Ajay Vasal, Global Leader for Data & AI, Genpact – Interview Series

Ajay Vasal, Global Leader for Data & AI, Genpact, is a veteran technology and strategy executive with more than two decades of experience spanning artificial intelligence, data strategy, digital transformation, mergers and acquisitions, and enterprise consulting. In his current role at Genpact, he leads initiatives focused on helping enterprises scale AI adoption and transition from assistive AI toward more autonomous, operationally embedded systems. Prior to joining Genpact, Vasal spent nearly nine years at Accenture, where he held several senior leadership positions including Global Data & AI Lead for Consumer Industries and Global Head of the Centre for Data & Insights. Throughout his career, he has focused on helping organizations unlock measurable business value from data through operational efficiency, revenue growth, and risk mitigation, while also leading strategic investments, M&A initiatives, and AI-driven transformation programs across industries.

Genpact is a global professional services and digital transformation company specializing in data, AI, analytics, intelligent automation, and operational modernization for large enterprises. Originally founded within General Electric before becoming an independent company, Genpact now serves organizations across industries including financial services, healthcare, manufacturing, retail, and consumer goods. The company has increasingly positioned itself around AI-powered enterprise transformation, combining analytics, automation, and domain expertise to help businesses modernize workflows, improve decision-making, and deploy large-scale AI systems responsibly.

You’ve spent years at Accenture shaping data and AI strategy across industries before stepping into your current role at Genpact. How has that experience influenced your perspective on why enterprises still struggle to extract real value from AI investments today?

One of the biggest lessons from working across industries is that most organizations do not fail because the technology is weak. The reality is that they struggle because they try to layer AI onto operating models that were never designed for autonomous execution. The findings from our recent research reinforce that point clearly. While enterprises are moving aggressively into AI adoption, many are still struggling to redesign the operating models needed to support autonomous execution at scale. That gap highlights how many enterprises are still working through questions around governance, accountability, and organizational readiness.

We are also seeing organizations move quickly before they have the right measurement frameworks in place. Seventy-one percent of executives believe agentic AI will deliver ROI faster than any previous wave of technology, yet 67% still rely on productivity metrics that were built for earlier automation models. My experience has shown that successful AI adoption is never just about deploying models. It is about redesigning workflows, aligning technology with business priorities, and helping people understand how their roles evolve alongside increasingly autonomous systems.

Your recent report, Autonomy Requires Trust in AI, shows that most executives believe agentic AI will fundamentally change business operations, yet most systems remain supervised. What is the biggest disconnect between belief and execution?

The disconnect is entirely centered around trust. Executives clearly believe agentic AI will reshape how work gets done, but belief in the technology does not automatically translate into comfort with handing over decision-making authority. Our research found that only 22% of organizations are comfortable allowing AI systems to operate with domain-level or broad autonomy.

Most enterprises are still comfortable using AI as an assistant that can recommend, summarize, or support workflows. The hesitation begins when AI systems are expected to act autonomously in ways that affect operations, customers, compliance, or financial outcomes. That hesitation is forcing many enterprises to keep humans tightly involved in approval and oversight processes even as they push toward more autonomous operating models.

Many enterprises are investing heavily in AI but failing to demonstrate meaningful returns. What are the most common mistakes organizations make when trying to measure AI-driven value?

A major mistake is that many organizations are still using productivity metrics designed for earlier waves of automation. The data shows that 67% of enterprises still depend on productivity-based measures that cannot fully capture the value of adaptive, decision-driven systems. That creates a real disconnect between expectations and how value is actually being measured.

Another issue is that many organizations have not yet defined AI-native success metrics. Only a small percentage of enterprises are measuring things like autonomous workflow completion, reduced escalation, or autonomous exception handling. If companies continue focusing only on cost reduction or hours saved, they will miss the broader operational impact agentic AI is designed to create.

The report suggests that organizational readiness is a bigger barrier than technology itself. What specific structural changes do companies need to make to unlock the full potential of agentic AI?

Organizations need to rethink how accountability, workflow ownership, and decision-making operate across the business. One of the clearest takeaways from the research is that organizational readiness is becoming a bigger barrier than technology itself. In fact, 33% of enterprises identified business processes not being ready for agentic integration as the top obstacle to adoption.

The companies making the most progress are redesigning workflows end-to-end so autonomous systems can operate within clearly defined boundaries and oversight structures. We also found that 44% of enterprises expect agentic AI to flatten management structures as systems absorb coordination tasks traditionally handled by middle management. Employees need much greater clarity around oversight responsibilities, intervention points, and where human judgment fits into an increasingly autonomous environment.

