Enterprise AI Adoption Moves Beyond Pilot Projects as Companies Focus on Scaling Production Systems

Beyond 'Pilot Purgatory': What does it take to build AI that works?

AI conversations have moved past the point of curiosity. Boards and leadership teams are no longer asking what AI might eventually do. They are asking where it is actually working, what measurable value it is creating - and why so many promising experiments still fail to become durable operating advantages. Across industries, companies have invested heavily in AI pilots, proofs of concept and impressive demos. Yet many remain stuck in what I think of as pilot purgatory: the place where a tool works in a controlled environment but never survives contact with the complexity, exceptions and accountability required in production. The problem usually isn't the model In my experience, AI initiatives rarely fail because the underlying technology is not powerful enough. They fail because of how the technology is applied. A model can be impressive in a sandbox and still be irrelevant to the business if it is not embedded into a real workflow, connected to the right data, governed appropriately and measured against outcomes that matter. That is why access to AI is no longer a differentiator. Anyone can buy access to models or integrate a third-party tool. The real advantage lies in the things that can't be bought off the shelf: proprietary data, deep domain expertise, and the discipline to continuously improve AI once it is operating at scale. For us, those principles come together in our Lean AI approach, rooted in a Lean operating model that drives continuous improvement through testing, learning, and acting. Instead of chasing technology for technology's sake, our Lean AI approach helps us move AI beyond experimentation and into production, where it can improve service, boost productivity and create real business value. Production AI requires discipline, not experimentation for its own sake This is where many organizations get stuck. They treat AI as a portfolio of experiments instead of an operating capability. Organizations that successfully operationalize AI tend to do the opposite. They prioritize AI opportunities based on business value and points of operational friction, identifying manual, repetitive and high-volume work. Then, they build and deploy agents where automation can improve speed, accuracy, scalability or service quality across the entire customer workflow. There is no hobby AI in this model. Every deployment needs a clear business case, a workflow owner, measurement, feedback loops, and a plan to scale. That discipline is especially important with agentic AI, because agents operate with more autonomy than traditional software. Progress is not always linear. Systems improve, encounter new edge cases, retrench and improve again. Human oversight isn't a temporary bridge either; it is part of the architecture. The application layer is where the moat gets built The AI ecosystem is often described in layers, from the underlying IT infrastructure and large language models to the applications built on top of them. Those foundational layers are essential, but they are not where most enterprises will build durable, competitive moats. The real advantage comes at the application layer -- where AI is integrated into workflows, systems, exceptions, data and human judgment that define how a business actually runs. If an enterprise does not own or deeply control that layer, it risks turning AI into another generic capability rather than a competitive advantage. Take supply chain logistics as one example. Moving a single shipment isn't a linear task. It may require coordinating moves by truck and ship and rail, customs documentation in multiple countries, handoffs at multiple facilities, and weather and market conditions that change by the hour. A generic AI tool does not understand that workflow out of the box. Context is the hard part Every industry has its own data and context that powers it. In supply chains, that context lives in historical pricing patterns, warehouse operations, customer-specific policies, shipment characteristics, driver performance, market cycles and the judgment of people who have solved messy freight problems for years. That context cannot simply be purchased. It has to be collected, structured, governed and applied. AI becomes more effective when it's built into your technology platform and can learn from those realities rather than relying on generic information alone. Just as important, employees add institutional knowledge through continuous feedback, teaching AI agents the same way they would train a new operations employee. Take something as seemingly straightforward as scheduling a truck to pick up freight. On the surface, it sounds like a narrow task. In practice, it requires understanding customer requirements, freight characteristics, facility policies, loading dock constraints, appointment systems and exceptions that may vary by location. An AI agent can only automate that work reliably if it has been engineered with the right context and oversight. The companies that get the most from AI will be the ones that move beyond pilots and treat it as an operating model. That means starting with real business problems, owning the application layer where differentiation happens, feeding agents with proprietary context, keeping humans in the loop and measuring outcomes relentlessly. AI will not reward the companies with the most demos. It will reward those that can operationalize learning faster than their competitors. Check out out list of the best cloud backup services. This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Securing adoption in the era of shadow AI

Securing AI adoption without sacrificing governance or innovation Artificial intelligence (AI) is rapidly becoming embedded in the modern workplace, with employees are increasingly turning to AI tools to work more efficiently and boost productivity. This growing demand for faster, more effective ways of working is driving the rise of shadow AI - the use of AI tools outside approved organizational controls and governance frameworks - which results in organizations quickly losing visibility into data usage and potential risks. The scale of this challenge is significant. While 90% of executives are confident in their organizations' visibility into AI tools, just 52% of employees admit to using AI tools without approval, often through personal accounts. As a result, organizations are left grappling with a widening gap between AI adoption and AI governance. The next frontier of AI risk When AI is used without formal oversight, it can bypass governance controls, increasing the risk of errors, regulatory breaches and sensitive data leakages. Organizations are most exposed when AI is already influencing business-critical activities, from customer service and operational decision-making to software development and content creation. The challenge will intensify as businesses move beyond large language models, which generate information, to large action models and agentic systems that can take action. These systems can diagnose issues, recommend actions and execute workflows with minimum human input, increasing both the speed and scale at which mistakes occur. A shadow agent operating outside approved governance frameworks could trigger harmful actions before organizations have the visibility and governance capabilities needed to intervene. There is also a longer-term risk that future AI systems will be trained on synthetic or lower-quality data, weakening performance and decision-making over time. Transparency and traceability will be critical to maintaining accountability, protecting ethical standards and preserving the effectiveness of AI systems as adoption continues to accelerate. AI governance as an enabler What works is AI governance that enables innovation while putting clear guardrails in place that are integrated, transparent, auditable, and aligned with existing risk and compliance frameworks. If AI is to be used safely, firms must be able to successfully identify exactly what went wrong and why when issues arise. In practice, mature governance starts with an approved AI tool stack that provides safe and trusted options for common use cases. This should be supported by risk-based policies that make clear the data being handled, what can and cannot be shared, which tools are permitted, and where human approval is required. Low-risk tasks such as drafting or summarizing content should not be governed in the same way as high-risk uses involving customer data, regulated information or business-critical decisions. Training is equally important. The challenge, beyond only enforcing controls, involves helping employees understand why those controls exist and how to use AI responsibly. As agents increasingly diagnose issues, recommend actions, and execute workflows with minimum input, human oversight and approval processes must scale alongside them. Interoperability will be critical to making this workable at scale, allowing organizations to operate across jurisdictions and multiple AI models without repeatedly rebuilding governance processes and systems from scratch. Making responsible adoption the easy choice For security and compliance leaders, the goal should be to make responsible AI adoption the path of least resistance. Employees turn to shadow AI when approved tools are unavailable, difficult to access or fail to meet their needs. Companies that focus solely on restricting usage risk driving activity further underground and losing out on the efficiency and innovation gains that AI can deliver. Organizations that successfully balance AI productivity and control over their systems recognize that shadow AI use is often a symptom of unmet demand. Employees typically turn to unauthorized tools because they are easier to access, faster to use or better suited to the task at hand. Rather than focusing on restrictions alone, leaders should understand where AI is already being used across the business and ensure approved alternatives are available for the most common use cases. With three-quarters of office professionals saying they would be likely to look for a new job that offered better AI skills development, firms that combine governance with opportunities to build AI skills are likely to see stronger adoption of approved tools and, as a result, less reliance on shadow AI. Building an AI-enabled culture means giving employees the tools, knowledge and confidence to innovate within clear boundaries. By doing so, shadow AI can be reduced while the speed and agility that workers increasingly expect is maintained. The organizations best positioned to succeed The businesses that strike the right balance for AI success will be those that view governance as a foundation for AI adoption and not a barrier to it. By making the secure, approved path the easiest path, shadow AI risk is reduced without sacrificing productivity. Embedding strong governance, supported by trusted and well-managed data foundations, avoids costly mistakes and allows AI to be deployed and scaled with greater safety and confidence. We've reviewed, rated, and ranked the best small business software. This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

