Why AI projects fail inside large enterprises and what leaders are learning: Puneet Asthana, ED & CTO, Shriram Group’s Capital Market Business
Enterprises across sectors are investing heavily in AI platforms, large language models, copilots, and intelligent automation. Yet despite the enthusiasm, many AI initiatives continue to struggle when moving from pilot projects to enterprise-wide deployment.
Puneet Asthana, ED & CTO, Shriram Group’s Capital Market Business (Shriram Wealth AMC | Broking | Research), says the reasons behind these failures are often far less technical than most organisations assume.
In an exclusive interaction with Express Computer, Asthana argues that successful AI adoption depends less on algorithms and more on fundamentals such as data discipline, governance, business ownership, and organisational readiness. As wealth management and capital markets firms increasingly embrace AI-driven experiences, he believes the winners will not necessarily be those with the most sophisticated models but those with the strongest foundations beneath them.
The scaling challenge begins after the pilot succeeds
Most organisations today are able to demonstrate promising AI pilots. Small teams identify a business problem, collaborate with technology vendors, and produce encouraging results. However, the real challenge emerges when organisations attempt to scale those pilots across business functions.
Many enterprises underestimate the operational complexity involved in scaling AI.
“Most enterprises do well with the pilot project. A small team gets together, picks a specific problem, works closely with a vendor, and the results look promising. But when it comes to scaling, that’s when the real challenges start showing up,” he says.
Asthana points to three recurring reasons behind failed AI initiatives: poor data quality, lack of business ownership, and governance frameworks that arrive too late in the process.
The first challenge is what he describes as “data debt”, organisations attempting to build intelligent systems on top of fragmented and inconsistent data environments. “You can’t build intelligent systems on top of messy data foundations. Organisations jump to the AI layer before they’ve done the plumbing. The AI model is only as good as what you feed it.”
The second challenge is ownership. AI projects often remain confined within technology departments instead of being championed by the business functions that ultimately benefit from them. “When AI is a technology team’s project and not a business team’s project, then the adoption dies. The RM, the fund manager, and the ops head have to own it. Technology is just the engine room,” he adds.
Governance represents the third major obstacle. In regulated sectors such as capital markets, compliance and risk requirements cannot be treated as post-implementation considerations. “Especially in regulated industries like capital markets, if audit trails and explainability aren’t built in from day one, you’re rebuilding later and delaying the rollout or ending up killing the project altogether.”
For Asthana, the lesson is straightforward: organisations often focus on AI models before addressing the fundamentals. “The organisations pulling ahead aren’t the ones with the best AI. They’re the ones with the best data discipline and business ownership underneath it,” he adds.
Technology is becoming central to business strategy
As financial institutions modernise their operations, technology leaders are assuming increasingly strategic responsibilities.
Asthana is cautious about the popular narrative that banks and wealth firms must become technology companies. Instead, he believes financial institutions must learn to use technology intelligently while remaining focused on their core responsibilities. However, he acknowledges that the role of technology leadership has evolved dramatically.
Today’s CTOs are no longer limited to infrastructure management and operational reliability. They are participating in product design, customer experience initiatives, distribution strategy, and growth discussions.
“The CTO role has moved from ‘keep the lights on’ to ‘help us grow AUM.’ That’s a genuine shift, and it comes with accountability that didn’t exist before.”
At the same time, client expectations are reshaping technology priorities. Wealth management customers increasingly compare financial experiences with the seamless digital journeys they encounter in consumer platforms.
Yet unlike consumer technology companies, financial institutions must balance innovation with regulatory obligations. “A fintech can move fast and break things. A regulated wealth company cannot. A broken trade execution or a compliance gap in a wealth platform is a regulatory event, a client trust issue, or a reputational problem,” he points out.
This has given rise to what Asthana describes as a “trust architecture”, an operating model that combines agility with auditability, transparency, and regulatory readiness.
AI success depends on organisational readiness
While discussions around AI often focus on models and platforms, Asthana believes the real barriers to adoption lie elsewhere. The most significant challenge remains data readiness. “Every organisation I speak with has an AI strategy. Very few have a data strategy that can actually support one.”
The effort required to clean, standardise, and govern enterprise data is often underestimated because it lacks the visibility and excitement associated with AI deployments.
Alongside data, implementing AI messes with traditional ways of doing things and challenges longtime habits. “A relationship manager who’s been operating on his guts for the last fifteen years will not trust a system telling him instructions.”
This means organisations must invest heavily in training, change management, and user adoption rather than assuming technology alone will drive transformation.
Asthana also highlights the importance of involving risk and compliance teams early in the process. According to him, successful organisations follow a clear sequence: establish data foundations first, integrate processes second, build intelligence third, and design governance throughout.
AI is reshaping, not replacing, wealth advisory
The wealth management industry is witnessing a significant shift in customer expectations. Investors increasingly demand personalised experiences, instant access to information, and digital engagement models that mirror leading consumer applications.
Asthana believes AI is finally making true personalisation at scale possible. “A client conversation no longer has to be generic. It can reflect their portfolio, tax position, recent actions, and long-term goals.”
However, he rejects the notion that AI will replace human advisors, particularly in the HNI and UHNI segments. Instead, he sees AI as a force multiplier that allows advisors to focus on higher-value interactions.
Much of an advisor’s time today is consumed by administrative tasks, reporting, documentation, and routine queries. “If you can save that time for him, then the advisor becomes better, not redundant. They have more conversations that matter. They catch more opportunities. They retain more clients,” he adds.
For Asthana, the future is not about replacing relationships but strengthening them. “The firms that will win aren’t replacing the human relationship. They’re making it more available.”
Governance will determine the future of AI adoption
As AI systems become more influential in financial decision-making, governance is emerging as a strategic differentiator rather than a compliance obligation.
Asthana remains unequivocal on this point. “Speed without governance isn’t innovation. In financial services, it’s just sophisticated risk with a better user interface.”
He argues that every AI deployment must clearly establish accountability, transparency, and explainability before entering production.”Real explainability means the logic is traceable from input to output, auditable by a compliance team, and defensible to a regulator or an unhappy client.”
His framework for evaluating AI initiatives revolves around three questions: “Who is this decision affecting? What happens when it’s wrong? And can we explain it in plain language to the person it affected?”
Human oversight, he insists, will remain essential for the foreseeable future. “Human-in-the-loop isn’t a limitation of current technology. It’s a deliberate design choice that reflects how much is actually at stake.”
The future belongs to organisations that align strategy and technology
Looking ahead, Asthana expects AI capabilities themselves to become increasingly commoditised. Competitive advantage will instead come from data assets, organisational capabilities, and the ability to integrate technology into business strategy.
He also believes agentic AI will fundamentally reshape operating models across financial services.
“We are moving from AI that simply answers questions to AI that can take actions, whether that’s booking appointments, executing tasks, escalating issues, or following up with clients.”
This evolution will force institutions to rethink compliance structures, supervisory controls, and customer-consent frameworks.
Ultimately, Asthana points out the traditional distinction between business strategy and technology strategy is disappearing.
“In capital markets, technology choices are strategy choices. Which platform you build on, which data you own, and which workflows you automate, these decisions determine what you can and can’t offer to clients five years from now.”
For financial institutions navigating the AI era, the biggest risk may not be adopting the wrong technology. It may be treating technology as execution rather than strategy. “The most dangerous position a financial institution can be in is treating technology as execution and business as strategy. Because by the time the strategy needs the technology, it’s already too late to build it.”
