping the next phase of global growth. (Image source: “Chicago Thaws into Spring” by Trey Ratcliff is licensed under CC BY-NC-SA 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post White House compares industrial revolution with AI era appeared first on AI News. View the full article

Artificial intelligence has moved into the US workplace, but its adoption remains uneven, fragmented, and tied to role, industry, and organisation. Findings from a Gallup Workforce survey covering the ******* to the end of December 2025 show how employees use AI, who benefits most from it, and where areas of uncertainty remain. The findings draw from a nationally-representative questioning of more than 23,000 US adults in full- and part-time work, conducted online in August 2025. Its conclusions are that instances of AI in the workplace are increasing, but its use is far from universal, and is concentrated among knowledge-based workers. The office AI Employees in technology, finance, and professional services are by far the biggest user group. More than three-quarters of those working in IT report using AI “at least a few times a year”. In finance and professional services, the figure is a touch under 60%. AI-enabled or aided roles tend to be those that involve significant digital workflow and information synthesis; tasks that correspond with AI’s current abilities. AI use is lower in sectors dominated by customer-facing or manual work. Only around a third of retail workers report comparable levels of use to their office counterparts, although those in healthcare and manufacturing do tend to deploy AI more often than those in retail, for example. The fact that current raft of AI platforms fit more naturally into desk-based, cognitive roles seems obvious – less so is a drop-off in user numbers in tightly-regulated environments. Do we, or don’t we? Gallup’s data reveals a significant number of workers ore unsure whether or not their employer had adopted AI – nearly a quarter of those surveyed weren’t sure. In the third quarter of 2025, just over a third of employees said their organisation had implemented AI. 40% said there was no adoption of AI in their place of work It’s worth noting that earlier versions of Gallup surveys didn’t include a “don’t know” option for questions about employers’ AI adoption, which encouraged respondents to guess. Belief in organisational AI adoption appeared to rise sharply between 2024 and 2025, therefore, Gallup says. Once uncertainty could be stated explicitly, it became clear a good number of employees were simply uninformed on the matter. It’s staff in non-managerial roles who are more likely to say they’re unaware of their organisation’s AI use, a tendency mirrored in part-time staff and hands-on roles. The further workers are from decision-making, it seems, the less sure they become. How workers use AI The way employees use AI are consistent: of those using AI at least once a year, the most common applications are consolidating information, searching for information, and “generating ideas”, tasks that have changed little since Gallup first measured workplace AI use in 2024. More than 60% of AI users refer to chatbots, with using AI for writing and editing coming some way behind. Coding assistants and data science tools remain niche, but popular. Employees who use AI often are far more likely to use any more advanced tools at their disposal; particularly true in the cases of coding assistants and data analysis. Although use figures are generally up, Gallup concludes that AI has yet to be embedded in daily work for most Americans. Around 45% of workers say they use AI “a few times a year”, but only about 10% use it every day. Conclusions Business leaders have an easy win: simply clarifying a position on AI use would be a positive, plus publicising the availability (or otherwise) of AI tools would be an easy way to improve adoption rates. The current abilities of AI pertain to desk-based, digital and data-centric workflows, although there are a myriad of platforms that will utilise AI in other roles. Exploring these more fully would certainly be bucking the trend, and may make the difference between an organisation’s long-term prospects and those of its direct competitors. A page detailing Gallup’s findings can be found on the company’s website. (Image source: “DIY Open Plan Office” by lower29 is licensed under CC BY-NC-SA 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Gallup Workforce shows details of AI adoption in US workplaces appeared first on AI News. View the full article

For banks trying to put AI into real use, the hardest questions often come before any model is trained. Can the data be used at all? Where is it allowed to be stored? Who is responsible once the system goes live? At Standard Chartered, these privacy-driven questions now shape how AI systems are built, and deployed at the bank. For global banks operating in many jurisdictions, these early decisions are rarely straightforward. Privacy rules differ by market, and the same AI system may face very different constraints depending on where it is deployed. At Standard Chartered, this has pushed privacy teams into a more active role in shaping how AI systems are designed, approved, and monitored in the organisation. “Data privacy functions have become the starting point of most AI regulations,” says David Hardoon, Global Head of AI Enablement at Standard Chartered. In practice, that means privacy requirements shape the type of data that can be used in AI systems, how transparent those systems need to be, and how they are monitored once they are live. Privacy shaping how AI runs The bank is already running AI systems in live environments. The transition from pilots brings practical challenges that are easy to underestimate early on. In small trials, data sources are limited and well understood. In production, AI systems often pull data from many upstream platforms, each with its own structure and quality issues. “When moving from a contained pilot into live operations, ensuring data quality becomes more challenging with multiple upstream systems and potential schema differences,” Hardoon says. David Hardoon, Global Head of AI Enablement at Standard Chartered Privacy rules add further constraints. In some cases, real customer data cannot be used to train models. Instead, teams may rely on anonymised data, which can affect how quickly systems are developed or how well they perform. Live deployments also operate at a much larger scale, increasing the impact of any gaps in controls. As Hardoon puts it, “As part of responsible and client-centric AI adoption, we prioritise adhering to principles of fairness, ethics, accountability, and transparency as data processing scope expands.” Geography and regulation decide where AI works Where AI systems are built and deployed is also shaped by geography. Data protection laws vary in regions, and some countries impose strict rules on where data must be stored and who can access it. These requirements play a direct role in how Standard Chartered deploys AI, particularly for systems that rely on client or personally identifiable information. “Data sovereignty is often a key consideration when operating in different markets and regions,” Hardoon says. In markets with data localisation rules, AI systems may need to be deployed locally, or designed so that sensitive data does not cross borders. In other cases, shared platforms can be used, provided the right controls are in place. This results in a mix of global and market-specific AI deployments, shaped by local regulation not a single technical preference. The same trade-offs appear in decisions about centralised AI platforms versus local solutions. Large organisations often aim to share models, tools, and oversight in markets to reduce duplication. Privacy laws do not always block this approach. “In general, privacy regulations do not explicitly prohibit transfer of data, but rather expect appropriate controls to be in place,” Hardoon says. There are limits: some data cannot move in borders at all, and certain privacy laws apply beyond the country where data was collected. The details can restrict which markets a central platform can serve and where local systems remain necessary. For banks, this often leads to a layered setup, with shared foundations combined with localised AI use cases where regulation demands it. Human oversight remains central As AI becomes more embedded in decision-making, questions around explainability and consent grow harder to avoid. Automation may speed up processes, but it does not remove responsibility. “Transparency and explainability have become more crucial than before,” Hardoon says. Even when working with external vendors, accountability remains internal. This has reinforced the need for human oversight in AI systems, particularly where outcomes affect customers or regulatory obligations. People also play a larger role in privacy risk than technology alone. Processes and controls can be well designed, but they depend on how staff understand and handle data. “People remain the most important factor when it comes to implementing privacy controls,” Hardoon says. At Standard Chartered, this has pushed a focus on training and awareness, so teams know what data can be used, how it should be handled, and where the boundaries lie. Scaling AI under growing regulatory scrutiny requires making privacy and governance easier to apply in practice. One approach the bank is taking is standardisation. By creating pre-approved templates, architectures, and data classifications, teams can move faster without bypassing controls. “Standardisation and re-usability are important,” Hardoon explains. Codifying rules around data residency, retention, and access helps turn complex requirements into clearer components that can be reused in AI projects. As more organisations move AI into everyday operations, privacy is not just a compliance hurdle. It is shaping how AI systems are built, where they run, and how much trust they can earn. In banking, that shift is already influencing what AI looks like in practice – and where its limits are set. (Photo by Corporate Locations) See also: The quiet work behind Citi’s 4,000-person internal AI rollout Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Inside Standard Chartered’s approach to running AI under privacy rules appeared first on AI News. View the full article

According to Databricks, enterprise AI adoption is shifting to agentic systems as organisations embrace intelligent workflows. Generative AI’s first wave promised business transformation but often delivered little more than isolated chatbots and stalled pilot programmes. Technology leaders found themselves managing high expectations with limited operational utility. However, new telemetry from Databricks suggests the market has turned a corner. Data from over 20,000 organisations – including 60 percent of the Fortune 500 – indicates a rapid shift toward “agentic” architectures where models do not just retrieve information but independently plan and execute workflows. This evolution represents a fundamental reallocation of engineering resources. Between June and October 2025, the use of multi-agent workflows on the Databricks platform grew by 327 percent. This surge signals that AI is graduating to a core component of system architecture. The ‘Supervisor Agent’ drives enterprise adoption of agentic AI Driving this growth is the ‘Supervisor Agent’. Rather than relying on a single model to handle every request, a supervisor acts as an orchestrator, breaking down complex queries and delegating tasks to specialised sub-agents or tools. Since its launch in July 2025, the Supervisor Agent has become the leading agent use case, accounting for 37 percent of usage by October. This pattern mirrors human organisational structures: a manager does not perform every task but ensures the team executes them. Similarly, a supervisor agent manages intent detection and compliance checks before routing work to domain-specific tools. Technology companies currently lead this adoption, building nearly four times more multi-agent systems than any other industry. Yet the utility extends across sectors. A financial services firm, for instance, might employ a multi-agent system to handle document retrieval and regulatory compliance simultaneously, delivering a verified client response without human intervention. Traditional infrastructure under pressure As agents graduate from answering questions to executing tasks, underlying data infrastructure faces new demands. Traditional Online Transaction Processing (OLTP) databases were designed for human-speed interactions with predictable transactions and infrequent schema changes. Agentic workflows invert these assumptions. AI agents now generate continuous, high-frequency read and write patterns, often creating and tearing down environments programmatically to test code or run scenarios. The scale of this automation is visible in the telemetry data. Two years ago, AI agents created just 0.1 percent of databases; today, that figure sits at 80 percent. Furthermore, 97 percent of database testing and development environments are now built by AI agents. This capability allows developers and “vibe coders” to spin up ephemeral environments in seconds rather than hours. Over 50,000 data and AI apps have been created since the Public Preview of Databricks Apps, with a 250 percent growth rate over the past six months. The multi-model standard Vendor lock-in remains a persistent risk for enterprise leaders as they seek to increase agentic AI adoption. The data indicates that organisations are actively mitigating this by adopting multi-model strategies. As of October 2025, 78 percent of companies utilised two or more Large Language Model (LLM) families, such as ChatGPT, Claude, Llama, and Gemini. The sophistication of this approach is increasing. The proportion of companies using three or more model families rose from 36 percent to 59 percent between August and October 2025. This diversity allows engineering teams to route simpler tasks to smaller and more cost-effective models while reserving frontier models for complex reasoning. Retail companies are setting the pace, with 83 percent employing two or more model families to balance performance and cost. A unified platform capable of integrating various proprietary and open-source models is rapidly becoming a prerequisite for the modern enterprise AI stack. Contrary to the big data legacy of batch processing, agentic AI operates primarily in the now. The report highlights that 96 percent of all inference requests are processed in real-time. This is particularly evident in sectors where latency correlates directly with value. The technology sector processes 32 real-time requests for every single batch request. In healthcare and life sciences, where applications may involve patient monitoring or clinical decision support, the ratio is 13 to one. For IT leaders, this reinforces the need for inference serving infrastructure capable of handling traffic spikes without degrading user experience. Governance accelerates enterprise AI deployments Perhaps the most counter-intuitive finding for many executives is the relationship between governance and velocity. Often viewed as a bottleneck, rigorous governance and evaluation frameworks function as accelerators for production deployment. Organisations using AI governance tools put over 12 times more AI projects into production compared to those that do not. Similarly, companies employing evaluation tools to systematically test model quality achieve nearly six times more production deployments. The rationale is straightforward. Governance provides necessary guardrails – such as defining how data is used and setting rate limits – which gives stakeholders the confidence to approve deployment. Without these controls, pilots often get stuck in the proof-of-concept phase due to unquantified safety or compliance risks. The value of ‘boring’ enterprise automation from agentic AI While autonomous agents often conjure images of futuristic capabilities, current enterprise value from agentic AI lies in automating the routine, mundane, yet necessary tasks. The top AI use cases vary by sector but focus on solving specific business problems: Manufacturing and automotive: 35% of use cases focus on predictive maintenance. Health and life sciences: 23% of use cases involve medical literature synthesis. Retail and consumer goods: 14% of use cases are dedicated to market intelligence. Furthermore, 40 percent of the top AI use cases address practical customer concerns such as customer support, advocacy, and onboarding. These applications drive measurable efficiency and build the organisational muscle required for more advanced agentic workflows. For the C-suite, the path forward involves less focus on the “magic” of AI and more on the engineering rigour surrounding it. Dael Williamson, EMEA CTO at Databricks, highlights that the conversation has shifted. “For businesses across EMEA, the conversation has moved on from AI experimentation to operational reality,” says Williamson. “AI agents are already running critical parts of enterprise infrastructure, but the organisations seeing real value are those treating governance and evaluation as foundations, not afterthoughts.” Williamson emphasises that competitive advantage is shifting back towards how companies build, rather than simply what they buy. “Open, interoperable platforms allow organisations to apply AI to their own enterprise data, rather than relying on embedded AI features that deliver short-term productivity but not long-term differentiation.” In highly regulated markets, this combination of openness and control is “what separates pilots from competitive advantage.” See also: Anthropic selected to build government AI assistant pilot Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Databricks: Enterprise AI adoption shifts to agentic systems appeared first on AI News. View the full article