The AI pilot scaling playbook: Enterprise-first AI, connected systems, and shared ownership

A Bosch roundtable offered a reckoning with how organisations learn, how they change, and how they decide whether innovation is a performance or a practice in the context of AI. A recently-concluded Bosch Conversations roundtable, The CXO Playbook: AI-led Connected Systems for Future Growth, a Bosch x Economic Times initiative, began with moderator Miloni Bhatt noting that, two to three years ago, global enthusiasm for artificial intelligence (AI) prompted boards to approve major investments, with US companies alone committing $30-40 billion to AI initiatives. She cited MIT's 2025 "State of AI in Business" (also called "The GenAI Divide") report, which found that 95% of enterprise generative-AI pilots delivered no measurable impact on the profit-and-loss statement, and only about 5% scaled beyond the lab to production use. Turning to India, Bhatt referred to industry data from NASSCOM indicating that two-thirds of Indian firms allocate less than 10% of their IT budgets to AI, and only a quarter of those have moved from pilot or demo stage to live production deployment. The real question, she said, was not whether AI could work, but what kind of AI works, where it works, and what organisational discipline is needed for it to work at scale. She emphasised that the small share of successful adopters were not distinguished by pursuing the most advanced or novel applications, but by executing fundamentals well: ensuring clean and well-governed data, clarifying ownership, and actively involving frontline and shop-floor staff in the transformation. Beyond the hype: What leaders say turns AI pilots into business value The shift from pilot projects to real-world impact defines the current shift in enterprise AI. And the organisations reaping the biggest benefits in this shifting landscape are those that have laid foundations for scalable deployment. At a Bosch x Economic Times roundtable, senior leaders from Bosch SDS, Ashok Leyland, Supreme Industries, Raymond, Crompton Greaves, Jio Platforms, Yokohama Off Highway Tires, SKF India, and Metro Brands discussed how to translate AI pilots into enterprise-wide value. They moved past the buzz to emphasise practical priorities: integrated systems, reliable data, clear business ownership, responsible rollout, sustainability, and trust. The consensus was that durable AI success depends on investing in fundamentals. When technology, people, and business goals are aligned, AI projects stop being isolated experiments and transition into measurable, business-changing outcomes. The advantage will go to organisations that link these elements to build practical, trusted AI with long-term impact. Watch the video for full insights from the discussion. In her view, this success hinges on disciplined engineering and operations rather than flashy technology. Bhatt framed the roundtable's agenda around identifying the frictions stalling AI adoption, discussing how to remove them, and outlining the potential payoff. She noted that consultancy forecasts from NASSCOM, BCG, and Accenture project multi‑billion‑dollar growth, particularly in manufacturing, once these barriers are overcome. Introducing the panel, Bhatt began with Rakesh Kumar Murugan, Global Head - Digital Transformation and Industry 4.0, Bosch SDS, whose role involves guiding businesses through precisely this kind of change. She asked him to identify the real "wall" organisations face, whether it is technology, people, or entrenched habits, and to explain what separates companies that manage to scale AI from those that remain stuck in cycles of pilots that never reach production. The real wall: People, process, and problem definitionRakesh Kumar Murugan, Global Head of Digital Transformation and Industry 4.0 at Bosch SDS, opened the substantive discussion by arguing that change management is often treated as the last topic, when it should be the first. "The last topic, whenever you think about it, is change management. That's supposed to be the first topic," he said, making clear that the problem is rarely the software in isolation. According to Murugan, technology may be the visible layer, but the real barrier is organisational. If businesses have not identified the right problem, if they have not contextualised the data, if they have not aligned leadership around a business-first objective, then even the most advanced system will remain inert. Too often, companies blame the tool when the failure lies in the business problem. The deeper question is whether companies have understood the business problem well enough to solve it. Murugan described how, even before the current wave of generative AI, Bosch had already been working on what he called "Semantic Stack", a structure that resembles today's data-product architecture. The logic behind it, he explained, was simple: gather the right data, contextualise it, and ensure that feedback from service stations, customers, and production systems flows into one space rather than remaining fragmented. The larger point is not to impress people with the sophistication of the model, but to make the model useful. "We never felt technology sets you," he said, arguing instead that the real differentiator is whether the company understands its own processes well enough to decide where AI can make a difference. He was particularly emphatic about the fact that many valuable use cases come from the shop floor, even if companies often assume that innovation must always originate at the top. "Shop-floor people may not know the digital black belt, six-sigma term," explaining that does not, however, limit them from identifying where the pain points are. In fact, he suggested, "a good amount of problem statements that can be solved through AI are actually rising from the bottom to the top." That, he said, is why AI adoption must be tied to productivity, transparency, and operational value rather than to abstract enthusiasm. In one case, he said, a customer was facing "30 to 35 days of unplanned downtime" affecting production, and the team had to go "seven to eight levels down" to identify whether the issue lay with machines, operators, or raw materials. The result, he said, was a substantial improvement in production. Elsewhere, he pointed to use cases in energy efficiency, first-pass yield, and maintenance cost reduction, arguing that the strongest return on investment (ROI) appears when AI is deployed against problems that are both visible and expensive. "It's always the return on investment first," he added. The problem of information silosFrom there, the conversation moved naturally to the question of whether companies are solving genuine problems or simply deploying AI in isolated initiatives. A company may have automation in one department, analytics in another, and a chatbot somewhere else, but if those systems do not know each other, they cannot produce intelligence in any meaningful sense. Dimitri Baumtrok, Head of International Sales, CONTACT Software, argued that organisations often do not begin with the most important question, namely what AI means for the business as a whole. Instead, they chase isolated tools, individual team fixes, or the fashionable language of the moment. "We love PowerPoints, we love reading about it, but do we really love generally doing something practical about it," he asked. Companies, Baumtrok suggested, have spent years creating point solutions, systems that function like separate organs or limbs, but have failed to build the nervous system that would allow those parts to work together. The real opportunity, in his view, lies in connecting systems so that they know of each other, rather than merely pushing information around in silos. "You need to build a nervous system and connect the different body parts so that they know of each other," he said, describing AI as a kind of digital reintegration rather than a collection of isolated experiments. The question of silos dominates internal company discussions on AI and workflows. More tools, more models, and more data streams can create the illusion of progress while deepening dependency on disconnected systems. And that is why critics say that the AI era has not abolished organisational fragmentation; in many ways, it has intensified it. Globally, enterprises are discovering that digital maturity cannot be measured by the number of technologies they can name, but by the degree to which those technologies are integrated into a coherent operational logic. In India, where many companies are simultaneously modernising, expanding, and managing legacy infrastructure, this challenge is even more acute. Gopi Sankar, Senior Vice President and Chief Engineer - Global Trucks and Buses, Ashok Leyland, offered a pragmatic view when he said, "We start in silos and then integrate it over a period of time." Not every transformation begins with a perfect system, and not every company can wait for an ideal architecture before beginning. But it is important to remain cognisant of the danger of fragmentation, especially if the goal is to move from small pockets of innovation to something more durable. His observation that there should be "a single brain across silos" captured the aspiration, where departments need not be abolished, but organisations must find a way for them to stop clashing and begin cooperating. Sudhir Kanvinde, Chief Information Officer, Supreme Industries, took that further by warning that fragmented adoption can create "shadow IT" and, worse, "shadow AI" if it is allowed to proliferate without governance. He argued that the first step is not model deployment but visibility: "You need to have visibility of your data." In manufacturing environments with old machines, partial connectivity, and large numbers of moving parts, the first