‘Doom-and-gloom framing that AI is destroying jobs misunderstands what AI actually does well’

Vijay Vijayasankar is the Global Agentic AI Officer at Genpact. He is responsible for driving Genpact’s agentic AI vision and accelerating innovation through strategic partnerships across data, AI, go-to-market, and delivery functions. Prior to joining Genpact, Vijay held senior leadership roles at IBM, MongoDB, and SAP. He holds a degree in mechanical engineering and an MBA from the University of Kerala. In an interaction with TOI, Vijay Vijayasankar shared Genpact’s AI-driven initiatives and more. Genpact says it has moved beyond traditional BPO to ‘AI-driven autonomous operations’, what does this shift mean in terms of revenue mix, and how much of your business today is already AI-led?At Genpact, the shift to AI-driven autonomous operations is less about isolating AI as a revenue line and more about embedding AI into the fabric of how work gets done. Our Advanced Technology Solutions business -- spanning data and AI, digital technology, advisory, and agentic solutions -- continues to grow and now represents a significant and increasing share of our overall business. That’s not a trend; it’s a clear signal of where enterprise demand is moving. More broadly, nearly half of our portfolio is now AI-infused, underscoring the central role AI plays across our clients. What's changing in the revenue mix is the nature of the engagement itself. Rather than traditional BPO with people-based contracts, enterprise clients are increasingly looking to buy outcomes like touchless invoice processing with our agentic accounts payable suite or faster financial close with our agentic record-to-report suite. You’ve positioned yourself as building “agentic AI systems” rather than offering services. How is this changing the way clients engage with you compared to traditional outsourcing contracts?Genpact has three decades of deep experience operating the G&A and COGS functions of leading global enterprises, so we understand the last mile nuances that make or break each process. This is how we can offer productized agentic AI solutions to clients which are differentiated from classic AI and software products as well as traditional services contracts. We codify the knowledge of data, process, and operator experience into our agentic solutions and offer it at scale, and we stand behind business outcomes. The conversation is no longer about labour arbitrage, it's about how intelligently work can be orchestrated and executed at scale, with humans stepping in for judgment and exception handling. And the market is responding. In 2025, we saw strong adoption of the Genpact AP Suite, our first agentic solution suite for accounts payable. What’s particularly telling is that a significant portion of that demand came from net new clients. That's not rotation of an existing book of business; that's evidence of a meaningfully expanded total addressable market. For existing clients who have rotated from FTE-led to agentic models, we're seeing both revenue growth and gross margin expansion that has outpaced even our own projections from mid-2025. With AI taking over repetitive processes, what kind of roles are actually growing within Genpact are you seeing a shift towards data, AI training, and domain-led roles? The easiest way to explain what is happening is that everyone in Genpact is becoming an AI Practitioner, and within that some colleagues are majoring as AI Builders. Builders design the systems: data engineers, AI architects, agentic solution developers. Practitioners are the domain experts: the finance leads, supply chain specialists, and risk analysts who now work alongside agents rather than alongside spreadsheets. I want to push back on the broader narrative here. The doom-and-gloom framing that AI is destroying jobs misunderstands what AI actually does well. Take programming – arguably the single most powerful AI use case today. Software engineering job numbers are not declining; if anything, they're ticking up. Why? Because the actual typing of code is maybe 10-20% of a software engineer's job. The thinking, the architecture, the problem definition, that's the other 80%. AI takes over the repetitive part and frees the human to do more of the valuable part. The same logic applies across every domain we work in. The humans who risk their jobs being displaced are the ones who resist working with AI to redefine their roles.As jobs are being redefined, people are being trained and repositioned to work with AI agents. A supply chain manager who six months ago was running reports is now managing agent workflows and interpreting AI-generated exception flags. Does this transition mean lower headcount growth compared to the traditional BPO model, or is it creating new categories of jobs within the organisation?The linear relationship between revenue growth and headcount growth is breaking – and that's by design, not by accident. In the traditional BPO model, to grow the business 10%, you added 10% more people. Agentic Operations changes that equation entirely. At Genpact, the composition of the workforce is changing more than the overall size. We continue to invest aggressively in AI talent – through both hiring technology experts and intentionally training and upskilling our teams. Clients are increasingly looking for outcomes rather than manpower, are contracts now moving towards AI-driven, outcome-based pricing models, and how does that impact margins?Yes, we see margins being accretive in the outcome-based model for Agentic Operations, while clients get a much better outcome. This is how the model is designed from the ground up. Let me introduce an analogy that resonates well with clients. Think of the civil society we live in – we need laws and law enforcement to uphold them. A local police officer or state trooper both operate within a defined jurisdiction. If a suspect crosses a state line, those cops can't follow. But a federal agent with broader jurisdiction (like FBI in US or CBI in India) can. That's how Agentic Operations works. AI agents handle what I call the “happy flow” – the well-defined, repeatable tasks within their jurisdiction, like standard invoice matching. When something falls outside that jurisdiction, the highly trained human – “the special agent” – steps in. The key insight is that you can't deploy your federal agent on simple noise complaints. The regular cop needs to hold the line on 80% of work, so your most skilled people are reserved for the exceptions that genuinely need them. That's the model. As AI becomes the execution layer, client expectations are naturally shifting toward outcomes over effort. This aligns with the broader move to AI-first operations, where value is delivered through AI-enabled precision, speed, and scale rather than manpower.From a delivery standpoint, this enables more predictable and efficient operations, driven by straight-through processing and reduced variability. Over time, this supports a transition toward value-linked engagement models, where pricing is increasingly tied to business outcomes. From a margin standpoint, the benefits are evident in the increasing efficiency and scalability of the model, even as we continue to invest in strategic priorities. The shift toward non-FTE, outcome-based delivery is a key driver. Within our Advanced Technology Solutions business, a large share of work is recurring and delivered through non-FTE models, combining durability with higher-quality economics at scale.With global players combining IT and BPM capabilities, do you see the industry moving towards a model where AI, consulting, and operations are fully integrated?The industry is clearly moving toward a model where AI, consulting, and operations are converging into a single, integrated capability – where the definition of success is the outcome that is underwritten, and not the value of hours spent building it.What differentiates leaders is not just their ability to deploy AI, but how deliberately they design for autonomy at scale, across data, architecture, orchestration, and governance. Part of what I see my own role doing is driving exactly this convergence inside enterprises. The Global Agentic AI Officer role exists precisely because organisations are in transition – moving from a largely human-oriented way of solving problems toward an increasingly autonomous enterprise model. We bring technology and process into the same room, connecting deep domain and industry knowledge, proprietary data, and agentic systems to truly integrate AI and transform their businesses.