Anthropic has been selected to build government AI assistant capabilities to modernise how citizens interact with complex state services. For both public and private sector technology leaders, the integration of LLMs into customer-facing platforms often stalls at the proof-of-concept stage. The ***’s Department for Science, Innovation, and Technology (DSIT) aims to bypass this common hurdle by operationalising its February 2025 Memorandum of Understanding with Anthropic. The joint project, announced today, prioritises the deployment of agentic AI systems that are designed to actively guide users through processes rather than simply retrieving static information. The decision to move beyond standard chatbot interfaces addresses a friction point in digital service delivery: the gap between information availability and user action. While government portals are data-rich, navigating them requires specific domain knowledge that many citizens lack. By employing an agentic system powered by Claude, the initiative seeks to provide tailored support that maintains context across multiple interactions. This approach mirrors the trajectory of private sector customer experience, where the value proposition is increasingly defined by the ability to execute tasks and route complex queries rather than just deflect support tickets. The case for agentic AI assistants in government The initial pilot focuses on employment, a high-volume domain where efficiency gains directly impact economic outcomes. The system is tasked with helping users find work, access training, and understand available support mechanisms. For the government, the operational logic involves an intelligent routing system that can assess individual circumstances and direct users to the correct service. This focus on employment services also serves as a stress test for context retention capabilities. Unlike simple transactional queries, job seeking is an ongoing process. The system’s ability to “remember” previous interactions allows users to pause and resume their journey without re-entering data; a functional requirement that is essential for high-friction workflows. For enterprise architects, this government implementation serves as a case study in managing stateful AI interactions within a secure environment. Implementing generative AI within a statutory framework necessitates a risk-averse deployment strategy. The project adheres to a “Scan, Pilot, Scale” framework, a deliberate methodology that forces iterative testing before wider rollout. This phased approach allows the department to validate safety protocols and efficacy in a controlled setting, minimising the potential for compliance failures that have plagued other public sector AI launches. Data sovereignty and user trust form the backbone of this governance model. Anthropic has stipulated that users will retain full control over their data, including the ability to opt out or dictate what the system remembers. By ensuring all personal information handling aligns with *** data protection laws, the initiative aims to preempt privacy concerns that typically stall adoption. Furthermore, the collaboration involves the *** AI Safety Institute to test and evaluate the models, ensuring that the safeguards developed inform the eventual deployment. Avoiding dependency on external AI providers like Anthropic Perhaps the most instructive aspect of this partnership for enterprise leaders is the focus on knowledge transfer. Rather than a traditional outsourced delivery model, Anthropic engineers will work alongside civil servants and software developers at the Government Digital Service. The explicit goal of this co-working arrangement is to build internal AI expertise that ensures the *** government can independently maintain the system once the initial engagement concludes. This addresses the issue of vendor lock-in, where public bodies become reliant on external providers for core infrastructure. By prioritising skills transfer during the build phase, the government is treating AI competence as a core operational asset rather than a procured commodity. This development is part of a broader trend of sovereign AI engagement, with Anthropic expanding its public sector footprint through similar education pilots in Iceland and Rwanda. It also reflects a deepening investment in the *** market, where the company’s London office is expanding its policy and applied AI functions. Pip White, Head of ***, Ireland, and Northern Europe at Anthropic, said: “This partnership with the *** government is central to our mission. It demonstrates how frontier AI can be deployed safely for the public benefit, setting the standard for how governments integrate AI into the services their citizens depend on.” For executives observing this rollout, it once again makes clear that successful AI integration is less about the underlying model and more about the governance, data architecture, and internal capability built around it. The transition from answering questions to guiding outcomes represents the next phase of digital maturity. See also: How Formula E uses Google Cloud AI to meet net zero targets Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Anthropic selected to build government AI assistant pilot appeared first on AI News. View the full article

If database technologies offered performance, flexibility and security, most professionals would be happy to get two of the three, and they might have to expect to accept some compromises, too. Systems optimised for speed demand manual tuning, while flexible platforms can impose costs when early designs become constraints. Security is, sadly, sometimes, a bolt-on, with DBAs relying on internal teams’ skills and knowledge not to introduce breaking changes. RavenDB, however, exists because its founder saw the cumulative costs of those common trade-offs, and the inherent problems stemming from them. They wanted a database system that didn’t force developers and administrators to choose. Abstracting away complexity Oren Eini, RavenDB’s founder and CTO was working as a freelance database performance consultant nearly two decades ago. In an exclusive interview he recounted how he encountered many capable teams “digging themselves into a hole” as the systems in their care grew in complexity. Problems he was presented with didn’t stem from developers not possessing the required skills, but rather from system architecture. Databases tend to guide their developers towards fragile designs and punish developers for following those paths, he says. RavenDB was a project that began as a way to reduce friction when the unstoppable force of what’s required meets the mountain of database schema. The platform’s emphasis is on performance and adaptability without (ironically) at some stage requiring the services of people like Oren. Armed with a bag full of experience and knowledge, he formed RavenDB, which has now been shipping for more than fifteen years – well before the current interest in AI-assisted development. The bottom line is that over time, the RavenDB database adapts to what the organisation cares about, rather than what it guessed it might care about when the database was first spun up. “When I talk to business people,” Eini says, “I tell them I take care of data ownership complexity.” For example, instead of expecting developers or DBAs to anticipate every possible query pattern, RavenDB observes queries as they are executed. If it detects that a query would benefit from an index, it creates one in the background, with minimal overhead on extant processing. This contrasts with most relational databases, where schema and indexing strategies are set by the initial developers, so are difficult to alter later, regardless of how an organisation may have changed. Oren draws the comparison with pouring a building’s foundations before deciding where the doors and support columns might go. It’s an approach that can work, but when the business changes direction over the years, the cost of regretting those early decisions can be alarming. Oren Eini (source: RavenDB) Speaking ahead of the company’s appearance at the upcoming TechEx Global event in London this year (February 4 & 5, Olympia), he cited an example of a European client that struggled to expand into US markets because its database assumed a simple VAT rate that it had consigned to a single field, a schema not suitable for the complexities of state and federal sales taxes. From seemingly simple decisions made in the past (and perhaps not given much thought – European VAT is fairly standard), the client was storing financial pain and technical debt for the next generation. Much of RavenDB’s attractiveness is manifest in practical details and small tweaks that make databases more performant and easier to address. Pagination, for example, requires two database calls in most systems (one to fetch a page of results, another to count matching records). RavenDB returns both in a single query. Individually, such optimisations may appear minor, but at scale they compound. Oren says. “If you smooth down the friction everywhere you go, you end up with a really good system where you don’t have to deal with friction.” Compounded removal of frictions improves performance and makes developers’ jobs simpler. Related data is embedded or included without the penalties associated with table joins in relational databases, so complex queries are completed in a single round trip. Software engineers don’t need to be database specialists. In their world, they just formulate SQL-like queries to RavenDB’s APIs. Compared to other NoSQL databases, Raven DB provides full ACID transactions by default, and reduced operational complexity: many of its baked-in features (ETL pipelines, subscriptions, full-text search, counters, time series, etc.) reduce the need for external systems. In contrast with DBAs and software developers addressing a competing database system and its necessary adjuncts, both developers and admins spend less time sweating the detail with Raven DB. That’s good news, not least for those that hold an organisation’s purse strings. Scaling to fit the purpose RavenDB is also built to scale, as painlessly as it handles complex queries. It can create multi-node clusters if wanted so supports huge numbers of concurrent users. Such clusters are created by RavenDB without time-consuming manual configuration. “With RavenDB, this is normal cost of business,” he says. In February this year, RavenDB Cloud announced version 7.2, and this being 2026, mention needs to be made of AI. Raven DB’s AI Assistant is, “in effect, […] a virtual DBA that comes inside of your database,” he says. The key word is inside. It’s designed for developers and administrators, not end users, answering their questions about indexing, storage usage or system behaviour. AI as a professional tool He’s sceptical about giving AIs unconfined access to any data store. Allowing an AI to act as a generic gatekeeper to sensitive information creates unavoidable security risks, because such systems are difficult to constrain reliably. For the DBA and software developer, it’s another story – AI is a useful tool that operates as a helping hand, configuring and addressing the data. RavenDB’s AI assistant inherits the permissions of the user invoking it, having no privileged access of its own. “Anything it knows about your RavenDB instance comes because, behind the scenes, it’s accessing your system with your permissions,” he says. The company’s AI strategy is to provide developers and admins with opinionated features: generating queries, explaining indexes, helping with schema exploration, and answering operational questions, with calls bounded by operator validation and privileges. Teams developing applications with RavenDB get support for vector search, native embeddings, server-side indexing, and agnostic integration with external LLMs. This, Oren says, lets organisations deliver useful AI-driven features in their applications quickly, without exposing the business to risk and compliance issues. Security and risk Security and risk comprise one of those areas where RavenDB draws a clear line between it and its competitors. We touched on the recent MongoBleed vulnerability, which exposed data from unauthenticated MongoDB instances due to an interaction between compression and authentication code. Oren describes the issue as an architectural failure caused by mixing general-purpose and security-critical code paths. “The reason this is a vulnerability,” he says, “is specifically the fact that you’re trying to mix concerns.” RavenDB uses established cryptographic infrastructure to handle authentication before any database logic is invoked. And even if a flaw emanated from elsewhere, the attack surface would be significantly smaller because unauthenticated users never reach the general code paths: that architectural separation limits the blast radius. While the internals of RavenDB are highly technical and specialised, business decision-makers can easily appreciate that delays caused by schema changes, performance tuning, or infrastructure changes will have significant economic impact. But RavenDB’s malleability and speed also remove what Oren describes as the “no, you can’t do that” conversations. Organisations running RavenDB reduce their dependency on specialist expertise, plus they get the ability to respond to changing business needs much more quickly. “[The database’s] role is to bring actual business value,” Eini says, arguing that infrastructure should, in operational contexts, fade into the background. As it stands, it often determines the scope of strategy discussions. Migration and getting started RavenDB uses a familiar SQL-like query language, and most teams will only need a day at most to get up to speed. Where friction does appear, Oren suggests, it is often due to assumptions carried over from other platforms around security and high availability. For RavenDB, these are built into the design so don’t cause extra workload that needs to be factored in. Coming about as the result of the experience of operational pain by the company’s founder himself, RavenDB’s difference stems from accumulated design decisions: background indexing, query-aware optimisation, the separation of security and authentication issues, and latterly, the need for constraints on AI tooling. In everyday use, developers experience fewer sharp edges, and in the longer term, business leaders see a reduction in costs, especially around the times of change. The combination is compelling enough to displace entrenched platforms in many contexts. To learn more, you can speak to RavenDB representatives at TechEx Global, held at Olympia, London, February 4 and 5. If what you’ve read here has awakened your interest, head over to the company’s website. (Image source: “#316 AVZ Database” by Ralf Appelt is licensed under CC BY-NC-SA 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Lowering the barriers databases place in the way of strategy, with RavenDB appeared first on AI News. View the full article