task is not to add intelligence on top of confusion. It is to understand what data exists, where it resides, what it can support, and what gaps must be closed. Kanvinde's message for organisations was clear: AI cannot be a side project owned by the IT team alone. "If we continue with only insights, we'll not be able to go to production," he warned. The decisive shift happens when business users own the use case, when plant heads and operational leaders see that the system can save time and improve performance in ways that matter to them directly. The move from pilot to production is not primarily technical; it is cultural, as it requires organisational readiness. That readiness also depends on ownership. Dr Biswajit Rath, Group Chief Data & AI Officer, Raymond Ltd, then brought the discussion back to first principles. He reminded the room that it is too easy to reduce the AI conversation to large language models (LLMs), small language models (SLMs), and cloud infrastructure, when in fact AI is a much larger family. "We are not talking about the AI family," he said, arguing that machine learning (ML), deep learning, and natural language processing (NLP) remain central to industrial automation and manufacturing. The ethical guardrailsAnita Pansare, CTO and ESG Leader, Crompton Greaves Consumer Electricals, steered the discussion into the terrain of responsibility, product safety, and customer value. She made it clear that in product development, AI cannot simply be thrown in because the industry is excited about it. "It's a stage gate process," she said, explaining that teams must first ask whether the design is technically robust, whether the failure modes have been considered, and whether the feature could affect safety or reliability. She was especially cautious about sectors such as healthcare, where an autonomous decision made by a model can have serious consequences if the system has not been properly validated. At the same time, Pansare insisted that responsible AI is not only about limiting risk, but also about creating practical value. She described how AI-enabled products can help consumers save energy, reduce carbon footprint, and make smarter decisions about electricity usage. In one example, she referred to connected appliances and smart dashboards that tell consumers how much energy they can save, and whether they want to switch something off. "It's at a cost," she acknowledged, returning to a recurring tension in the discussion: the willingness to pay often lags behind the promise of the technology. For farmers, she said, convenience can begin with something as simple as an SMS command to turn a pump on or off. The point is not novelty for its own sake, but practical relief for users who cannot afford to spend time travelling back and forth to the field. She described how weather data and scheduling can be combined so that a pump makes the right decision on behalf of the farmer. "We learn and we say, how about really we give you a smart decision," she said, illustrating how AI becomes meaningful only when it reduces friction in ordinary life. Several speakers also underlined the importance of data privacy, governance, and sovereign AI. Bharkatiya noted that with India's DPDP Act and Europe's GDPR in view, data leakage is a major concern, and enterprises must ensure that sensitive data does not leave organisational boundaries. Murugan added that responsible and sovereign usage of AI will demand more small language models (SLMs) and knowledge repositories hosted on-premise, rather than exposing data to the cloud and large external models. In critical use cases, speakers agreed, human-in-the-loop decision-making remains essential until organisations are fully confident in the system's reliability. The right AI for the enterpriseThat idea resonated with Sanoj Somasundar, Chief Technology Officer India, Director - Technology Development, SKF India, who said the startup ecosystem is increasingly shifting towards "specific data" and "specific needs" that can create competitive leverage and customer value. The broader point, which several speakers reinforced, is that AI is becoming less about generic capability and more about specificity. In other words, the companies that succeed will be the ones that know which tools belong inside the business and which ones do not. Sanjay Bharkatiya, Vice President and Head of Engineering - CIAM, Jio Platforms Limited, highlighted the risks of dependence on a small number of external AI providers. "We are actually dependent on a handful of companies now, as we speak today," he said, pointing to the vulnerability created by concentrated platforms. Businesses, he suggested, need to work on their own use cases and build some of that intelligence in-house. If a company's processes, data, and decisions rely too heavily on outside systems, it may gain convenience but lose resilience. "Localised content, you know, the language of a company has to be in that LLM," said Sankar, noting that a large external model will take longer to adapt to a firm's internal vocabulary. Murugan's warning about cloud dependence was especially pointed. He noted that some companies have spent millions moving data to cloud providers, only to discover that the cost outweighed the benefit. "Business pays for it but business doesn't get any impact out of it," he said. That, in his view, is precisely why the next phase of enterprise AI will be shaped not by broad, indiscriminate deployment, but by careful contextualisation and tighter control over data, language models, and knowledge repositories. Smaller models, he suggested, may be better suited for enterprise manufacturing contexts. The ROI of AIOne of the strongest thematic threads in the roundtable was the tension between the old corporate habit of measuring everything through classic ROI and the newer requirement to judge AI differently. Jitendra Mangave, CIO and CTO, Metro Brands, noted: "AI is a real value and we should have different frameworks to find the value. Traditional ROI methods will never work for AI. That's the biggest learning." The real value of AI may appear in faster decisions, improved trust, less waste, more adoption or better behaviour over time. "AI will succeed when your user will have trust in it and you embed it in their regular journey, in their daily life, in their workloads," added Mangave. The speakers, in different ways, all returned to the same lesson: the future will not be built by those who merely adopt technology, but by those who learn how to organise themselves around it. The road aheadBy the end of the roundtable, a clear consensus had emerged: the decisive advantage will not belong to companies that simply adopt AI, but to those that reorganise themselves around it. The discussion moved repeatedly from the allure of new models to the discipline of use-case selection, from isolated pilots to connected systems, and from generic tools to contextual intelligence that speaks the language of the enterprise. Several themes stood out. First, success depends on solving the right business problem. Murugan's insistence that change management must come first, and that technology is only the visible layer, framed the entire conversation. When AI is tied to concrete operational pain points, including unplanned downtime, energy costs, first-pass yield, and maintenance spend, it ceases to be a demo and becomes a driver of measurable impact. Second, the architecture of AI matters as much as the algorithms. Baumtrok's metaphor of the "nervous system" captured a point that recurred throughout: point solutions and siloed tools create the illusion of progress while deepening fragmentation. The real opportunity lies in connecting systems so they "know of each other", enabling data, decisions, and workflows to flow across departments rather than remaining trapped in isolated stacks. Third, ownership and governance are non-negotiable. Kanvinde's warning about "shadow AI", Rath's emphasis on joint ownership between business and technology, and the repeated calls for human-in-the-loop oversight in critical use cases all pointed to the same conclusion: AI cannot be a side project run by IT alone. It must be embedded in business processes, with clear accountability, visibility of data, and robust governance frameworks. Fourth, the debate on sovereign AI underscored that enterprises must be deliberate about where their data and models live. Bharkatiya's caution about dependence on a handful of external providers, Murugan's argument for on-premise small models and knowledge repositories, and Pansare's stage-gate approach to product safety all reinforced the need for responsible, context-aware AI that respects data boundaries and regulatory requirements. Finally, the ROI discussion broadened the lens beyond traditional financial metrics. Mangave's point that AI's value may appear in faster decisions, greater trust, reduced waste, and improved behaviour over time suggested that organisations need new frameworks to judge success. The payoff is not only in cost savings or revenue growth, but in the capacity to make better decisions, at speed, across the enterprise. Taken together, these insights sketch a playbook for the next phase of AI adoption: choose use cases that matter, connect systems rather than fragment them, build contextual intelligence that understands the business, govern data and models responsibly, and drive adoption through shared ownership across the organisation. The roundtable's message was not that AI is overhyped, but that its promise will be realised only by those who treat it as an organisational challenge, not just a technological one. In Video: Beyond the hype: What leaders say turns AI pilots into business value (This article is generated and published by ET Spotlight team. You can get in touch with them on [email protected])