Only a small percentage of organizations are comfortable granting AI agents real autonomy. What will it take for enterprises to trust AI systems with decision-making authority?

Trust in agentic AI ultimately comes down to accountability and control. Enterprises are still cautious about handing over decision-making authority because leaders want confidence that autonomous systems can operate within clear guardrails and governance structures.

The companies moving ahead are designing systems where escalation paths, intervention triggers, and governance controls are built directly into the operating model. Most organizations are granting autonomy incrementally based on business context and risk tolerance rather than making a sudden leap to full autonomy. As enterprises gain confidence through successful deployments, trust grows because leaders can clearly see how decisions are made, where accountability sits, and how human oversight is maintained.

Enterprises expect to scale agentic AI quickly, yet many still rely on outdated productivity metrics. What should more effective, AI-native performance metrics look like?

AI-native metrics need to measure execution and outcomes rather than activity. Traditional productivity measures focus on whether people are working faster, but agentic AI changes the equation because the system itself is beginning to carry a part of the operational workload. That is important because many organizations are still measuring agentic AI through a productivity lens that was designed for earlier automation waves rather than autonomous execution.

More effective measures should focus on autonomous workflow completion, reduced escalation rates, faster decision execution, and how effectively systems manage exceptions with limited human intervention. Those metrics provide a much clearer picture of whether AI is truly improving execution at scale.

You highlight that process redesign is the leading barrier to adoption. Why does workflow transformation matter more than model performance in this next phase of AI?

Workflow transformation matters because even the most advanced AI systems cannot scale effectively inside broken or fragmented processes. Our research found that 33% of enterprises identified process readiness as the leading barrier to agentic AI adoption.

What makes this next phase different is that agentic AI is designed to execute across workflows rather than simply assist within them. Organizations must redesign processes around autonomous execution by clarifying decision ownership, removing unnecessary handoffs, and embedding governance directly into workflows. In many ways, workflow redesign is now the true foundation for scalable AI adoption.

The report suggests that AI will flatten organizational structures as coordination tasks become automated. How do you see leadership roles and middle management evolving as a result?

As agentic AI takes on more coordination and operational oversight tasks, leadership roles are likely to shift from managing routine execution, towards guiding strategy, judgment, and governance. Our research found that many enterprises expect agentic AI to reduce layers of coordination across the organization as autonomous systems take on more operational oversight responsibilities.

Middle management has historically played a major role in routing decisions and maintaining operational continuity across teams. As autonomous systems begin handling more of those activities, leaders will need to focus more on exception management, accountability, and ensuring systems align with business objectives. Human leadership will become even more important in areas where context, ethics, and judgment still matter most.

Genpact positions itself as an agentic and advanced technology solutions company. How is the company approaching agentic AI differently from traditional consulting or technology firms?

Genpact’s approach is grounded in execution and outcomes and what we often describe as solving the “last mile” problem of AI adoption. A lot of organizations have already proven that AI models can generate insights or improve productivity, but the real challenge is embedding those capabilities into live business workflows where work actually gets done. That is where many AI initiatives stall. Our focus is on helping clients operationalize agentic AI across the last mile of execution so systems can drive measurable outcomes inside finance, supply chain, customer service, and other core enterprise functions.

What differentiates our approach is the combination of deep process expertise with data, technology, and AI capabilities. We are not just deploying models or building pilots. We are redesigning workflows, embedding governance into operations, and helping enterprises create the accountability structures needed for autonomous execution at scale. Because we have also embedded agentic AI into our own operations, we bring practical experience around what it takes to move from experimentation to enterprise-wide impact.

Looking ahead, do you believe agentic AI will widen the gap between leading and lagging enterprises, and what should companies be doing now to avoid falling behind?

Yes, I do believe the gap will widen because agentic AI compounds value differently than earlier technology waves. The organizations that successfully redesign workflows, establish accountability frameworks, and build trust into autonomous systems will create operational advantages that become difficult for competitors to replicate. What is striking is how quickly this shift is happening. Twenty-nine percent of enterprises expect agentic AI to scale across the business within the next 12 months.

For companies looking to stay competitive, the priority now should be preparing the operating model for autonomous execution. That includes redesigning workflows, defining ownership structures, building AI-native measurement frameworks, and investing in workforce readiness. The organizations that move early on those foundations will be much better positioned than those still approaching AI as a standalone technology initiative.

Thank you for the great interview, readers who wish to learn more should visit Genpact.