The severe weather experienced at present in the US has placed significant strain on the airline industry in the country, with knock-on effects of changes to schedules and routes affecting the rest of the world. It’s at times like this that companies have to respond to queries from customers at a much greater rate than during normal operations, and there are – in the specific case of the air sector – operational decisions that need to be taken quickly, yet inside the strictest safety boundaries. Several airlines are turning to generative AI to help them during these types of events, and more generally, to help turn them into more efficient and reactive organisations. Last year, Air France-KLM built a cloud-based generative AI ‘factory’ for use throughout the organisation, which it described as letting it make AI development more consistent and reusable. It formed a partnership with Accenture and Google Cloud for its factory, using it to test and deploy generative AI models. It produces measurable outcomes in ground operations, engineering and maintenance, and customer-facing functions. The partnership group has stated that enterprise deployment of generative AI has increased development speed by more than 35%. The AI factory was built on earlier work undertaken by the airline and Accenture, which involved migrating core applications to the cloud. Since then, Air France-KLM has created a private AI assistant and RAG tools linking LLMs with internal search to support tasks like diagnosing and repairing aircraft damage. The factory is also used by employees, who get trained on how to use AI tools in order that they can use the power of LLMs to make a positive impact to the business. Weather and when AI is used United Airlines is similarly exploring AI in its operations. In an interview with CIO.com, CIO Jason Birnbaum described AI as a way to “shorten decision cycles” during irregular operations such as the recent outages caused by the current extreme cold snap. The company’s AI journey began with the use of AI to respond to passenger enquiries. When flights are delayed or cancelled, customer service representatives are expected to respond quickly and informatively, yet retain a company-mandated communication style – honed during the company’s ‘Every Flight Has A Story’ programme. During extended periods of disruption, maintaining the output from what the company terms ‘storytellers’ difficult. Jason Birnbaum said, “Considering the number of delays versus storytellers, we couldn’t have a person write a new message with every event. So we focused on prioritising the most impactful situations. […] The data piece was simple: the basic facts of the flight and the running chat between the attendants, pilots, gate agents, and the operations people associated with the flight. We fed that information — with additional data on weather, for example — into the AI model, to generate a good draft customer message.” “The trick then was to have it understand the nuances of United Airlines’ communications style and what we wanted to emphasise. That’s where prompt engineering came in, not to train the model to understand flight data, but to use the words United prefers. Let’s take safety, for instance. We can emphasise safety with without scaring people, and the AI tool is learning to make the right word choice. […] The AI model was very good at looking back in time to bring previous flight data into the current situation. Even our human storytellers didn’t include reasons for flight delays, and that kind of information can be very useful to a customer.” Boston Consulting Group’s measure of AI maturity in industries pegs airlines at ‘average’, having moved from slightly below average in the past year. Only one of the 36 airlines surveyed met the highest criteria for being prepared for an AI-enabled future. The analysis suggests that by 2030, carriers that embed AI at the core of their workflows could achieve operating margins that are 5% to 6% points higher than those of peers. It’s thought that generative AI will become part of the operational core of airlines and airports, where decisions about schedules, crew allocations, aircraft rotations, and passenger recovery have to be made quickly. Microsoft claims data-driven AI systems can reduce the root causes of flight delays by up to 35% through improved disruption forecasting, which can limit the negative effects of the spread of disruption. Airlines using AI-driven personalisation report revenue increases of around 10% to 15% per passenger, according to Microsoft, which also says that AI-based tools such as self-service customer interfaces can lead to cost reductions of up to 30%. (Image source: “airplane” by Kuster & Wildhaber Photography is licensed under CC BY-ND 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Cold snap highlight’s airlines’ proactive use of AI appeared first on AI News. View the full article

Enterprise AI has moved from isolated prototypes to systems that shape real decisions: drafting customer responses, summarising internal knowledge, generating code, accelerating research, and powering agent workflows that can trigger actions in business systems. That creates a new security surface, one that sits between people, proprietary data, and automated execution. AI security tools exist to make those questions operational. Some focus on governance and discovery. Others harden AI applications and agents at runtime. Some emphasise testing and red teaming before deployment. Others help security operations teams handle the new class of alerts AI introduces in SaaS and identity layers. What counts as an “AI security tool” in enterprise environments? “AI security” is an umbrella term. In practice, tools tend to fall into a few functional buckets, and many products cover more than one. AI discovery & governance: identifies AI use in employees, apps, and third parties; tracks ownership and risk LLM & agent runtime protection: enforces guardrails at inference time (prompt injection defenses, sensitive data controls, tool-use restrictions) AI security testing & red teaming: tests models and workflows against adversarial techniques before (and after) production release AI supply chain security: assesses risks in models, datasets, packages, and dependencies used in AI systems SaaS & identity-centric AI risk control: manages risk where AI lives inside SaaS apps and integrations, permissions, data exposure, account takeover, risky OAuth scopes A mature AI security programme typically needs at least two layers: one for governance and discovery, and another for runtime protection or operational response, depending on whether your AI footprint is primarily “employee use” or “production AI apps.” Top 10 AI security tools for enterprises in 2026 1) Koi Koi is the best AI security tool for enterprises because of its approach to AI security from the software control layer, helping enterprises govern what gets installed and adopted in endpoints, including AI-adjacent tooling like extensions, packages, and developer assistants. The matters because AI exposure often enters through tools that look harmless: browser extensions that read page content, IDE add-ons that access repositories, packages pulled from public registries, and fast-moving “helper” apps that become embedded in daily workflows. Rather than treating AI security as a purely model-level concern, Koi focuses on controlling the intake and spread of tools that can create data exposure or supply chain risk. In practice, that means turning ad-hoc installs into a governed process: visibility into what’s being requested, policy-based decisions, and workflows that reduce shadow adoption. For security teams, it provides a way to enforce consistency in departments without relying on manual policing. Key features include: Visibility into installed and requested tools in endpoints Policy-based allow/block decisions for software adoption Approval workflows that reduce shadow AI tooling sprawl Controls designed to address extension/package risk and tool governance Evidence trails for what was approved, by whom, and under what policy 2) Noma Security Noma Security is often evaluated as a platform for securing AI systems and agent workflows at the enterprise level. It focuses on discovery, governance, and protection of AI applications in teams, especially when multiple business units deploy different models, pipelines, and agent-driven processes. A key reason enterprises shortlist tools like Noma is scale: once AI adoption spreads, security teams need a consistent way to understand what exists, what it touches, and which workflows represent elevated risk. That includes mapping AI apps to data sources, identifying where sensitive information may flow, and applying governance controls that keep pace with change. Key features include: AI system discovery and inventory in teams Governance controls for AI applications and agents Risk context around data access and workflow behaviour Policies that support enterprise oversight and accountability Operational workflows designed for multi-team AI environments 3) Aim Security Aim Security is positioned around securing enterprise adoption of GenAI, especially the use layer where employees interact with AI tools and where third-party applications add embedded AI features. The makes it particularly relevant for organisations where the most immediate AI risk is not a custom LLM app, but workforce use and the difficulty of enforcing policy in diverse tools. Aim’s value tends to show up when enterprises need visibility into AI use patterns and practical controls to reduce data exposure. The goal is to protect the business without blocking productivity: enforce policy, guide use, and reduce unsafe interactions while preserving legitimate workflows. Key features include: Visibility into enterprise GenAI use and risk patterns Policy enforcement to reduce sensitive data exposure Controls for third-party AI tools and embedded AI features Governance workflows aligned with enterprise security needs Central management in distributed user populations 4) Mindgard Mindgard stands out for AI security testing and red teaming, helping enterprises pressure-test AI applications and workflows against adversarial techniques. The is especially important for organisations deploying RAG and agent workflows, where risk often comes from unexpected interaction effects: retrieved content influencing instructions, tool calls being triggered in unsafe contexts, or prompts leaking sensitive context. Mindgard’s value is proactive: instead of waiting for issues to surface in production, it helps teams identify weak points early. For security and engineering leaders, this supports a repeatable process, similar to application security testing, where AI systems are tested and improved over time. Key features include: Automated testing and red teaming for AI workflows Coverage for adversarial behaviours like injection and jailbreak patterns Findings designed to be actionable for engineering teams Support for iterative testing in releases Security validation aligned with enterprise deployment cycles 5) Protect AI Protect AI is often evaluated as a platform approach that spans multiple layers of AI security, including supply chain risk. The is relevant for enterprises that depend on external models, libraries, datasets, and frameworks, where risk can be inherited through dependencies not created internally. Protect AI tends to appeal to organisations that want to standardise security practices in AI development and deployment, including the upstream components that feed into models and pipelines. For teams that have both AI engineering and security responsibilities, that lifecycle perspective can reduce gaps between “build” and “secure.” Key features include: Platform coverage in AI development and deployment stages Supply chain security focus for AI/ML dependencies Risk identification for models and related components Workflows designed to standardise AI security practices Support for governance and continuous improvement 6) Radiant Security Radiant Security is oriented toward security operations enablement using agentic automation. In the AI security context, that matters because AI adoption increases both the number and novelty of security signals, new SaaS events, new integrations, new data paths, while SOC bandwidth stays limited. Radiant focuses on reducing investigation time by automating triage and guiding response actions. The key difference between helpful automation and dangerous automation is transparency and control. Platforms in this category need to make it easy for analysts to understand why something is flagged and what actions are being recommended. Key features include: Automated triage designed to reduce analyst workload Guided investigation and response workflows Operational focus: reducing noise and speeding decisions Integrations aligned with enterprise SOC processes Controls that keep humans in the loop where needed 7) Lakera Lakera is known for runtime guardrails that address risks like prompt injection, jailbreaks, and sensitive data exposure. Tools in this category focus on controlling AI interactions at inference time, where prompts, retrieved content, and outputs converge in production workflows. Lakera tends to be most valuable when an organisation has AI applications that are exposed to untrusted inputs or where the AI system’s behaviour must be constrained to reduce leakage and unsafe output. It’s particularly relevant for RAG apps that retrieve external or semi-trusted content. Key features include: Prompt injection and jailbreak defense at runtime Controls to reduce sensitive data exposure in AI interactions Guardrails for AI application behaviour Visibility and governance for AI use patterns Policy tuning designed for enterprise deployment realities 8) CalypsoAI CalypsoAI is positioned around inference-time protection for AI applications and agents, with emphasis on securing the moment where AI produces output and triggers actions. The is where enterprises often discover risk: the model output becomes input to a workflow, and guardrails must prevent unsafe decisions or tool use. In practice, CalypsoAI is evaluated for centralising controls in multiple models and applications, reducing the burden of implementing one-off protections in every AI project. The is particularly helpful when different teams ship AI features at different speeds. Key features include: Inference-time controls for AI apps and agents Centralised policy enforcement in AI deployments Security guardrails designed for multi-model environments Monitoring and visibility into AI interactions Enterprise integration support for SOC workflows 9) Cranium Cranium is often positioned around enterprise AI discovery, governance, and ongoing risk management. Its value is particularly strong when AI adoption is decentralised and security teams need a reliable way to identify what exists, who owns it, and what it touches. Cranium supports the governance side of AI security: building inventories, establishing control frameworks, and maintaining continuous oversight as new tools and features appear. The is especially relevant when regulators, customers, or internal stakeholders expect evidence of AI risk management practices. Key features include: Discovery and inventory of AI use in the enterprise Governance workflows aligned with oversight and accountability Risk visibility in internal and third-party AI systems Support for continuous monitoring and remediation cycles Evidence and reporting for enterprise AI programmes 10) Reco Reco is best known for SaaS security and identity-driven risk management, which is increasingly relevant to AI because so much “AI exposure” exists inside SaaS tools, copilots, AI-powered features, app integrations, permissions, and shared data. Rather than focusing on model behaviour, Reco helps enterprises manage the surrounding risks: account compromise, risky permissions, exposed files, overintegrations, and configuration drift. For many organisations, reducing AI risk starts with controlling the platforms where AI interacts with data and identity. Key features include: SaaS security posture and configuration risk management Identity threat detection and response for SaaS environments Data exposure visibility (files, sharing, permissions) Detection of risky integrations and access patterns Workflows aligned with enterprise identity and security operations Why AI security matters for enterprises AI creates security issues that don’t behave like traditional software risk. The three drivers below are why many enterprises are building dedicated AI security abilities. 1) AI can turn small mistakes into repeated leakage A single prompt can expose sensitive context: internal names, customer details, incident timelines, contract terms, design decisions, or proprietary code. Multiply that in thousands of interactions, and leakage becomes systematic not accidental. 2) AI introduces a manipulable instruction layer AI systems can be influenced by malicious inputs, direct prompts, indirect injection through retrieved content, or embedded instructions inside documents. A workflow may “look normal” while being steered into unsafe output or unsafe actions. 3) Agents expand blast radius from content to execution When AI can call tools, access files, trigger tickets, modify systems, or deploy changes, a security problem is not “wrong text.” It becomes “wrong action,” “wrong access,” or “unapproved execution.” That’s a different level of risk, and it requires controls designed for decision and action pathways, not just data. The risks AI security tools are built to address Enterprises adopt AI security tools because these risks show up fast, and internal controls are rarely built to see them end-to-end: Shadow AI and tool sprawl: employees adopt new AI tools faster than security can approve them Sensitive data exposure: prompts, uploads, and RAG outputs can leak regulated or proprietary data Prompt injection and jailbreaks: manipulation of system behaviour through crafted inputs Agent over-permissioning: agent workflows get excessive access “to make it work” Third-party AI embedded in SaaS: features ship inside platforms with complex permission and sharing models AI supply chain risk: models, packages, extensions, and dependencies bring inherited vulnerabilities The best tools help you turn these into manageable workflows: discovery → policy → enforcement → evidence. What Strong Enterprise AI Security Looks Like AI security succeeds when it becomes a practical operating model, not a set of warnings. High-performing programmes typically have: Clear ownership: who owns AI approvals, policies, and exceptions Risk tiers: lightweight governance for low-risk use, stronger controls for systems touching sensitive data Guardrails that don’t break productivity: strong security without constant “security vs business” conflict Auditability: the ability to show what is used, what is allowed, and why decisions were made Continuous adaptation: policies evolve as new tools and workflows emerge This is why vendor selection matters. The wrong tool can create dashboards without control, or controls without adoption. How to choose AI security tools for enterprises Avoid the trap of buying “the AI security platform.” Instead, choose tools based on how your enterprise uses AI. Map your AI footprint first Is most use employee-driven (ChatGPT, copilots, browser tools)? Are you building internal LLM apps with RAG, connectors, and access to proprietary knowledge? Do you have agents that can execute actions in systems? Is AI risk mostly inside SaaS platforms with sharing and permissions? Decide what must be controlled vs observed Some enterprises need immediate enforcement (block/allow, DLP-like controls, approvals). Others need discovery and evidence first. Prioritise integration and operational fit A great AI security tool that can’t integrate into identity, ticketing, SIEM, or data governance workflows will struggle in enterprise environments. Run pilots that mimic real workflows Test with scenarios your teams actually face: Sensitive data in prompts Indirect injection via retrieved documents User-level vs admin-level access differences An agent workflow that has to request elevated permissions Choose for sustainability The best tool is the one your teams will actually use after month three, when the novelty wears off and real adoption begins. Enterprises don’t “secure AI” by declaring policies. They secure AI by building repeatable control loops: discover, govern, enforce, validate, and prove. The tools above represent different layers of that loop. The best choice depends on where your risk concentrates, workforce use, production AI apps, agent execution pathways, supply chain exposure, or SaaS/identity sprawl. Image source: Unsplash The post Top 10 AI security tools for enterprises in 2026 appeared first on AI News. View the full article