As AI moves beyond pilots, Dell Technologies sees India entering a new phase of enterprise adoption

Agentic AI, infrastructure modernisation, and growing boardroom interest are driving Indian enterprises from experimentation to execution. India's artificial intelligence (AI) journey is reaching a pivotal moment. After nearly two years of pilots, proofs of concept, and experimentation, enterprises are increasingly looking beyond AI as a technology showcase and towards deploying it at scale across business operations. According to Venkat Sitaram, Senior Director and Country Head, Infrastructure Solutions Group, Dell Technologies India, the clearest indicator of this shift is that organisations are beginning to move AI initiatives into production environments. "When you talk about inflection points, the first thing that comes to my mind is most pilots moving into production. And there are clear proof points. And that's when you can really say that it has reached an inflection point because the adoption rate has gone up," Sitaram said in an interview with Dhruv Mohan of The Economic Times. At the centre of this transition is the rapid emergence of agentic AI, which Sitaram believes is fundamentally changing how enterprises think about productivity, automation and business workflows. The Real AI Challenge Isn't Adoption. It's Scaling It. India's enterprises are no longer asking whether AI works. They're asking how quickly they can scale it.In an exclusive conversation with The Economic Times, Venkat Sitaram, Senior Director and Country Head, Infrastructure Solutions Group, Dell Technologies India, said organisations are increasingly moving AI projects from pilots into production. He explained why agentic AI, infrastructure modernisation, and growing boardroom interest are reshaping enterprise adoption, and why the next challenge is execution, not experimentation.Watch the full podcast. "The biggest catalyst for this is agentic AI because the cognitive work of agentic AI when it starts running into your business workflows has changed dramatically. So it's made analytics easy, it's made coding easy, right? And you don't need so many human interventions," Sitaram said. For Indian enterprises, the challenge is no longer whether AI has potential, but how quickly they can turn that potential into measurable business outcomes. Budgets, skills and prioritisation remain hurdlesDespite growing enthusiasm, enterprises continue to face familiar obstacles in scaling AI deployments. "Budgets are not growing," Sitaram said, describing cost pressures as a universal concern. Alongside funding constraints, he identified skills shortages and poor prioritisation of use cases as the two other major barriers slowing adoption. "Number two, skills. Number three, lack of right prioritisation of use cases sometimes leads to longer experimentation cycles. And that's where we see many of them have not progressed," he explained. However, he believes awareness around these challenges has improved significantly over the past year. Organisations are becoming more disciplined about identifying use cases with measurable outcomes rather than pursuing AI initiatives simply because of market hype. At the same time, the nature of AI deployment itself is changing. Increasingly, inferencing is occurring at the edge, bringing intelligence closer to business users rather than relying entirely on centralised infrastructure. "More and more inferencing is happening on the edge and that has become the business layer. What that would mean is you can have a simple edge device, an agentic edge device, and then start giving the power to that edge user." This shift is accelerating adoption across industries and reinforcing the importance of infrastructure planning. Infrastructure becomes the foundation of AI strategyAs organisations scale AI initiatives, infrastructure is emerging as a critical differentiator. "The adoption of AI is linked to technology infrastructure planning. And the right one size fits all approach will not work. You got to have rightly sized infrastructure," Sitaram said. Dell's strategy reflects this view. The company recently launched PowerStore Elite, which Sitaram described as more than a conventional storage upgrade. "It's not yet another storage array. It's a data platform. It's a modern data platform." The platform incorporates AI-driven operational capabilities designed to automate workload management, optimise performance and streamline recovery processes. According to Sitaram, this allows enterprise IT teams to redirect resources away from routine operational tasks and towards higher-value initiatives. "In PowerStore Elite with the AI ops integration, the IT operations work otherwise in managing workloads, placing the right workloads on performance and the recovery restore operations are all automated, made predictive." He also argued that AI is fundamentally reshaping the economics of enterprise storage as organisations seek to manage larger volumes of data and support increasingly sophisticated workloads. "The economics of storage is changing in an environment. Who is changing the economics of the environment? Storage economics is AI." Future-proofing investments is another concern for enterprises making long-term infrastructure decisions. Sitaram highlighted lifecycle upgrade capabilities that allow organisations to adopt new generations of infrastructure without disruptive migrations. "We're telling customers that start with PowerStore Elite Gen 3 now and later if there is a Gen 4 that comes in after a few years, that can seamlessly coexist with this and you don't need to have any disruption." Agentic AI is raising infrastructure demandsThe rise of agentic AI is also increasing demands on enterprise infrastructure. "AI demands a lot of compute power, highly optimised storage and last but not the least, high bandwidth network links," Sitaram said. In his view, all three foundational layers of enterprise technology, compute, storage, and networking, must evolve simultaneously to support the next wave of AI adoption. Dell's approach combines infrastructure with advisory services and pre-validated architectures intended to help organisations shorten deployment cycles. "Pilot to production is the execution that you talk about. If you want pace, you need something that is qualified, ready, ready to use." The company engages customers through what it calls accelerated workshops, where business leaders and technology teams evaluate use cases, define measurable outcomes and develop implementation roadmaps. "We work with customers by engaging them with what we call an accelerated workshop. And in that workshop, we clearly discuss how and where they could start and what could be the outcomes, measurable outcomes, and use cases that need to be prioritised." These engagements are often followed by visits to Dell briefing centres, where customers can examine real-world deployments, simulations, and reference architectures before committing to large-scale investments. Making the economics workFor many organisations, the business case remains the deciding factor in AI adoption. Sitaram believes enterprises are increasingly evaluating AI through the lens of productivity gains and operational outcomes rather than focusing solely on upfront costs. "The cost of investment therefore becomes something that you can always invest and then take returns which are multi X in a certain period of time." While initial spending can appear substantial, he argues the returns can be transformative. "You're delivering outcomes which are 100X. So on the 100X, a 2X investment in the initial stage may look maybe a little more, but once you see the cycle, then the results are phenomenal." To ease adoption, Dell is also promoting alternative financing models through Dell Apex, including consumption-based and pay-as-you-grow structures. "We sometimes help customers do an Opex modeling of this, consume and pay as you grow models. And with the hand holding, so that is helping them get a quick fast start." Implementation timelines vary depending on architecture and workload requirements, but Sitaram said organisations are often able to deploy AI infrastructure within months rather than years. "We've seen best cases varying anywhere a smaller implementation from 60 days and going up to 180 days." Private cloud and cyber resilience gain importanceAs enterprises modernise their infrastructure, private cloud deployments are becoming increasingly relevant, particularly as AI workloads rely more heavily on containers and distributed architectures. According to Sitaram, enterprises are looking for environments that offer flexibility, control and scalability while avoiding large upfront commitments. "You're not offloading too much in one go. You can buy as you grow." At the same time, cybersecurity concerns are intensifying as AI systems become more sophisticated. "AI adoption has increased the sophistication of cyber attacks. Newer and newer forms of attacks are emerging." For that reason, Sitaram believes cyber resilience must underpin every AI strategy. "While AI is at the core, cyber resiliency should be the foundation layer." AI moves into the boardroomLooking ahead, Dell expects AI adoption to expand well beyond large enterprises and regulated sectors. "We're seeing this catching up with even small, medium enterprises, AI startups." More importantly, AI is increasingly becoming a board-level priority rather than a purely technology-led initiative. "Anywhere and everywhere you see there is a talk and buzz about how I can do and what I can do with AI, it's become board conversations, it's become management meetings." Those conversations increasingly centre on competitive advantage. "Are we leveraging enough? And are we seeing that as a competitive differentiator? Yes." For Dell, which has operated in India for nearly three decades, that shift signals a long-term opportunity. As enterprises move from experimentation to execution, the focus is increasingly turning towards infrastructure readiness, operational outcomes and the ability to scale AI responsibly. The companies that succeed, Sitaram suggests, will be those that can bridge the gap between ambition and deployment, and do so quickly.

From Technical Rollout to Business Mandate: Scaling Trustworthy AI at Enterprise Scale