Big retailers are committing more heavily to agentic AI-led commerce, and accepting some loss of customer proximity and data control in the process. As reported by Retail Dive, the opening weeks of 2026 have seen Etsy, Target and Walmart push product ranges onto third-party AI platforms, forming new partnerships with Google’s Gemini and Microsoft’s Copilot, after last year’s collaborations with OpenAI’s ChatGPT. These let consumers purchase goods inside the AI’s conversation interface. Amazon and Walmart have been investing in their own consumer-facing AI assistants, Rufus and Sparky respectively to change how shoppers interact with their brands. Agentic AI is beginning to redraw direct-to-consumer engagement, and industry figures regard this trend as an important moment in online retail. “I think this has the potential to disrupt retail in the same way the internet once did,” Kartik Hosanagar, a marketing professor at the Wharton School of the University of Pennsylvania, told the website’s reporters. Partnering with AIs like ChatGPT or Gemini engages consumers wherever they happen to be and may choose to shop. Adobe’s 2025 Holiday Shopping report found that AI-driven traffic to US e-commerce sites grew 758% year on year between in November 2025, and Cyber Monday saw a 670% increase in AI-referred retail visits. “What we expect is a deepening of consumer engagement,” Katherine ******, a partner at Kearney specialising in food, drug and mass-market retail, said in an email to Retail Dive. “More shoppers will rely on AI for purchasing, and across a wider range of missions. As retailers’ capabilities within these tools improve, adoption should accelerate further.” Meeting customers on AI platforms comes with trade-offs, according to industry observers, with questions around data ownership and the risk that retailers are sidelined. 81% of retail executives believe generative AI will erode brand loyalty by 2027, according to Deloitte’s 2026 Retail Industry Global Outlook, published earlier this month. Retailers’ websites or apps provide a stream of behavioural data, and if discovery, evaluation, and purchase happen externally, any insight doesn’t reach the retailer. “This fundamentally changes where power sits,” Hosanagar said. “Control over the agent increasingly means control over the customer relationship.” Google and Alphabet CEO Sundar Pichai has unveiled new commerce tools for Gemini, outlining how it will support customers from discovery to final purchase. Nikki Baird, vice president of strategy and product at Aptos, says this raises difficult questions. “What he’s describing is Google owning the data across discovery, decision and transaction. Even if some information is shared back, missing context from those stages leaves retailers with a much poorer understanding of their customers.” Pichai reassured retailers collaboration remains central to Google. “From nearly three decades of working with retailers, we know success only comes when we work together,” he told an NRF audience. “Our aim is to use our full technology stack to help shape the next era of retail.” Yet agentic systems’ features like instant checkout absorb the shopping experience into one platform. “If research, discovery and purchase all happen on OpenAI rather than Walmart.com, you’re effectively giving away the brand experience. At that point, the retailer risks becoming little more than a fulfilment operation,” Hosanagar said. Amazon has not announced plans to sell directly through ChatGPT, doubling down on its own AI initiatives. Earlier this month, the company launched a dedicated site for Alexa+, its generative AI assistant that helps users research and plan purchases. Yet participation in third-party AI commerce may become unavoidable. When OpenAI launched its Instant Checkout feature on ChatGPT last September, it suggested that enabling the function could influence how merchants are ranked in search results, in addition to price and product quality. Uploading product catalogues to AI chat platforms may be the first step in a transformation of online retail. According to Deloitte, roughly half of retail executives expect the current multi-stage shopping process to reduce to a single AI-driven interaction by 2027. For now the industry remains at an early stage of any transition. “The real inflection point is when consumers rely on an autonomous agent to shop on their behalf,” Hosanagar told Retail Dive. “Retailers will engage less with humans directly and more with their representatives — AI agents. That agent processes information differently, requires data in new formats and responds to persuasion in ways unlike a person.” Today, consumers can access ChatGPT on their phones while in-store, effectively consulting an always-available expert. “It’s not just the internet in your pocket,” Baird told Retail Dive. “It’s like having a highly knowledgeable store associate who knows every retailer.” This may prompt retailers to equip frontline staff with their own AI tools, offering instant insight into customer preferences or shopping history. Alternatively, a retailer’s AI agent could proactively notify customers when a favoured item is back in stock, helping associates convert interest into sales. “The goal is to enable store associates to perform at their best,” Baird said. (Image source: “Shopping trauma!” by Elsie esq. is licensed under CC BY 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Retailers examine options for on-AI retail appeared first on AI News. View the full article

AI continues to reshape technology and business; yet for the network, enterprise connectivity in the AI age means being always-on, and extra vigilant for sovereignty and security besides. This means that speed is not the only requirement. As Julian Skeels, chief digital officer at Expereo notes, it is more about ‘certainty.’ “AI workloads are distributed, they’re continuous, they’re incredibly latency-sensitive. Inference, monitoring, retrieval and remediation never stop, so that changes the network’s role,” says Skeels. “In the world of AI, networking actually becomes a system dependency,” he adds. “When the network degrades, the application degrades immediately. “An AI-ready network needs to make data movement deterministic. It’s not just about it being fast; it’s about it being predictable, and observable, and governable, and resilient – and to do all those things under continual change.” Many CIOs, however, are struggling right now with what Skeels describes as ‘connectivity everywhere but visibility nowhere.’ “They’re dealing with hybrid networks, multiple clouds, multiple providers and portals that create a constant operational drag to their teams,” says Skeels. “What they want is clarity and control – not more tools.” Skeels arrived at Expereo last year with myriad cross-industry experience in product and digital transformation initiatives under his belt. He found an industry ripe for accelerative change, and a company determined to lead the way and ensure pricing global connectivity should take minutes rather than weeks. “When I came to Expereo, I saw that global connectivity has, I would say, largely resisted real digital transformation for a long time,” notes Skeels. “Most customers will still experience it as slow, and manual, and opaque, and fragmented across the dozens of providers and portals they need to work with. “We believe, though, that with emerging technologies such as agentic AI, that’s finally changing,” adds Skeels. “Our ambition here is to make global connectivity as simple, and immediate, and transparent as cloud computing is for our customers.” Enabling such change for customers requires that mix of speed and visibility – and this is where the expereoOne platform comes in, to provide what the company calls ‘visibility at the speed of life’ and give customers a single, global view of what is being deployed, how it is performing, and what it costs. Beyond visibility, customers also need proactivity, as Skeels explains. “We’re deeply integrated into our customers’ order management, their ITSM, their ERP systems, which makes working with Expereo at scale absolutely seamless,” he says. “The key point is that better visibility isn’t about more dashboards. It’s about connecting network behaviour to their business outcomes in terms of resilience, security experience, and cost.” Skeels is speaking at the Digital Transformation Expo Global on February 4-5 around designing the AI-ready network – and his session promises to subvert the usual advice for those in attendance. “I want to challenge a few things,” notes Skeels. “I want to ask people to consider even unlearning things they’ve learned in the past. “A lot of what we’ve taken for granted about networks no longer holds in an AI world.” Watch the full conversation between Julian Skeels and TechEx’s James Bourne below: Photo by Pixabay The post Expereo: Enterprise connectivity amid AI surge with ‘visibility at the speed of life’ appeared first on AI News. View the full article

Formula E is using Google Cloud AI to meet its net zero targets by driving efficiency across its global logistics and commercial operations. As part of an expanded multi-year agreement, the electric racing series will integrate Gemini models into its ecosystem to support performance analysis, back-office workflows, and event logistics. The collaboration demonstrates how sports organisations are utilising cloud infrastructure to drive tangible business outcomes, rather than just securing surface-level sponsorship. The partnership focuses on optimising business operations, ranging from race management to the fan experience. Operational twins and carbon data to achieve net zero targets While marketing visibility often drives sports partnerships, this agreement builds on a technical foundation first formalised in January 2025. The elevation to “Principal Partner” involves Formula E adopting Google Cloud technologies for business-critical functions. The immediate application involves optimising the complex logistics of a global championship. Advanced AI modelling of the back office and the creation of race and event digital twins allow the organisation to simulate and optimise site builds virtually. This application directly affects Scope 3 emissions. The capability to plan infrastructure virtually minimises the need for physical on-site reconnaissance and reduces the transport of heavy equipment. For a championship that is the only sport-certified net zero carbon entity since inception, maintaining this status requires finding efficiencies in the supply chain. The digital twin approach delivers a quantifiable reduction in the operational carbon footprint while maintaining performance. Beyond logistical modelling, the Google Cloud AI partnership extends into the workforce productivity layer. Formula E is deploying Google Workspace with Gemini AI to enable greater agility and efficiency across its organisation. The organisation intends to use these tools to accelerate performance and deliver faster operations. This reflects a broader trend where generative AI tools are provisioned to reduce administrative latency in distributed workforces. The viability of these implementations to achieve net zero targets is supported by previous collaborative projects. Formula E recently utilised Google’s AI Studio and Gemini models to execute the ‘Mountain Recharge’ initiative. Engineers used the models to map an optimal route for the GENBETA car during a mountain descent. The AI identified and analysed specific braking zones, calculating the necessary regenerative braking required to harvest enough energy to complete a full lap of the Monaco circuit subsequently. This specific use case demonstrates how high-dimensional data – including topography, friction, and energy consumption – can be processed to define physical execution. Using Google Cloud AI to enhance Formula E’s data product The partnership also addresses the commercial requirement to retain and grow a digital audience. Formula E has integrated a ‘Strategy Agent’ into its live broadcasts. This tool processes real-time data to provide viewers with tailored insights and predictions regarding race strategy and driver performance. Millions of viewers have utilised these insights, which explain complex race dynamics as they unfold. This mirrors the enterprise challenge of observability (i.e. taking vast streams of real-time technical data and synthesising them into understandable narratives for stakeholders.) Beyond helping to achieve net zero targets, the leadership at both organisations frames this expansion as a necessary evolution of their technical stack. Jeff Dodds, CEO of Formula E, said: “Our expanded partnership with Google Cloud is a true game-changer for Formula E and for motorsport as a whole. We are already pushing the boundaries of technology in sport, and this Principal Partnership confirms our vision. “The integration of Google Cloud’s AI capabilities will unlock a new dimension of real-time performance optimisation and strategic decision-making, both for the Championship and for our global broadcast audience. This collaboration will redefine how fans experience our races and set a new benchmark for technology integration in sport worldwide.” Tara Brady, President of Google Cloud EMEA, added: “Formula E is a hub of innovation, where milliseconds can define success. This expanded partnership is a testament to the power of Google Cloud’s AI and data analytics, showing how our technology can deliver a competitive advantage in the most demanding scenarios.” The progression from the initial partnership in January 2025 to this expanded scope suggests the pilot programs provided sufficient ROI to warrant a broader rollout. As organisations face pressure to balance performance with net zero targets, the use of virtual simulation to optimise physical deployment remains a high-value area for investment. See also: Controlling AI agent sprawl: The CIO’s guide to governance Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post How Formula E uses Google Cloud AI to meet net zero targets appeared first on AI News. View the full article