Enterprise AI is rapidly transitioning from a simple technology experiment to a core operational mandate that dictates modern business survival. To unlock true, enterprise-wide value, organizations must move beyond viewing AI accuracy, governance, and explainability as mere technical checklists and begin treating them as vital enablers of trust, predictability, and business agility. When embedded directly into engineering workflows from day one, robust governance ceases to be a compliance bottleneck and becomes a powerful value accelerator -- allowing teams to innovate with speed while keeping humans firmly accountable for business outcomes. True transformation happens when companies stop managing AI as a standard IT rollout and start leveraging it as a dynamic power source that shapes entire business models. By pairing autonomous capabilities with human oversight, enterprises build the trusted foundation needed to achieve scalable adoption, measurable ROI, and long-term competitive advantage. "As AI increasingly shapes critical enterprise decisions, success belongs to those who align cutting-edge innovation with uncompromising trust and clear human accountability," says Bismi Ravindran, VP & Head of Platform - Engineering at Ascendion. CXOToday: Why are AI accuracy, governance, and explainability no longer just technical concerns but critical business priorities for enterprises today? Bismi: The journey to becoming an AI-powered company, which started as a technical journey, is now rapidly becoming an organizational mandate. AI is increasingly becoming a decision-maker, advisor, and operator embedded into core business processes. AI accuracy, governance, and explainability are not technology goals; they are enablers of fundamental characteristics that enterprises depend on to trust consistency, predictability, accountability, compliance, and scalability. Accuracy creates trust; consistent accuracy leads to confidence. Governance creates consistency and control, ensuring that outcomes are repeatable, auditable, and compliant. Explainability creates accountability transforming AI from a black box to a trusted advisor. What that essentially means is that AI is starting to impact every department, every workflow, and every P&L, so it cannot be owned by just one team or one executive. The companies that are getting this right treat AI as a new power source driving new operating models, rather than a technology rollout. AI is more like electricity than a new CRM system. When organizations start getting this right, it leads to scalable adoption, faster innovation and sustainable business value. CXOToday: How can business leaders strike the right balance between accelerating AI adoption and ensuring responsible, transparent, and accountable AI systems? Bismi: Responsible AI and rapid AI adoption cannot be considered competing priorities. In reality, the ones with strong AI governance foundations are the organizations with the fastest adoption. Organizations slow down not because of a lack of guardrails, but because of a lack of strong vision and direction. A lack of clarity in trust, accountability and ownership plays a large part too. First, we never encourage organizations to start with accelerating AI adoption as the central point of view. If you want to make meaningful progress, the smartest place to start is by assessing opportunities across the key business levers that matter most: revenue growth, cost optimization, risk reduction, customer experience, employee experience, productivity, quality, speed to market, and innovation. Align the C-suite internally so that it is equally accountable for the outcome, then redesign that workflow end-to-end with AI built in. Second, there is no question of balance. Responsibility and accountability must be an integral part of enterprise success when building AI systems. We have our own proprietary software, AAVA™, which introduces golden agents with built-in governance, certification, and protocol for enterprise-scale software engineering. This works alongside the human interface, where we gauge exactly where and how much value AI is truly adding, or identify deviations that make AI risky. The key differentiator between the success of one enterprise and another lies in how we leverage the carbon-plus-silicon partnership to harness the full potential of the technology. Last but not least, demand the 50/50 value proposition. Find a partner willing to own the impact AI delivers and pay for business impact, not effort. Cost has been, and will remain, part of the economic model. However, the leaders pulling ahead are structuring new engagements to unlock capital, drive innovation, and pay down years of accumulated technical and process debt. If you are getting brain surgery, you are not paying the surgeon for the hours they spend in the operating room; you are paying for decades of training, mentorship, and practice that improve your personal outcome. The same logic applies to AI-powered technical operators. Do not box yourself out of upside value with rate cards designed for a pre-AI world. CXOToday: How can leaders ensure governance is a value accelerator rather than a compliance bottleneck? Bismi: Governance becomes a bottleneck only when it is treated as a gate at the end of the process. The most successful organizations treat governance as an accelerator embedded into the AI lifecycle from the beginning. Governance should be about creating the trust, consistency, and predictability required to scale AI faster and with greater confidence. It should not be about slowing teams down with reviews, approvals, and compliance checklists. At an individual level, we see how easily AI content can manipulate decision-making, while at the enterprise level, the defining question is whether your clients' customers actually trust the AI powering your products. Overcoming this gap requires prioritizing safety, governance, and the race to data intimacy. Why? Because embedding transparent guardrails, automated risk tracking, and clear human-in-the-loop accountability directly into engineering workflows prevents large-scale failures without slowing down innovation velocity. When trust is secured at the foundation, speed and value naturally follow. CXOToday: When AI systems move towards autonomous actions, who should be made accountable for outcomes? Bismi: When AI systems move toward autonomous action, high autonomy without high accountability is a breakdown in enterprise leadership. While autonomy and accountability might look like a dichotomy, they are two sides of the same coin: AI executes at speeds humans cannot match, but humans must remain the trusted force behind every enterprise decision. Accountability must remain firmly with the humans and leaders who define the objectives, establish the guardrails, approve the boundaries, and oversee the outcomes. We already see this tension in modern IT operations, where siloed teams, fragmented systems, and reactive monitoring slow down resolution when automated changes fail. The fix is straightforward. Connect real-user monitoring directly to change tracking, deploy agentic AI for complex diagnostics, and build a blameless culture where teams share context instead of shifting blame. As organizations progress on their AI maturity journey, they should clearly define decision rights and accountability boundaries. AI can own actions, but leadership must own outcomes because only executive leadership has the authority to break down these organizational silos, fund integrated guardrails, and foster the culture required to safely operate autonomous AI at scale. CXOToday: How important is data quality and governance to achieve AI accuracy? Bismi: "Garbage in, garbage out" is an immutable law, not just in computing, but in life. Just as the quality of the food we eat or the health habits, we follow dictates our long-term well-being, high-quality data is the absolute top requirement for Generative AI success. However, data quality cannot exist without data governance. Governance establishes the foundational rules, policies, and ownership that dictate how data is collected, structured, updated, and secured across its lifecycle. Without active data governance, organizations quickly fall into the trap of poor data integrity, multiplying security vulnerabilities, and unchecked bias creeping into AI models. In our endeavour to achieve AI accuracy, these principles are critical. AI is increasingly leveraged to build modern data foundations, embedding intelligence across the data lifecycle to rapidly and securely maximize "return on data." Clients are already realizing significant value: our engineers use tools like AAVA to turn raw data into actionable insights, improving customer experiences, business decision accuracy, and operational velocity. In the data-driven battlefield of modern business, harnessing the transformative power of AI with humans firmly "in the loop" empowers organizations to unlock data's full potential ethically and safely, dramatically improving both speed and cost efficiency. Accurate data drives trust, governance ensures compliance and accountability, secure access protects enterprise assets, and standardized definitions create the consistency and scalability required for AI to deliver value at scale. CXOToday: How should organizations measure whether their AI systems are truly trustworthy, beyond traditional metrics like accuracy? Bismi: We're in the middle of an S-curve shift. That is when an old system is collapsing and the new one hasn't yet stabilized. Humanity's been here before: fire, electricity, the internet, the cloud. Each new source of economic power sparked fear before it fuelled growth. This time, particularly for knowledge workers, it's personal. This is coming for our minds. As an organization, we have spent years living at the intersection of AI hype and hard reality. And this is what we have learnt. Accuracy builds confidence in an answer; trustworthiness builds confidence in adoption. Enterprises should measure not only whether AI is right, but whether it is explainable, fair, secure, accountable, and consistently delivering business value. We engineer these principles into the design the core of what we describe as Engineering to the Power of AI. This is our method and platform to pair humans with AI. Every agent we deploy includes identity anchoring, context-aware authorization, immutable audit trails, and dynamic trust scoring to monitor performance. To measure and ensure our systems are truly trustworthy beyond traditional metrics like accuracy, four things will build trust in the AI valley: * Business Impact: Prove AI arbitrage by measuring and showing real-world velocity, cost, and impact gains * Human Accountability: Percentage of high-impact decisions with human oversight and ownership. Keep humans at the wheel to own the "why" while AI handles the "how." * Adoption and behavior changes: Celebrate the humans using AI through rituals like Agent-a-thons to normalize and humanize the journey. Measure employee usage, repeatability, and willingness to rely on agentic AI.

7 Reasons Enterprise AI Projects Fail After the Pilot Stage

Enterprise AI has moved beyond experimentation. Over the past two years, organisations have invested heavily in AI pilots across customer service, operations, software development, finance, and supply chains. Yet despite this momentum, very few initiatives successfully make the transition from pilot to enterprise-wide deployment. The challenge is no longer whether AI works. In most cases, it does. The bigger question is whether organisations are prepared to integrate AI into the realities of day-to-day business operations. Recent industry research puts a hard number on the scale of the problem: roughly 80% of enterprise AI initiatives fail to deliver their intended business value. Nearly a third are abandoned before they ever reach production. This is not a model performance issue. It is an operationalisation issue. Having spent over 25 years across banking, financial services, and enterprise technology, and now working with enterprises deploying AI in regulated environments, I have seen the same seven failure patterns repeat with striking consistency. * Pilots are built on ideal conditions Most pilots are developed using clean, structured, and carefully prepared datasets. Production environments are very different. Data is fragmented, inconsistent, and spread across multiple systems. As projects move into real-world environments, data quality quickly becomes one of the biggest obstacles to scaling AI successfully. * Success is never clearly defined Many AI initiatives begin with enthusiasm but without clearly agreed business outcomes. Teams focus on building a working model instead of defining measurable objectives such as cost reduction, productivity improvement, turnaround time, or customer experience. Without clear success metrics, it becomes difficult to justify further investment once the pilot ends. * Legacy technology slows enterprise AI Most organisations continue to operate on technology that was never designed for AI-driven workflows. Legacy applications, disconnected systems, and lengthy deployment cycles make it difficult to integrate AI into everyday business processes. The technology may be ready, but the underlying enterprise architecture often is not. This is precisely where I see the most enterprises stall. The instinct is to treat legacy modernisation and AI adoption as sequential projects: rebuild the core first, add AI later. That sequencing is expensive and slow. The organisations that move faster are the ones that build a configuration and orchestration layer over what already exists, rather than waiting to replace it. * Organisations underestimate the change required Technology alone rarely determines the success of an AI initiative. Scaling AI requires collaboration between business teams, technology teams, compliance, security, and operations. A pilot may succeed with a small project team, but enterprise-wide adoption demands organisational alignment that many businesses underestimate. * Operational readiness receives little attention Building an AI model is only one part of the journey. Production environments require governance, monitoring, security, model updates, and performance management. Without these operational capabilities, many promising pilots fail to deliver sustainable business value after deployment. This is where architecture matters more than ambition. The enterprises that succeed treat governance, auditability, and human approval as design principles built into the system from day one, not compliance work added after a pilot succeeds. AI should be confined to design-time configuration, with deterministic, auditable systems handling execution in production. That separation is what makes operational readiness achievable rather than aspirational. * Expectations are higher than reality Many organisations expect AI to deliver transformational results within a single planning cycle. In reality, enterprise AI creates value over time through continuous improvement and refinement. Unrealistic expectations often lead organisations to abandon initiatives before meaningful outcomes are achieved. The most effective way I have seen organisations manage this is by starting small and shipping fast. A single, well-defined workflow or dashboard delivered in days builds more organisational confidence than a twelve-month transformation roadmap. Speed early in the journey buys the patience needed for the larger initiatives that follow. * Executive sponsorship does not last long enough Enterprise AI programmes often span multiple budget cycles. When leadership priorities shift, projects lose momentum, funding, and organisational focus. Sustained executive sponsorship is often the difference between isolated pilots and enterprise-wide transformation. Sponsorship rarely fades because leaders stop believing in AI. It fades because the initiative stops producing visible wins. Programmes that deliver measurable outcomes every few weeks, rather than promising one large outcome a year from now, keep sponsorship alive because the business case renews itself continuously. The next phase of AI is execution The conversation around enterprise AI is changing. For the last few years, organisations have focused on adopting AI. The next phase will focus on delivering measurable business outcomes. Competitive advantage will not come from simply deploying more AI models. It will come from integrating AI into business processes, modernising enterprise technology without ripping it out, strengthening governance as a design principle, and enabling faster execution across the organisation. The organisations that succeed will not necessarily be the ones investing the most in AI. They will be the ones with the right architecture: AI used responsibly at the design stage, deterministic and auditable systems in execution, and the discipline to measure business outcomes rather than pilot activity. That is what separates enterprises still explaining their AI strategy from those already running on it. Shrish Anand Lal is the Executive Director and Chief Business Officer at New Street, the company behind MiFiX.ai. With over 25 years of experience across banking, financial services, and enterprise technology, he leads enterprise growth, business development, and strategic partnerships globally, and the views expressed in this article are his own