For many organisations, the AI debate has moved on from whether to adopt the technology to a harder question: why do the results feel uneven? New tools are in place, pilots are running, and budgets are rising, yet clear AI returns remain elusive. According to Cloudflare’s 2026 App Innovation Report, the difference often has less to do with AI itself and more to do with the state of the applications underneath it. The report, based on a survey of more than 2,300 senior leaders in APAC, EMEA, and the Americas, points to application modernisation as the clearest divider between organisations seeing real AI value and those still struggling. Companies that are ahead of schedule in modernising their applications are nearly three times more likely to report a clear payoff from their AI investments. In APAC, the link is even more explicit: 92% of leaders say updating their software was the single most important factor in improving their AI abilities. Modernisation, not experimentation, drives AI returns The finding re-frames AI success as a foundation problem not a tooling problem. AI systems depend on fast access to data, flexible architectures, and reliable integration points. Legacy applications, fragmented infrastructure, and brittle workflows make it harder for AI projects to move beyond isolated use cases. Modernised applications, by contrast, give organisations room to experiment, scale, and adapt without constant rework. The report describes this relationship as a reinforcing cycle. Organisations modernise applications to support AI, then use AI results to justify deeper modernisation. Leaders in this group report far higher confidence that their infrastructure can support AI development, and that confidence translates into action. In APAC, 90% of leading organisations have already integrated AI into existing applications, compared with much lower levels among those behind schedule. Around 80% plan to increase that integration further over the next year. The shift marks a change in mindset, as earlier waves of AI adoption focused on testing and pilots. Now, the emphasis is on integration. AI is not treated as a standalone project but as part of everyday systems, from internal workflows to customer-facing applications. The report shows that leading organisations are using AI to improve internal processes, build content-driven applications, and support revenue-generating work, while lagging organisations remain more cautious and fragmented in their approach. The cost of delay shows up in security and confidence The cost of falling behind is becoming clearer as well. Organisations that lag on modernisation tend to modernise reactively, often after a security incident or operational failure. In APAC, these organisations report lower confidence in both their infrastructure and their teams’ ability to support AI. That lack of confidence slows decision-making and limits how far AI projects can go. Instead of expanding use cases, teams spend time managing risk, fixing gaps, and dealing with technical debt. Security plays a central role in this dynamic. The report shows that organisations with strong alignment between security and application teams are far more likely to scale AI successfully. Where that alignment is weak, security issues consume time and attention, pushing modernisation and AI work further down the priority list. Many lagging organisations report difficulty tracking risks in applications and APIs, which makes it harder to move quickly without increasing exposure. For leaders, security is treated as part of application design not an add-on. That approach reduces the amount of reactive work needed after incidents and frees teams to focus on building and improving systems. Over time, this also lowers the operational drag that can stall AI efforts. The report suggests that reliability has become a practical limit on speed: organisations that cannot maintain stable, secure systems struggle to move AI projects into production. Fewer tools, clearer foundations, faster AI integration Another pressure point highlighted in the APAC data is tool sprawl. Nearly all organisations report challenges in managing large and complex technology stacks, but leaders are responding more aggressively. About 86% of APAC leaders say they are actively cutting redundant tools and addressing shadow IT. The goal is not just cost control, but clarity. Fewer platforms and integrations make it easier to modernise applications, apply consistent security controls, and integrate AI without friction. Developer time is also a factor. In organisations with a modernised foundation, developers spend more time maintaining and improving systems that already work. In lagging organisations, developers are more likely to rebuild from scratch or spend time on configuration and remediation. That difference affects how quickly new AI abilities can be introduced and refined. When teams are tied up fixing problems, AI becomes harder to prioritise. Taken together, the findings suggest that AI success is less about racing to deploy new models and more about removing the obstacles that slow everything else down. Application modernisation creates the conditions for AI to deliver value, while fragmented systems and reactive practices limit what AI can achieve. Without that foundation, organisations find it harder to turn AI investment into measurable AI returns. For APAC organisations, the message is that AI investment without modernisation tends to produce shallow results. Modernisation without integration plans risks becoming an ongoing rebuild. The organisations seeing the strongest returns are those that treat application updates, security alignment, and AI integration as connected work, not separate initiatives. The report does not suggest a single path forward, but it does draw a clear line between organisations that act early and those that wait. The advantage not comes from having AI, but from having applications ready to use it. (Photo by Julio Lopez) See also: Controlling AI agent sprawl: The CIO’s guide to governance Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Modernising apps triples the odds of AI returns, Cloudflare says appeared first on AI News. View the full article

Anthropic’s Economic Index offers a look at how organisations and individuals are actually using large language models. The report contains the company’s analysis of a million consumer interactions on Claude.ai, plus a million enterprise API calls, all dated from November 2025. The report notes that its figures are based on observations, rather than, for example, a sample of business decision-makers or generic survey. Limited use cases dominate Use of Anthropic’s AI tends to cluster around a relatively small number of tasks, with the ten most frequently-performed tasks accounting for almost a quarter of consumer interactions, and nearly a third of enterprise API traffic. There’s a focus on the use of Claude for code creation and modification, as readers might expect. This concentration of use of AI as a software development tool has remained fairly constant over time, suggesting that the model’s value is largely based around these types of tasks, with no emerging use of Claude for other purposes of any empirical significance. This suggests that broad, general rollouts of AI are less likely to be successful than those focused on tasks where large language models are proven to be effective. Augmentation outperforms automation On consumer platforms, collaborative use – where users iterate on queries to the AI over the course of a virtual conversation – is more common than using the AI to produce automated workflows. Enterprise API usage shows the opposite, as businesses attempt to gain savings through automating tasks. However, while Claude succeeds on shorter tasks, the observed quality of outcomes declines the more complex the task (or series of tasks) is, and the longer the required ‘thinking time’ required. This implies automation is most effective for routine, well-defined tasks that are simpler, require fewer logical steps, and where responses to queries can be quick. Tasks estimated to take humans several hours show significantly lower completion rates than shorter tasks. For longer tasks to succeed, users have to iterate and correct outputs. Users breaking down large tasks into manageable steps and posing each separately (either interactively or via API) have improved success rates. The company’s observations show most queries put to the LLMs are associated with white-collar roles (although poorer countries tend to use Claude in academic settings more commonly than, for instance, the US). For example, travel agents can lose complex planning tasks to the LLM and retain elements of their more transactional work, while some roles, such as property managers, show the opposite: routine administrative tasks can be handled by the AI, and tasks needing higher-judgement remain with the human professional.. Productivity gains lessened by reliability The report notes that claims of AI boosting annual labour productivity by 1.8% (over a decade) are likely best to be reduced to 1-1.2%, due to the need to factor in extra labour and costs. While a 1% efficiency gain over a decade is still economically meaningful, the need for activities such as validation, error handling, and reworking will lower success rates and therefore there should be a similar adjustment in the minds of a business’s decision-makers. Potential gains to an organisation deploying AI also depend on whether tasks given to the LLM complement or substitute work. In the latter case, the success of substituting an AI for tasks normally done by a human depends on how complex the work is. It’s noteworthy that the report finds a near-perfect correlation between the sophistication of users’ prompts to the LLM and successful outcomes. Thus, how people use AI shapes what it delivers. Key takeaways for leaders AI implementation delivers value fastest in specific, well-defined areas. Complementary systems (AI+human) outperform full automation for complex work. Reliability and necessary extra work ‘around’ the AI reduce predicted productivity gains. Changes to workforces’ makeup depend on the mix of tasks and their complexity, not specific job roles. (Image source: “the virtual construction worker” by antjeverena is licensed under CC BY-NC-SA 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Anthropic’s usage stats paint a detailed picture of AI success appeared first on AI News. View the full article

Cyber threats don’t follow predictable patterns, forcing security teams to rethink how protection works at scale. Defensive AI is emerging as a practical response, combining machine learning with human oversight. Cybersecurity rarely fails because teams lack tools. It fails because threats move faster than detection can keep pace. As digital systems expand, attackers adapt in real time while static defences fall behind. This reality explains why AI security explained has become a central topic in modern cyber defense conversations. Why cyber defense needs machine learning now Attack techniques today are fluid. Phishing messages change wording in hours. Malware alters behaviour to avoid detection. Rule-based security struggles in this environment. Machine learning fills this void by learning how systems are expected to behave. In other words, it does not wait for a recognised pattern but searches for something that does not seem to fit. The is important when a threat is either new or camouflaged. For security teams, this change reduces blind spots. Machine learning processes data volumes that no human team could review manually. It connects subtle signals in networks, endpoints and cloud services. You see the benefit when response times shrink. Early detection limits damage. Faster containment protects data and continuity. In global environments, that speed often determines whether an incident stays manageable. How defensive AI identifies threats in real time Machine learning models are interested in behaviour and not in assumptions. Models learn by observing how users and applications interact. When activity breaks from expected patterns, alerts surface. This approach works even when the threat has never appeared before. Zero-day attacks really become visible because behaviour, not history, triggers concern. Common detection techniques include: Behavioural base-lining to spot unusual activity Anomaly detection in network and application traffic Classification models trained on diverse threat patterns Real-time analysis is essential. Modern attacks spread quickly in interconnected systems. Machine learning continuously evaluates streaming data, letting security teams react before damage escalates. This ability proves especially valuable in cloud environments. Resources change constantly. Traditional perimeter defences lose relevance. Behaviour-based monitoring adapts as systems evolve. Embedding defense across the AI security lifecycle Effective cyber defense does not start at deployment. It begins earlier and continues throughout a system’s lifespan. Machine learning technology evaluates development configurations and dependencies during development. High-risk configuration items and exposed services are identified before deployment to production. That makes them less exposed in the long run. Once systems go live, monitoring shifts to runtime behaviour. Access requests, inference activity and data flows receive constant attention. Unusual patterns prompt investigation. Post-deployment oversight remains critical. Use patterns change. Models age. Defensive AI detects drift that may signal misuse or emerging vulnerabilities. The lifecycle view reduces fragmentation. Security becomes consistent in stages not reactive after incidents occur. Over time, that consistency builds operational confidence. Defensive AI in complex enterprise environments Enterprise infrastructure rarely exists in one place. Cloud platforms, remote work and third-party services increase complexity. Defensive AI addresses this by correlating signals in environments. Isolated alerts become connected stories. Security teams gain context instead of noise. Machine learning also helps prioritise risk. Not every alert requires immediate action. By scoring threats based on behaviour and impact, AI reduces alert fatigue. This prioritisation improves efficiency. Analysts spend time where it matters most. Routine anomalies are monitored and not escalated. As organisations operate in regions, consistency becomes vital. Defensive AI applies the same analytical standards globally. That uniformity supports reliable protection without slowing operations. Human judgement in an AI-driven defense model Defensive AI is most effective when paired with human expertise. Automation deals with speed and volume. Human judgement and accountability are provided by humans. The ensures there is no blind trust in systems unaware of what is happening in the real world. Security specialists are involved in model training and testing. Human judgement is used to decide which behaviours are most significant. Context is always important for interpretation, particularly when business dynamics, roles and geographic considerations apply. Explainability is also a factor in trust. It is necessary to know the reason a warning was issued. Modern defensive systems are increasingly providing a reason for a decision, letting analysts review the results and make decisions with confidence not hesitation. The combination produces stronger results. AI points out potential dangers early, in large spaces. Humans make decisions about actions, focus on impact and mitigate effects. AI and humans create a robust defense system. In light of the increasingly adaptable nature of threats in cyberspace, this synergy has become imperative. The role of defensive AI in supporting the underlying foundation through analysis has been made possible through human oversight. Conclusions Cybersecurity exists in a reality that is defined by speed, scale and continuous change. The static nature of cyber-defense makes it inadequate in this reality, as attack vectors change faster than static cyber-defense measures can keep pace. Defensive AI represents a useful evolution. Machine learning improves detection, reduces response time and helps build resistance in complex systems by recognising nuanced patterns of human behaviour. But when paired with experienced human monitoring, defensive AI goes beyond automation. It can become an assured means of protecting contemporary digital infrastructure, facilitating stable security operations that don’t diminish responsibility or decision-making. Image source: Unsplash The post Defensive AI and how machine learning strengthens cyber defense appeared first on AI News. View the full article

Corporate networks are filling up with AI agents, creating a governance blind spot for leaders managing multi-cloud infrastructures. As distinct business units race to adopt generative technologies, CIOs especially find their ecosystems populated by fragmented and unmonitored assets. This mirrors the shadow IT challenges of the cloud era, but involves autonomous actors capable of executing business logic and accessing sensitive data. IDC projects the number of actively deployed AI agents will exceed one billion by 2029—a forty-fold increase from current levels. In the first half of 2025 alone, agent creation surged by 119 percent. For enterprise leadership, the immediate challenge shifts from building these agents to locating, auditing, and governing them across platforms. Salesforce has responded to this fragmentation by expanding its MuleSoft Agent Fabric capabilities, introducing automated discovery tools designed to centralise the management of AI agents regardless of their origin. Automating discovery Visibility remains the core issue for security and operations teams. When marketing teams deploy AI agents on one platform and logistics teams build on another, effective governance becomes difficult as central IT loses a consolidated view of the organisation’s digital workforce. MuleSoft’s updated architecture addresses this via ‘Agent Scanners’. These tools continuously patrol major ecosystems – including Salesforce Agentforce, Amazon Bedrock, and Google Vertex AI – to identify running agents. Rather than relying on developers to manually register their deployments, the system automates detection. Finding an agent is only the first step; compliance leaders need to understand the logic behind it. The scanners extract metadata detailing the agent’s capabilities, the LLMs driving it, and the specific data endpoints it is authorised to access. This information is then normalised into standard Agent-to-Agent (A2A) specifications, creating a uniform profile for assets regardless of the underlying vendor. Andrew Comstock, SVP and GM of MuleSoft, said: “The most successful organisations of the next decade will be those that harness the full diversity of the multi-cloud AI landscape. The expanded capabilities of MuleSoft Agent Fabric give you the freedom to innovate across any platform while maintaining the unified visibility and control needed to scale.” Governance and cost control for AI agents Unmanaged agents create financial inefficiency and risk exposure. Consider a CISO in the banking sector. Under standard operations, verifying a new loan-processing agent involves manually chasing documentation from development teams. Automated cataloguing allows security teams to immediately view which financial databases an agent accesses and verify its authorisation levels without manual intervention. This capability ensures security teams view real-time data rather than outdated snapshots. From a financial perspective, visibility drives consolidation. Large enterprises frequently suffer from redundancy where regional teams independently procure or build similar tools. A multinational manufacturer, for instance, might have three separate teams paying for distinct summarisation agents on different platforms. By using the MuleSoft Agent Visualizer to filter the estate by job type, operations leaders can identify these overlaps. Consolidating these into a single high-performing asset reduces redundant licensing costs and allows budget reallocation toward novel development. Transitioning successfully to an ‘Agentic Enterprise’ Innovation often occurs at the edges, where data scientists build bespoke tools outside formal procurement channels. The expanded Agent Fabric addresses this by allowing the registration of “homegrown” agents and Model Context Protocol (MCP) servers via URL. This is particularly relevant for sectors like logistics, where teams may build internal tools for proprietary database optimisation. Instead of remaining hidden, these assets can be registered and made discoverable for reuse across the company. Jonathan Harvey, Head of AI Operations at Capita, said: “Agent Scanners will let us focus on innovation instead of inventory management. Knowing that every agent is automatically discovered and catalogued allows our teams to collaborate, reuse work, and build smarter multi-agent solutions.” Similarly, AT&T is utilising the framework to orchestrate agents across customer support, chat, and voice interactions. Brad Ringer, Enterprise & Integration Architect at AT&T, explained: “With AI moving so fast, MuleSoft Agent Fabric provides the framework we need to scale. It brings together and helps us orchestrate all of the agents and MCP servers we’re building in customer support, chat, and voice interactions. It isn’t just a tool; it’s a huge enabler for everything we’re doing next.” The transition to an “Agentic Enterprise” requires a change in governance around how IT assets are tracked, rendering the days of managing integrations via stale spreadsheets incompatible with the speed of AI agent deployment. Leaders must assume their inventory of AI agents is incomplete and deploy automated scanning tools to establish a baseline of truth. Once this baseline is established, governance policies should mandate that all agents – whether bought or built – expose their capabilities and data access privileges in a standardised format like A2A to facilitate monitoring. Finally, executives can use the visibility provided by these tools to audit spend, identifying duplicate functionalities across cloud environments and merging them to control the Total Cost of Ownership (TCO). As organisations move from pilot programmes to mass deployment, the differentiator will not be the intelligence of individual agents, but the coherence of the network that connects them. See also: Balancing AI cost efficiency with data sovereignty Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Controlling AI agent sprawl: The CIO’s guide to governance appeared first on AI News. View the full article

Primary healthcare systems across parts of Africa are under growing strain, caught between rising demand, chronic staff shortages, and shrinking international aid budgets. In that context, AI is being tested in healthcare less as a breakthrough technology and more as a way to keep basic services running. According to reporting by Reuters, the Gates Foundation and OpenAI are backing a new initiative, Horizon1000, that aims to introduce AI tools into primary healthcare clinics across several African countries. The project will begin in Rwanda and is intended to reach 1,000 clinics and surrounding communities by 2028, supported by a combined $50 million investment. The timing is not accidental as global development assistance for health fell by just under 27% last year compared to 2024, the Gates Foundation estimates, following cuts that began in the United States and spread to other major donors such as Britain and Germany. Those reductions have coincided with the first rise in preventable child deaths this century, adding pressure to health systems already stretched thin. Rather than focusing on advanced diagnostics or research, Horizon1000 is framed around everyday tasks that consume time in under-resourced clinics. AI tools under the programme are expected to assist with patient intake, triage, record keeping, appointment scheduling, and access to medical guidance, particularly in settings where one doctor may serve tens of thousands of people. Gates Foundation and OpenAI focus on AI support in healthcare “In poorer countries with enormous health worker shortages and lack of health systems infrastructure, AI can be a gamechanger in expanding access to quality care,” Bill Gates wrote in a blog post announcing the initiative. Speaking to Reuters at the World Economic Forum in Davos, Gates said the technology could help health systems recover after aid cuts slowed progress. “Our commitment is that that revolution will at least happen in the poor countries as quickly as it happens in the rich countries,” he said. The focus, according to both partners, is on supporting healthcare workers rather than replacing them. OpenAI is expected to provide technical expertise and AI systems, while the Gates Foundation will work with African governments and health authorities to oversee deployment and alignment with national guidelines. Rwanda was chosen as the first pilot country in part because of its existing digital health efforts. The country established an AI health hub in Kigali last year and has positioned itself as a testbed for health technology projects. Paula Ingabire, Rwanda’s minister of information and communications technology and innovation, said the goal is to reduce administrative burdens while expanding access. “It is about using AI responsibly to reduce the burden on healthcare workers, to improve the quality of care, and to reach more patients,” Ingabire said in a video statement released alongside the launch. Under Horizon1000, AI tools may also be used before patients reach clinics. Gates told Reuters the systems could support pregnant women and **** patients with guidance ahead of visits, especially when language barriers exist between patients and providers. What the AI tools are expected to handle Once patients arrive, AI could help link records, reduce paperwork, and speed up routine processes. “A typical visit, we think, can be about twice as fast and much better quality,” Gates said. Those expectations highlight both the promise and the limits of the approach. While AI may help streamline workflows, its impact depends on reliable data, stable power and connectivity, trained staff, and clear oversight. Many previous digital health pilots in low-income settings have struggled to scale beyond initial trials once funding or external support tapered off. Horizon1000’s designers say they are trying to avoid that pattern by working closely with local governments and health leaders rather than deploying one-size-fits-all systems. Tools are meant to be adapted to local clinical rules, languages, and care models. Even so, questions remain about long-term maintenance, data governance, and who bears responsibility if systems fail or produce errors. The initiative also reflects a broader shift in how AI is being positioned in global health. Instead of headline-grabbing claims about medical breakthroughs, the emphasis here is on narrow, operational use cases that address staffing gaps and administrative overload. In that sense, AI is being treated less as a cure for weak health systems and more as a temporary support amid declining resources. OpenAI’s involvement comes as the company expands its presence in healthcare, following earlier work on health-related applications. At the same time, it faces growing scrutiny over how its systems are trained, deployed, and governed, especially in sensitive sectors like medicine. A test of AI’s limits in healthcare systems For African health systems, the stakes are practical rather than symbolic. Sub-Saharan Africa faces an estimated shortage of nearly six million healthcare workers, a gap that training alone cannot close in the near term. If AI tools can help clinicians see more patients, reduce errors, or manage workloads more effectively, they may offer some relief. If they add complexity or require constant outside support, they risk becoming another layer of dependency. Horizon1000 sits at that intersection. As aid budgets tighten and healthcare demands rise, the project offers a test of whether AI can play a useful, limited role in primary care without overstating its reach. The outcome will depend less on the technology itself than on how well it fits into the systems meant to use it. See also: SAP and Fresenius to build sovereign AI backbone for healthcare Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Gates Foundation and OpenAI test AI in African healthcare appeared first on AI News. View the full article

AI cost efficiency and data sovereignty are at odds, forcing a rethink of enterprise risk frameworks for global organisations. For over a year, the generative AI narrative focused on a race for capability, often measuring success by parameter counts and flawed benchmark scores. Boardroom conversations, however, are undergoing a necessary correction. While the allure of low-cost, high-performance models offers a tempting path to rapid innovation, the hidden liabilities associated with data residency and state influence are forcing a reassessment of vendor selection. China-based AI laboratory DeepSeek recently became a focal point for this industry-wide debate. According to Bill Conner, former adviser to Interpol and GCHQ, and current CEO of Jitterbit, DeepSeek’s initial reception was positive because it challenged the status quo by demonstrating that “high-performing large language models do not necessarily require Silicon Valley–scale budgets.” For businesses looking to trim the immense costs associated with generative AI pilots, this efficiency was understandably attractive. Conner observes that these “reported low training costs undeniably reignited industry conversations around efficiency, optimisation, and ‘good enough’ AI.” AI and data sovereignty risks Enthusiasm for cut-price performance has collided with geopolitical realities. Operational efficiency cannot be decoupled from data security, particularly when that data fuels models hosted in jurisdictions with different legal frameworks regarding privacy and state access. Recent disclosures regarding DeepSeek have altered the math for Western enterprises. Conner highlights “recent US government revelations indicating DeepSeek is not only storing data in China but actively sharing it with state intelligence services.” This disclosure moves the issue beyond standard GDPR or CCPA compliance. The “risk profile escalates beyond typical privacy concerns into the realm of national security.” For enterprise leaders, this presents a specific hazard. LLM integration is rarely a standalone event; it involves connecting the model to proprietary data lakes, customer information systems, and intellectual property repositories. If the underlying AI model possesses a “back door” or obliges data sharing with a foreign intelligence apparatus, sovereignty is eliminated and the enterprise effectively bypasses its own security perimeter and erases any cost efficiency benefits. Conner warns that “DeepSeek’s entanglement with military procurement networks and alleged export control evasion tactics should serve as a critical warning sign for CEOs, CIOs, and risk officers alike.” Utilising such technology could inadvertently entangle a company in sanctions violations or supply chain compromises. Success is no longer just about code generation or document summaries; it is about the provider’s legal and ethical framework. Especially in industries like finance, healthcare, and defence, tolerance for ambiguity regarding data lineage is zero. Technical teams may prioritise AI performance benchmarks and ease of integration during the proof-of-concept phase, potentially overlooking the geopolitical provenance of the tool and the need for data sovereignty. Risk officers and CIOs must enforce a governance layer that interrogates the “who” and “where” of the model, not just the “what.” Governance over AI cost efficiency Deciding to adopt or ban a specific AI model is a matter of corporate responsibility. Shareholders and customers expect that their data remains secure and used solely for intended business purposes. Conner frames this explicitly for Western leadership, stating that “for Western CEOs, CIOs, and risk officers, this is not a question of model performance or cost efficiency.” Instead, “it is a governance, accountability, and fiduciary responsibility issue.” Enterprises “cannot justify integrating a system where data residency, usage intent, and state influence are fundamentally opaque.” This opacity creates an unacceptable liability. Even if a model offers 95 percent of a competitor’s performance at half the cost, the potential for regulatory fines, reputational damage, and loss of intellectual property erases those savings instantly. The DeepSeek case study serves as a prompt to audit current AI supply chains. Leaders must ensure they have full visibility into where model inference occurs and who holds the keys to the underlying data. As the market for generative AI matures, trust, transparency, and data sovereignty will likely outweigh the appeal of raw cost efficiency. See also: SAP and Fresenius to build sovereign AI backbone for healthcare Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Balancing AI cost efficiency with data sovereignty appeared first on AI News. View the full article