Beyond the AI Pilot Trap: Industrializing Enterprise AI for Measurable Value

Scaling enterprise AI is no longer a challenge of technology, but of strategy, engineering, and organizational readiness. While proofs of concept demonstrate potential in controlled environments, moving beyond the pilot trap requires a shift from curiosity-led experimentation to a value-realization mindset. True transformation happens when organizations treat AI not as a collection of isolated innovation projects, but as a core enterprise capability built on trusted data, scalable engineering, responsible governance, and clear business ownership. As agentic systems and autonomous workflows reshape the corporate landscape, the winners will be the organizations that modernize their underlying architectures and align cross-functional teams around shared outcomes. Moving from proof to production demands robust platforms, continuously governed data, and a workforce empowered to reimagine how the business creates value. Bridging this gap between experimentation and lasting impact is central to the strategy outlined by Mayank Verma, Global Head of Data and AI at Xebia. CXOToday: Why do enterprises continue to struggle with scaling AI beyond successful pilots? Mayank: Enterprises rarely struggle because the pilot failed; they struggle because the pilot was never designed for enterprise scale. Proof of concept can demonstrate potential in a controlled environment, but production AI must perform reliably across real users, real data, regulatory constraints, and changing business conditions. The common blockers are clear: unclear ownership, fragmented data, limited engineering discipline, and governance that arrives too late. Many pilots are also led by technology curiosity rather than measurable business value, which makes it difficult to justify investment once the excitement fades. AI also does not scale like traditional software. It requires continuous monitoring, model improvement, data quality management, security controls, and human oversight. Without these foundations, even a promising pilot becomes difficult to industrialize. At Xebia, we encourage clients to move from an experimentation mindset to a value-realization mindset. The question should not be, "Where can we use AI?" It should be, "Which business outcomes can AI improve, and what operating model will help us scale them responsibly?" CXOToday: What is the 'AI pilot trap,' and what are the biggest barriers to enterprise-wide AI adoption? Mayank: The AI pilot trap is the gap between proof and production. It happens when organizations celebrate a successful demonstration but do not build the enterprise capabilities needed to repeat, govern, secure, and scale that success across business functions. The barriers are usually structural, not technical: disconnected data estates, unclear decision rights, legacy systems, limited platform maturity, and weak linkage to commercial outcomes. Too often, organizations start with "How do we use Generative AI?" when the sharper question is "Where can AI create measurable value?" As AI becomes more autonomous through agentic systems, these challenges become even more pronounced because AI is only as reliable as the quality of the data and processes behind it. The organizations that are scaling successfully are treating AI as an enterprise capability not a collection of isolated innovation projects. They are investing equally in data, engineering, governance, and organizational change. CXOToday: What does an effective AI operating model look like? Mayank: An effective AI operating model starts with business outcomes, not technology choices. AI shouldn't sit in an innovation lab it should become part of how the business operates. The model rests on four pillars: business ownership, trusted data, AI engineering, and governance. Business teams define the outcomes, data provides the foundation, engineering industrializes AI through scalable platforms, and governance ensures security, compliance, and responsible AI. Equally important is collaboration. AI can no longer be owned solely by IT. It requires business leaders, data teams, engineers, and governance functions to work together from the outset. At Xebia, we see AI engineering as the bridge between experimentation and enterprise value. When organizations combine strong engineering discipline with business alignment and trusted data, AI moves beyond pilots to become a scalable business capability. CXOToday: How can organizations better align data, engineering, governance, and business teams? Mayank: The best alignment mechanism is a shared business outcome. When teams are aligned around revenue growth, customer experience, productivity, risk reduction, or operational resilience, AI stops being a handoff between departments and becomes a joint transformation agenda. Business teams should define the problem and success metrics; data teams should ensure trusted, contextual information; engineering teams should build scalable and reusable AI platforms; and governance teams should shape the guardrails from day one rather than reviewing risk after deployment. Shared accountability is critical. Cross-functional AI product teams with common KPIs consistently outperform siloed approaches because AI systems continue to evolve after launch and require continuous business, data, engineering, and risk input. Ultimately, trusted data, engineering excellence, and governance are not independent functions they are interconnected capabilities that determine whether AI creates sustainable business value. CXOToday: How is AI reshaping infrastructure and cloud investment decisions? Mayank: AI is fundamentally changing how enterprises think about infrastructure. The conversation is no longer cloud-first, it's outcome-first. Enterprises are now evaluating hybrid cloud, sovereign AI, edge computing, GPU infrastructure, and modular architectures through the lens of workload sensitivity, latency, regulatory requirements, cost, and business criticality. As infrastructure decisions should follow the value case not the other way around. Generative AI and agentic AI also demand greater flexibility. Enterprises need architecture that allow workloads to move seamlessly across cloud, on-premises, and edge environments while maintaining governance and security. The winners will be organizations that avoid lock-in, design for portability, and build cloud-agnostic foundations that can adapt as models, regulatory expectations, and AI workload patterns continue to change. CXOToday: What infrastructure is required to support Generative AI and Agentic AI workloads? Mayank: Generative AI and Agentic AI require more than GPU capacity. They require an AI-ready enterprise fabric that connects compute, data, security, orchestration, observability, governance, and engineering discipline. For Generative AI, enterprises need scalable compute, modern data platforms, vector databases, orchestration frameworks, prompt and model management, and robust security. Agentic AI raises the bar because autonomous agents interact with multiple systems, trigger actions, and make decisions that require stronger control, traceability, and human oversight. The right infrastructure depends on the business use case. A customer-facing AI assistant, an internal knowledge system, a regulated decisioning workflow, and an edge-based industrial use case will each demand different choices across public cloud, private cloud, sovereign environments, and hybrid deployments. Equally important is designing modular architectures that can evolve as AI technologies mature. The pace of innovation is unprecedented, and enterprises need platforms that allow them to adopt new models, tools, and deployment strategies without rebuilding their entire technology landscape. CXOToday: How are enterprise data architectures evolving in the age of AI? Mayank: AI is fundamentally changing what enterprises consider valuable data. Traditionally, organizations focus on structured information stored in data warehouses and transactional systems. Today, much of an organization's intelligence resides in unstructured assets such as documents, policies, emails, technical manuals, and knowledge repositories. The challenge is no longer the volume of data but its quality, accessibility, and context. Many enterprises have abundant data but lack the trusted, governed information that AI systems need to generate reliable outcomes. As agentic AI becomes more prevalent, this becomes even more critical because autonomous systems rely on contextual enterprise knowledge to make informed decisions. Modern data architecture therefore needs to unify structured and unstructured information through strong metadata management, semantic layers, knowledge graphs, vector search, governance, and AI-ready pipelines. Organizations that can turn enterprise knowledge into trusted, contextual intelligence will have a clear advantage in scaling AI. CXOToday: Is workforce readiness becoming more important than technology investments? Mayank: Workforce readiness is becoming the decisive factor. Technology is increasingly accessible, but the ability to redesign processes, make informed decisions, and adopt AI responsibly depends on people. AI transformation requires employees to understand where AI can support their work, where human judgement remains essential, and how processes should change when intelligence becomes embedded into everyday workflows. That requires continuous learning, experimentation, and leadership commitment. Equally important is creating multidisciplinary teams where domain experts, engineers, data scientists, and business leaders work together to solve real business problems. AI delivers its greatest value when technical expertise is combined with deep industry knowledge. Ultimately, AI doesn't replace human expertise, it amplifies it. Organizations that invest in both technology and talent will create a far more sustainable competitive advantage than those focusing on technology alone. CXOToday: What AI skills will enterprises need over the next two to three years? Mayank: The demand is shifting from isolated AI expertise to integrated AI engineering capabilities. While data science remains important, enterprises increasingly need professionals who can build, deploy, govern, and continuously improve AI systems at scale. Skills such as AI engineering, platform engineering, MLOps, LLMOps, data engineering, AI security, and responsible AI will become foundational. At the same time, organizations will need business leaders who understand how to translate AI capabilities into measurable commercial outcomes. Perhaps the most valuable skill will be the ability to combine domain expertise with AI. Technology alone doesn't create competitive advantage it's the ability to apply AI to solve industry-specific challenges that drives real business value. The future workforce will therefore be defined less by job titles and more by the ability to collaborate across technology, business, and engineering disciplines. CXOToday: How should organizations approach AI governance, responsible AI, and risk management? Mayank: Responsible AI should not be viewed as a compliance exercise it should be treated as a business capability. As AI becomes embedded in critical enterprise decisions, governance becomes essential for building trust, ensuring regulatory compliance, and managing operational risk. Governance needs to be built into every stage of the AI lifecycle from data quality and model development to deployment, monitoring, and continuous improvement. Waiting until after deployment to address governance often creates unnecessary risk and slows adoption. As organizations adopt autonomous AI systems, strong governance becomes even more important. Trusted data, Zero Trust architectures, explainability, human oversight, and clear accountability are all essential for ensuring AI operates safely and responsibly. As autonomous systems rely entirely on trusted data and trusted pipelines, making governance a non-negotiable foundation for enterprise AI. The organizations that will lead in AI are not those taking the greatest risks they are the ones building innovation on a foundation of trust, transparency, and responsible engineering. CXOToday: Which industries are leading AI industrialization, and what can others learn from them? Mayank: The strongest AI industrialization is happening in sectors where AI is tied directly to measurable business outcomes. Financial services, manufacturing, healthcare, retail, and telecommunications are leading because they are using AI for fraud prevention, predictive maintenance, intelligent customer engagement, clinical and operational decision support, and supply chain optimization. What differentiates these organizations is that they don't treat AI as a standalone technology initiative. Instead, they integrate AI into core business processes and build the engineering, governance, and data capabilities required to sustain it at scale. Another common characteristic is their willingness to modernize legacy systems rather than simply layering AI on top of existing infrastructure. They understand that AI delivers its greatest value when supported by modern data platforms, scalable engineering practices, and cross-functional collaboration. The lesson for every enterprise is clear successful AI transformation is not about deploying the most advanced model. It is about selecting high-impact use cases, modernizing the underlying data and technology foundation, building responsible guardrails, and scaling what works with discipline. CXOToday: What differentiates organizations that are successfully scaling AI across the enterprise? Mayank: Organizations that scale AI successfully share one mindset: they treat AI as a business transformation capability, not a series of disconnected technology experiments. First, they begin with a clear business objective. Every AI initiative is linked to measurable outcomes, whether that's improving customer experience, increasing productivity, reducing costs, or accelerating innovation. Second, they invest in the fundamentals. Trusted data, AI engineering, governance, and modern platforms are treated as strategic capabilities rather than project-level requirements. Third, they foster collaboration across business, technology, and governance teams. AI doesn't succeed in silos. It requires shared accountability, continuous iteration, and executive sponsorship. Most importantly, they build for change. The AI landscape is evolving quickly, so leading organizations design flexible, modular ecosystems that allow them to adopt new models, tools, and deployment patterns without disrupting existing operations. Ultimately, organizations that scale AI successfully don't chase every new technology trend -- they build the organizational capabilities needed to continuously innovate. CXOToday: What is the growing role of AI engineering in enterprise transformation? Mayank: AI engineering is becoming the foundation of enterprise AI transformation. While Generative AI has made AI more accessible, enterprise success depends on the ability to operationalize it reliably, securely, and at scale. A model is only one component of enterprise AI. Organizations also need robust data pipelines, platform engineering, model orchestration, observability, evaluation frameworks, security, and continuous monitoring. AI engineering brings these disciplines together, so AI systems perform consistently in real business environments. As AI evolves toward autonomous and agentic systems, engineering becomes even more critical. Enterprises need platforms that can manage multiple models, integrate with business applications, enforce policy controls, capture feedback, and detect drift or unintended behavior. At Xebia, we see AI engineering as the bridge between innovation and business value. It enables organizations to move beyond isolated pilots and build AI solutions that are scalable, resilient, and aligned with enterprise objectives. In the years ahead, AI engineering will become as fundamental to business transformation as software engineering has been over the last two decades. CXOToday: Based on Xebia's experience, what are some practical best practices for enterprises looking to scale AI? Mayank: The first best practice is to anchor every AI initiative in a clear business outcome. Organizations that define the problem, success metric, owner, and path to adoption upfront are far more likely to move from experimentation to measurable impact. Second, prioritize data readiness. AI can only generate reliable outcomes when it is built on trusted, well-governed, and context-rich data. In many enterprises, preparing the right data is more challenging and more valuable than choosing the model itself. Third, invest early in AI engineering and governance. Scalable platforms, reusable components, evaluation mechanisms, security controls, and responsible AI frameworks create the consistency needed to expand AI across business functions. Finally, organizations should adopt an iterative approach. AI transformation is not a one-time implementation but a continuous journey of learning, refinement, and improvement. Those that build flexibility into their architecture and operating model are far better positioned to adapt as AI technologies continue to evolve. CXOToday: What is one key piece of advice for CIOs and business leaders looking to move beyond AI pilots in 2026? Mayank: My advice would be simple: don't think incrementally think transformational. The conversation should no longer be about launching another AI pilot. It should be about reimagining how the organization will operate in an AI-first world, how it serves customers, empowers employees, improves decisions, and creates new sources of value. That means asking harder questions: Is the data trusted? Is the architecture flexible? Are governance and security built in? Do teams have the skills to adopt AI? And, most importantly, is every AI investment connected to a business outcome leaders are willing to own? Technology will continue to evolve rapidly, but organizations that build strong foundations trusted data, modern engineering, responsible governance, and a business-first AI strategy will be best positioned to capture long-term value. The goal is not simply to automate existing processes. It is to reimagine how the enterprise creates value in an AI-driven world. Organizations that build that foundation now will be better positioned to define the next decade of digital transformation.