For many large companies, artificial intelligence still lives in side projects. Small teams test tools, run pilots, and present results that struggle to spread beyond a few departments. Citi has taken a different path, where instead of keeping AI limited to specialists, the bank has spent the past two years pushing the technology into daily work in the organisation. That effort has resulted in an internal AI workforce of roughly 4,000 employees, drawn from roles that range from technology and operations to risk and customer support. The figure was first reported by Business Insider, which detailed how Citi built its “AI Champions” and “AI Accelerators” programmes to encourage participation not central control. The scale of integration is notable, as Citi employs around 182,000 people globally, and more than 70% of them now use firm-approved AI tools in some form, according to the same report. That level of use places Citi ahead of many peers that still restrict AI access to technical teams or innovation labs. From central pilots to team-level adoption Rather than start with tools, Citi focused on people. The bank invited employees to volunteer as AI Champions, giving them access to training, internal resources, and early versions of approved AI systems. The employees then supported colleagues in their own teams, acting as local points of contact not formal trainers. The approach reflects a practical view of adoption. New tools often fail not because they lack features, but because staff do not know when or how to use them. By embedding support inside teams, Citi reduced the gap between experimentation and routine work. Training played a central role. Employees could earn internal badges by completing courses or demonstrating how they used AI to improve their own tasks. The badges did not come with promotions or pay rises, but they helped create visibility and credibility in the organisation. According to Business Insider, this peer-driven model helped AI spread faster than top-down mandates. Everyday use, with guardrails Citi’s leadership has framed the effort as a response to scale not novelty. With operations spanning retail banking, investment services, compliance, and customer support, small efficiency gains can add up quickly. AI tools are being used to summarise documents, draft internal notes, analyse data sets, and assist with software development. None of these uses are new on their own, but the difference lies in how they are applied. The focus on everyday tasks also shapes Citi’s risk posture. The bank has limited employees to firm-approved tools, with guardrails around what data can be used and how outputs are handled. That constraint has slowed some experiments, but it has also made managers more comfortable allowing broader access. In regulated industries, trust often matters more than speed. What Citi’s approach shows about scaling AI The structure of Citi’s programme suggests a lesson for other large enterprises. AI adoption does not require every employee to become an expert. It requires enough people to understand the tools well enough to apply them responsibly and explain them to others. By training thousands instead of dozens, Citi reduced its reliance on a small group of specialists. There is also a cultural signal at play. Encouraging employees from non-technical roles to participate sends a message that AI is not only for engineers or data scientists. It becomes part of how work gets done, similar to spreadsheets or presentation software in earlier decades. That shift aligns with broader industry trends. Surveys from firms like McKinsey have shown that many companies struggle to move AI projects into production, often citing talent gaps and unclear ownership. Citi’s model sidesteps some of those issues by distributing ownership in teams, while keeping governance centralised. Still, the approach is not without limits. Peer-led adoption depends on sustained interest, and not all teams move at the same pace. There is also the risk that informal support networks become uneven, with some groups benefiting more than others. Citi has tried to address this by rotating Champions and updating training content as tools change. What stands out is the bank’s willingness to treat AI as infrastructure not innovation. Instead of asking whether AI could transform the business, Citi asked where it could remove friction from existing work. That framing makes progress easier to measure and reduces pressure to produce dramatic results. The experience also challenges a common assumption that AI adoption must start at the top. Citi’s senior leadership supported the effort, but much of the momentum came from employees who volunteered time to learn and teach. In large organisations, that bottom-up energy can be hard to generate, yet it often determines whether new technology sticks. As more companies move from pilots to production, Citi’s experiment offers a useful case study. It shows that scale does not come from buying more tools, but from helping people feel confident using the ones they already have. For enterprises wondering why AI progress feels slow, the answer may lie less in strategy decks and more in how work actually gets done, one team at a time. (Photo by Declan Sun) See also: JPMorgan Chase treats AI spending as core infrastructure Want to learn more about AI and big data from industry leaders? Check outAI & Big Data Expo taking place in Amsterdam, California, and London. This comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post The quiet work behind Citi’s 4,000-person internal AI rollout appeared first on AI News. View the full article

SAP and Fresenius are building a sovereign AI platform for healthcare that brings secure data processing to clinical settings. For data leaders in the medical sector, deploying AI requires strict governance that public cloud solutions often lack. This collaboration addresses that gap by creating a “controlled environment” where AI models can operate without compromising data sovereignty. Moving AI from pilot to production The project aims to build an open and integrated ecosystem allowing hospitals to use AI securely. Rather than running isolated experiments, the companies plan to create a digital backbone for a sovereign and AI-supported healthcare system. Michael Sen, CEO of Fresenius, said: “Together with SAP, we can accelerate the digital transformation of the ******* and European healthcare systems and enable a sovereign European solution that is so important in today’s global landscape. “We are making data and AI everyday companions that are secure, simple and scalable for doctors and hospital teams. This creates more room for what truly matters: caring for patients.” The technical base uses SAP Business AI and the SAP Business Data Cloud. By leveraging these components, the platform creates a compliant, sovereign foundation for operating AI models in healthcare. This infrastructure handles health data responsibly, a requirement for scaling automated processes in patient care. The partnership tackles data fragmentation through SAP’s “AnyEMR” strategy, which supports the integration of diverse hospital information systems (HIS). Using open industry standards like HL7 FHIR, the platform connects HIS, electronic medical records (EMRs), and other medical applications. This connectivity allows Fresenius to develop AI-supported solutions that increase efficiency across the care chain. The goal is to build an individual, scalable platform that enables connected, data-driven healthcare processes. Investing in sovereign AI to advance healthcare Both companies intend to invest a “mid three-digit million euro amount” in the medium term. The funds target the digital transformation of ******* and European healthcare systems using AI-supported solutions. Plans include joint investments in startups and scaleups, alongside internal technological developments. This approach aims to build a broader library of tools that plug into the sovereign platform. Christian Klein, CEO of SAP SE, commented: “With SAP’s leading technology and Fresenius’ deep healthcare expertise, we aim to create a sovereign, interoperable healthcare platform for Fresenius worldwide. “Together, we want to set new standards for data sovereignty, security, and innovation in healthcare. Thanks to SAP, Fresenius can harness the full potential of digital and AI-supported processes and sustainably improve patient care.” This deal indicates that the next phase of healthcare AI in Europe will focus on sovereign infrastructure. Scalable AI requires a controlled environment to satisfy regulatory demands—without a sovereign data backbone, AI initiatives risk stalling due to compliance concerns. See also: Scaling AI value beyond pilot phase purgatory Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post SAP and Fresenius to build sovereign AI backbone for healthcare appeared first on AI News. View the full article

Scaling AI value from isolated pilots to enterprise-wide adoption remains a primary hurdle for many organisations. While experimentation with generative models has become ubiquitous, industrialising these tools (i.e. wrapping them in necessary governance, security, and integration layers) often stalls. Addressing the gap between investment and operational return, IBM has introduced a new service model designed to help businesses assemble, rather than purely build, their internal AI infrastructure. Adopting asset-based consulting Traditional consultancy models typically rely on human labour to solve integration problems, a process that is often slow and capital-intensive. IBM is among the companies aiming to alter this dynamic by offering an asset-based consulting service. This approach combines standard advisory expertise with a catalogue of pre-built software assets, aiming to help clients construct and govern their own AI platforms. Instead of commissioning bespoke development for every workflow, organisations can leverage existing architectures to redesign processes and connect AI agents to legacy systems. This method helps companies to achieve value by scaling new agentic applications without necessitating alterations to their existing core infrastructure, AI models, or preferred cloud providers. Managing a multi-cloud environment A frequent concern for enterprise leaders is vendor lock-in, particularly when adopting proprietary platforms. IBM’s strategy acknowledges the reality of the heterogeneous enterprise IT landscape. The service supports a multi-vendor foundation, compatible with Amazon Web Services, Google Cloud, and Microsoft Azure, alongside IBM watsonx. This approach extends to the models themselves, supporting both open- and closed-source variants. By allowing companies to build upon their current investments rather than demanding a replacement strategy, the service addresses a barrier to adoption: the fear of technical debt accumulation when switching ecosystems. The technical backbone of this offering is IBM Consulting Advantage, the company’s internal delivery platform. Having utilised this system to support over 150 client engagements, IBM reports that the platform has boosted its own consultants’ productivity by up to 50 percent. The premise is that if these tools can accelerate delivery for IBM’s own teams, they should offer similar velocity for clients. The service provides access to a marketplace of industry-specific AI agents and applications. For business leaders, this suggests a “platform-first” focus, where attention turns from managing individual models to managing a cohesive ecosystem of digital and human workers. Active deployment of a platform-centric approach to scaling AI value The efficacy of such a platform-centric approach is best viewed through active deployment. Pearson, the global learning company, is currently utilising this service to construct a custom platform. Their implementation combines human expertise with agentic assistants to manage everyday work and decision-making processes, illustrating how the technology functions in a live operational environment. Similarly, a manufacturing firm has employed IBM’s solution to formalise its generative AI strategy. For this client, the focus was on identifying high-value use cases, testing targeted prototypes, and aligning leaders around a scalable strategy. The result was the deployment of AI assistants using multiple technologies within a secured, governed environment, laying a foundation for wider expansion across the enterprise. Despite the attention surrounding generative AI, the realisation of balance-sheet impact is not guaranteed. “Many organisations are investing in AI, but achieving real value at scale remains a major challenge,” notes Mohamad Ali, SVP and Head of IBM Consulting. “We have solved many of these challenges inside IBM by using AI to transform our own operations and deliver measurable results, giving us a proven playbook to help clients succeed.” The conversation is gradually moving away from the capabilities of specific LLMs and towards the architecture required to run them safely. Success in scaling AI and achieving value will likely depend on an organisation’s ability to integrate these solutions without creating new silos. Leaders must ensure that as they adopt pre-built agentic workflows, they maintain rigorous data lineage and governance standards. See also: JPMorgan Chase treats AI spending as core infrastructure Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Scaling AI value beyond pilot phase purgatory appeared first on AI News. View the full article

Artificial intelligence has shifted rapidly from a peripheral innovation to a structural component of modern financial services. In banking, payments, and wealth management, to name but three sub-sectors, AI is now embedded in budgeting tools, fraud detection systems, KYC, AML, and customer engagement platforms. Credit unions sit in this broader fintech transformation, facing similar technological pressures and operating under distinct cooperative models built on trust, proffered services in competitive markets, and community alignment. Consumer behaviour suggests AI is already part of everyday financial decision-making. Research from Velera indicates that 55% of consumers use AI tools for financial planning or budgeting, while 42% are comfortable using AI to complete financial transactions. Adoption is highest among younger demographics, with 80% of Gen Z and younger millennials using AI for financial planning and close to that proportion expressing ‘comfort’ with agentic AI. These patterns mirror trends in the wider fintech sector, where AI-driven personal finance tools and conversational interfaces have become more common. There is a particular a dual challenge for credit unions. Member expectations are shaped by large fintech companies’ digital platforms and apps, and large digital banks are deploying AI at scale. At the average Union, internal readiness remains limited. A CULytics survey shows that although 42% of credit unions have implemented AI in specific operational areas, only 8% report using it in multiple parts of the business. The gap between market expectations and institutional ability defines the current phase of AI adoption in the cooperative-based financial sector. AI as a trust-based extension of financial services Unlike many fintech startups, credit unions benefit from high levels of consumer trust. Velera reports that 85% of consumers see credit unions as reliable sources of financial advice, and 63% of CU members say they would attend AI-related educational sessions if such were offered. These findings position credit unions as being able to frame AI as an advisory tool to be embedded in existing relationships. In fintech, “explainable AI” and transparent digital finance are mainstays as identity verification, and regulation watch the technology closely. Regulators and consumers clearly expect transparency into how decisions are made by AI back ends. Credit unions can use this expectation by integrating AI into education programmes, fraud awareness efforts and financial literacy. Where AI delivers tangible value Personalisation is a leading use case for AI. Machine learning models let financial institutions move beyond static customer segmentation, via behavioural signals and life-stage indicators. The approach is already common in other sectors, and in the industry, in fintech lending and digital banking platforms. Credit unions can adopt similar techniques, ones that tailor offers, communications, and make product recommendations. Member service represents another potential high-impact area. According to CULytics, 58% of credit unions now use chatbots or virtual assistants, the most-adopted AI application in the sector. Cornerstone Advisors reports that deployment is accelerating among credit unions than banks, using AI to handle routine enquiries and preserve staff capacity. Fraud prevention has emerged as an AI use case in the sector. Alloy reports a 92% net increase in AI fraud prevention investment among credit unions in 2025, compared with lower prioritisation among banks. As digital payments get more widely-adopted, AI-driven fraud detection is important to balance security with low-friction user experiences. In this respect, credit unions face the same pressures as mainstream fintech payment providers and neobanks, where false declines and delayed responses can directly erode customer trust. Operational efficiency and lending decisions also feature prominently. Research from Inclind and CULytics shows AI being applied to reconciliation, underwriting, and internal business analytics. Users report reduced manual workloads and faster credit decisions. Cornerstone Advisors identifies lending as the third-most common AI function among credit unions, placing them closer to fintech lenders than traditional banks in this area. Structural barriers to scaling AI Despite clear use cases, scaling AI in credit unions remains difficult. Data readiness is the most frequently cited constraint. Cornerstone Advisors reports that only 11% of credit unions rate their data strategy as very effective (nearly a quarter consider it ineffective). Without accessible, well-governed data, AI systems cannot deliver reliable outcomes, regardless of the underlying sophistication of the LLM. Trust and explainability also limit the technology’s expansion. In regulated financial environments, opaque “****** box” models create risk for institutions that as a matter of course have to justify their decisions to members. PYMNTS Intelligence highlights the importance of breaking down data silos and using shared intelligence models to improve transparency and auditability. Consortium-based approaches, like those used by Velera in thousands of credit unions, reflect a trend in the financial sector towards pooled data. Integration presents a further challenge. CULytics finds that 83% of credit unions cite integration with legacy systems as an obstacle to AI, a familiar issue to many financial institutions. Limited in-house expertise in AI compounds this, again suggesting fintech partnerships, credit union service organisations (CUSOs), or externally-managed platforms as ways to accelerate deployment. From experimentation to embedded practice As AI becomes embedded in financial services, credit unions face a choice similar to that which has been confronted by banks and the wider fintech sector: placing AI as a foundational ability. Evidence suggests progress depends on disciplined execution. That means prioritising high-trust, high-impact use cases, so institutions can deliver visible benefits and not undermine members’ confidence in their trusted institutions. Strengthening data governance and accountability ensures AI-assisted decisions remain explainable and defensible. Partner-led integration might reduce technical complexity, while education and transparency align AI adoption with the values that underpin the cooperative organisation. (Image source: “Credit Union Building” by Dano is licensed under CC BY 2.0.) Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post Credit unions, fintech and the AI inflection of financial services appeared first on AI News. View the full article

Inside large banks, artificial intelligence has moved into a category once reserved for payment systems, data centres, and core risk controls. At JPMorgan Chase, AI is framed as infrastructure the bank believes it cannot afford to neglect. That position came through clearly in recent comments from CEO Jamie Dimon, who defended the bank’s rising technology budget and warned that institutions that fall behind on AI risk losing ground to competitors. The argument was not about replacing people but about staying functional in an industry where speed, scale, and cost discipline matter every day. JPMorgan has been investing heavily in technology for years, but AI has changed the tone of that spending. What once sat with innovation projects is now folded into the bank’s baseline operating costs. That includes internal AI tools that support research, document drafting, internal reviews, and other routine tasks in the organisation. From experimentation to infrastructure The shift in language reflects a deeper change in how the bank views risk. AI is considered part of the systems required to keep pace with competitors that are automating internal work. Rather than encouraging workers to rely on public AI systems, JPMorgan has focused on building and governing its own internal platforms. That decision reflects long-held concerns in banking about data exposure, client confidentiality, and regulatory monitoring. Banks operate in an environment where mistakes carry high costs. Any system that touches sensitive data or influences choices must be auditable and explainable. Public AI tools, trained on datasets and updated frequently, make that difficult. Internal systems give JPMorgan more control, even if they take longer to deploy. The approach also reduces the potential of uncontrolled “shadow AI,” in which employees use unapproved tools to speed up work. While such tools can improve productivity, they create gaps in oversight that regulators tend to notice quickly. A cautious approach to workforce change JPMorgan has been careful in how it talks about AI’s impact on jobs. The bank has avoided claims that AI will dramatically reduce headcount. Instead, it presents AI as a way to reduce manual work and improve consistency. Tasks that once required multiple review cycles can now be completed faster, with employees still responsible for final judgement. The framing positions AI as support not substitution, which matters in a sector sensitive to political and regulatory reaction. The scale of the organisation makes this approach practical. JPMorgan employs hundreds of thousands of people worldwide. Even tiny efficiency gains, applied broadly, can translate into meaningful cost savings over time. The upfront investment required to build and maintain internal AI systems is substantial. Dimon acknowledges that technology spending can have an impact on short-term performance, especially when market conditions are uncertain. His response is that cutting back on technology now may improve margins in the near term, but it risks weakening the bank’s position later. In that sense, AI spending is treated as a form of insurance against falling behind. JPMorgan, AI, and the risk of falling behind rivals JPMorgan’s stance reflects pressure in the banking sector. Rivals are investing in AI to speed up fraud detection, streamline compliance work, and improve internal reporting. As these tools become more common, expectations rise. Regulators may assume banks have access to advanced monitoring systems. Clients may expect faster responses and fewer errors. In that environment, lagging on AI can look less like caution and more like mismanagement. JPMorgan has not suggested that AI will solve structural challenges or eliminate risk. Many AI projects struggle to move beyond narrow uses, and integrating them into complex systems remains difficult. The harder work lies in governance. Deciding which teams can use AI, under what conditions, and with what oversight requires clear rules. Errors need defined escalation paths. Responsibility must be assigned when systems produce flawed output. Across large enterprises, AI adoption is not limited by access to models or computing power, but constrained by process, policy, and trust. For other end-user companies, JPMorgan’s approach offers a useful reference point. AI is treated as part of the machinery that keeps the organisation running. That does not guarantee success. Returns may take years to appear, and some investments will not pay off. But the bank’s position is that the greater risk lies in doing too little, not too much. (Photo by IKECHUKWU JULIUS UGWU) See also: Banks operationalise as Plumery AI launches standardised integration Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information. AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here. The post JPMorgan Chase treats AI spending as core infrastructure appeared first on AI News. View the full article

For the majority of web users, generative AI is AI. Large Language Models (LLMs) like GPT and Claude are the de facto gateway to artificial intelligence and the infinite possibilities it has to offer. After mastering our syntax and remixing our memes, LLMs have captured the public imagination. They’re easy to use and fun. And – the odd hallucination aside – they’re smart. But while the public plays around with their favourite flavour of LLM, those who live, breathe, and sleep AI – researchers, tech heads, developers – are focused on ******* things. That’s because the ultimate goal for AI max-ers is artificial general intelligence (AGI). That’s the endgame. To the professionals, LLMs are a sideshow. Entertaining and eminently useful, but ultimately ‘narrow AI.’ They’re good at what they do because they’ve been trained on specific datasets, but incapable of straying out of their lane and attempting to solve larger problems. The diminishing returns and inherent limitations of deep learning models is prompting exploration of smarter solutions capable of actual cognition. Models that lie somewhere between the LLM and AGI. One system that falls into this bracket – smarter than an LLM and a foretaste of future AI – is OpenCog Hyperon, an open-source framework developed by SingularityNET. With its ‘neural-symbolic’ approach, Hyperon is designed to bridge the gap between statistical pattern matching and logical reasoning, offering a roadmap that joins the dots between today’s chatbots and tomorrow’s infinite thinking machines. Hybrid architecture for AGI SingularityNET has positioned OpenCog Hyperon as a next-generation AGI research platform that integrates multiple AI models into a unified cognitive architecture. Unlike LLM-centric systems, Hyperon is built around neural-symbolic integration in which AI can learn from data and reason about knowledge. That’s because withneural-symbolic AI, neural learning components and symbolic reasoning mechanisms are interwoven so that one can inform and enhance the other. This overcomes one of the primary limitations of purely statistical models by incorporating structured, interpretable reasoning processes. At its core, OpenCog Hyperon combines probabilistic logic and symbolic reasoning with evolutionary programme synthesis and multi-agent learning. That’s a lot of terms to take it, so let’s try and break down how this all works in practice. To understand OpenCog Hyperon – and specifically why neural-symbolic AI is such a big deal – we need to understand how LLMs work and where they come up short. The limits of LLMs Generative AI operates primarily on probabilistic associations. When an LLM answers a question, it doesn’t ‘know’ the answer in the way a human instinctively does. Instead, it calculates the most probable sequence of words to follow the prompt based on its training data. Most of the time, this ‘impersonation of a person’ comes in very convincingly, providing the human user with not only the output they expect, but one that is correct. LLMs specialise in pattern recognition on an industrial scale and they’re very good at it. But the limitations of these models are well documented. There’s hallucination, of course, which we’ve already touched on, where plausible-sounding but factually incorrect information is presented. Nothing gaslights harder than an LLM eager to please its master. But a greater problem, particularly once you get into more complex problem-solving, is a lack of reasoning. LLMs aren’t adept at logically deducing new truths from established facts if those specific patterns weren’t in the training set. If they’ve seen the pattern before, they can predict its appearance again. If they haven’t, they hit a wall. AGI, in comparison, describes artificial intelligence that can genuinely understand and apply knowledge. It doesn’t just guess the right answer with a high degree of certainty – it knows it, and it’s got the working to back it up. Naturally, this ability calls for explicit reasoning skills and memory management – not to mention the ability to generalise when given limited data. Which is why AGI is still some way off – how far off depends on which human (or LLM) you ask. But in the meantime, whether AGI be months, years, or decades away, we have neural-symbolic AI, which has the potential to put your LLM in the shade. Dynamic knowledge on demand To understand neural-symbolic AI in action, let’s return toOpenCog Hyperon. At its heart is the Atomspace Metagraph, a flexible graph structure that represents diverse forms of knowledge including declarative, procedural, sensory, and goal-directed, all contained in a single substrate. The metagraph can encode relationships and structures in ways that support not just inference, but logical deduction and contextual reasoning. If this sounds a lot like AGI, it’s because it is. ‘Diet AGI,’ if you like, provides a taster of where artificial intelligence is headed next. So that developers can build with the Atomspace Metagraph and use its expressive power, Hyperon has created MeTTa (Meta Type Talk), a novel programming language designed specifically for AGI development. Unlike general-purpose languages like Python, MeTTa is a cognitive substrate that blends elements of logic and probabilistic programming. Programmes in MeTTa operate directly on the metagraph, querying and rewriting knowledge structures, and supporting self-modifying code, which is essential for systems that learn how to improve themselves. "We're emerging from a couple of years spent on building tooling. We've finally got all our infrastructure working at scale for Hyperon, which is exciting." Our CEO, Dr. @bengoertzel, joined Robb Wilson and Josh Tyson on the Invisible Machines podcast to discuss the present and… pic.twitter.com/8TqU8cnC2L — SingularityNET (@SingularityNET) January 19, 2026 Robust reasoning as gateway to AGI The neural-symbolic approach at the heart of Hyperon addresses a key limitation of purely statistical AI, namely that narrow models struggle with tasks requiring multi-step reasoning. Abstract problems bamboozle LLMs with their pure pattern recognition. Throw neural learning into the mix, however, and reasoning becomes smarter and more human. If narrow AI does a good impersonation of a person, neural-symbolic AI does an uncanny one. That being said, it’s important to contextualise neural-symbolic AI. Hyperon’s hybrid design doesn’t mean an AGI breakthrough is imminent. But it represents a promising research direction that explicitly tackles cognitive representation and self-directed learning not relying on statistical pattern matching alone. And in the here and now, this concept isn’t constrained to some big brain whitepaper – it’s out there in the wild and being actively used to create powerful solutions. The LLM isn’t dead – narrow AI will continue to improve – but its days are numbered and its obsolescence inevitable. It’s only a matter of time. First neural-symbolic AI. Then, hopefully, AGI – the final boss of artificial intelligence. Image source: Depositphotos The post OpenCog Hyperon and AGI: Beyond large language models appeared first on AI News. View the full article