Navigating Enterprise AI: Insights on Governance, Agentic AI, and Top-Line Growth

The rush to deploy enterprise AI has left many organizations trapped in an expensive cycle of experimentation, delivering impressive laboratory demos but negligible P&L impact. Crossing the divide from speculative pilots to scalable, production-grade systems requires a fundamental shift in priorities: establishing unified, well-governed data foundations long before chasing ambitious use cases. From navigating the guardrails of Agentic AI to redefining marketing ROI through rapid experimentation, the true value of artificial intelligence lies not in automating legacy workflows, but in committing to top-line growth and systemic process redesign. In this comprehensive Q&A, industry leaders dissect the real-world mechanics of scaling AI responsibly, drawing vital governance lessons from heavily regulated sectors like BFSI to establish lineage, auditability, and trust from day one. Crucially, the discussion shifts the strategic focus for CXOs away from simple cost reduction toward unlocking entirely new revenue opportunities and accelerating execution speed across the enterprise. Offering a pragmatic roadmap for executive leadership, these critical insights on bridging the gap between technological potential and measurable business value are delivered by Raghvendra Kushwah, Co-founder of Eucloid Data Solutions. CXOToday: Is enterprise AI moving beyond experimentation to measurable business outcomes? Raghvendra: Selectively, yes, but the honest picture is that most enterprises are still stuck between the two. Industry research suggests the vast majority of generative AI pilots deliver no measurable P&L impact, and that matches what we see on the ground. The organizations that are crossing over share a pattern. They invested in their data foundation before chasing use cases, they picked problems with genuine scale, and they changed their processes to absorb AI rather than bolting it onto workflows designed for a different era. One of our healthcare clients in the US spent close to a year consolidating more than 65 source systems onto a unified platform before pursuing AI seriously. Their first AI deployments followed almost immediately after, at a fraction of traditional build costs. The outcomes are measurable when the sequencing is right. Experimentation without that groundwork stays experimentation. CXOToday: What role does data quality and governance play in scaling AI successfully? Raghvendra: It is the difference between a demo and a production system. AI is only as good as the data layer beneath it, and fragmented, inconsistent data produces confident, wrong answers faster than any human analyst could. But I would push the point further: governance is where promising pilots most often die. A pilot can impress everyone in the lab and still fail at the production gate, when security and compliance teams ask entirely reasonable questions about how the system handles sensitive data, what the model saw, and who can trace it. If lineage, access controls, and auditability are built into the foundation from day one, those questions have ready answers. Bolted on afterwards, they become the wall the project never gets over. This is why we advise clients to treat governance as an enabler of scale, not a tax on it. CXOToday: How are enterprises approaching Agentic AI, and where are they seeing the biggest impact? Raghvendra: With more enthusiasm than discipline, in many cases. The appeal of agents is obvious, with systems that plan and execute multi-step work rather than answering single questions. But the fundamental constraint has not changed. AI today is not at a stage where it runs autonomously, and agents amplify both the capability and the risk. The impact we see is real but bounded: agents accelerating data engineering workflows, handling multi-step analytics requests, and compressing work that used to take days into hours, always with a human in the loop for decisions that matter. The organizations getting value are the ones that deploy agents where the cost of an error is low or recoverable, and keep humans on the approval path where it is not. The ones getting burned expected autonomy and planned no oversight. My advice is to treat agentic AI as a force multiplier for your teams, not a replacement for them. CXOToday: How can organisations improve marketing ROI using AI-driven analytics and customer insights? Raghvendra: The starting point is the data foundation. Marketing is one of the most fragmented functions in the enterprise, with customer information spread across CRM, web analytics, campaign platforms, and commerce tools, each holding a different view of the same customer. Without unifying that, AI-driven insights are little more than better-presented guesswork. Once that foundation is trusted, the familiar gains follow, be it sharper segmentation and propensity modelling, or attribution that goes beyond last-click. The bigger shift, though, is pace. The constraint has rarely been a shortage of ideas, it has been the wait between asking a question and getting an answer. When analytics sits on a governed data layer that can be queried conversationally, the marketer with the question can answer it in minutes rather than waiting days for an analyst. That compression changes behaviour more than any single model does. Teams test more variants and correct underperforming spend while a campaign is still running, instead of discovering it in a quarterly review. The discipline that decides whether any of it counts is measurement. Organisations need to define what improved ROI actually means, and instrument for it, before deploying AI. Marketing has a long history of adopting new tools without changing the processes around them, and AI will deliver the same disappointing results unless the workflows and decision-making practices evolve with it. CXOToday: What lessons from the BFSI sector can help other industries adopt AI more responsibly? Raghvendra: BFSI offers a useful lesson precisely because it has always had to innovate within tighter constraints than most industries. Banks and financial institutions already operate within established governance, security and risk frameworks, shaped by requirements such as RBI's IT and cybersecurity directions in India, Basel Committee standards, PCI DSS for payment data, and applicable data-protection laws. AI needs a similar level of restraint. Organizations need clear rules around what data can be used, who can access it, how decisions are explained, where human oversight is required, and what happens when a system gets something wrong. The lesson from BFSI is that governance should not be introduced after AI has been deployed. It should be built into the system from the beginning. Strong guardrails do not necessarily slow innovation down. They give organizations the confidence to deploy AI at scale because the risks, responsibilities, and boundaries are clearly understood. CXOToday: What are the biggest AI priorities CXOs should focus on over the next few years? Raghvendra: Three things. First, aim at the top line. Most AI budgets are still justified on cost reduction, and that is the smaller half of the opportunity. The use cases worth funding are the ones that help you sell more: faster product and content generation, new offerings that were never economical to build manually, better conversion across the customer journey. Growth use cases are also more forgiving of imperfection, which makes them the better place to build organisational confidence. Second, be realistic about what technology actually does to cost. It rarely takes cost out. What it does is make a capability cheap enough that people use far more of it than they did before, and that wider adoption is where the return sits. If a marketing team could afford one campaign variant and can now afford ten, the outcome is more revenue, not a smaller team. Leaders who go in expecting the cost line to fall usually end up disappointed on both counts, because they neither reduce spend nor pursue the growth that was available. Third, and this is the most underestimated: take the leap of faith on process change. Enterprises typically run a pilot inside existing workflows, wait for it to demonstrate value, and plan to redesign the process only once the results are convincing. The difficulty is that the value usually depends on those very changes. Tested in an environment that was never adapted for it, the pilot underperforms and the technology takes the blame, even though the experiment was constrained from the start. That commitment has to come from the top, and it has to reach the middle of the organisation, where transformation actually stalls. The technology is ready enough. The appetite to redesign how work gets done is usually what is missing.