There's a particular kind of corporate announcement that is part product launch, part territorial claim. Infosys's unveiling of its AI First Value Framework on February 17, 2026 (timed to coincide with its Investor AI Day in Bengaluru) falls squarely into that category.
The headline number is hard to ignore: a $300–$400 billion incremental AI services opportunity by 2030. But the framework itself, and the arguments Infosys chairman Nandan Nilekani made alongside it, tell a more interesting story than the dollar figure alone. They reveal a market rapidly shifting from experimental pilot programs to institutionalized, multi-year enterprise commitments. It also reveals a race among major IT services firms to own that transition before their clients figure out what they actually need.
For CIOs and enterprise transformation leaders, the timing matters. The vendors are moving faster than most buyers.
What the AI First Value Framework Actually Is
The framework is Infosys's structured answer to a question that most large enterprises are still struggling to frame: how do you go from "we're doing AI pilots" to "AI is how we actually run the business?"
It organizes Infosys's AI services into six value pools, each targeting a distinct layer of the enterprise transformation challenge:
- AI Strategy & Engineering: Designing AI architectures, platforms, and operating models using proprietary tools, AI agents, and third-party integrations.
- Data for AI: Getting enterprise data into a state where AI can actually use it, through AI-grade data engineering, synthetic data generation, and building what Infosys calls "trusted data assets."
- Process AI: Redesigning end-to-end workflows by integrating AI agents alongside human teams, with the goal of improving efficiency across functions.
- Agentic Legacy Modernization: Using AI agents to reverse-engineer and modernize legacy systems without bringing everything to a halt. This is the most technically ambitious pool and arguably the one with the highest stakes for most enterprises.
- Physical AI: Embedding AI into physical products, devices, and operations using digital twins, robotics, and edge intelligence.
- AI Trust: Governance, security testing, regulatory compliance, and responsible AI practices across the lifecycle.
The entire framework runs on Infosys Topaz, the company's generative and agentic AI suite, which now includes a composable layer called Topaz Fabric. Infosys says it's already working with 90% of its top 200 clients across more than 4,600 AI projects, with 500-plus enterprise AI agents built and over 30 new offerings launched across the six pools. AI work accounted for 5.5% of revenue in Q3 FY26, and CEO Salil Parekh has made clear that number is supposed to grow significantly.
At that scale, this isn't a pilot program. It's a market position being established before competitors can fully respond.
Nilekani's Argument: Why This Time Is Different
Nandan Nilekani co-founded Aadhaar, the world's largest biometric identity system, after helping build India's IT services industry in the 1990s. When he says the current AI transition is fundamentally different from what came before, the observation carries weight.
His central metaphor at AI Day 2026 was blunt: AI is not an upgrade. It's "root-and-branch surgery."
Previous technology transitions (the internet, mobile, cloud) allowed enterprises to layer new capabilities on top of existing structures. A company could launch a website, then a mobile app, then migrate to cloud infrastructure, largely without dismantling how the business actually operated at its core. The organizational chart stayed intact. The processes stayed mostly the same. The tech changed around the edges.
AI breaks that pattern. Capturing real value from AI requires rethinking customer journeys, breaking data silos that have calcified over decades, and adopting what Nilekani calls "non-deterministic approaches" — meaning enterprises have to get comfortable with AI systems that don't produce the same output every time.
He also identified what he called a "deployment gap": the widening distance between what AI can do and what enterprises have actually deployed. The technology is advancing faster than the organizational capacity to absorb it. Critically, the real difficulty isn't building new things with AI. Greenfield projects are relatively simple. The hard part is modernizing the sprawling legacy systems that run most large enterprises: the 30-year-old mainframe code, the siloed databases, the processes that have never been properly documented. That's where most of the cost and risk lives, and that's precisely where IT services firms are positioning themselves as irreplaceable.
Nilekani also flagged something that doesn't usually make it into the marketing slides: white-collar resentment. As AI reshapes knowledge work, the risk of backlash from employees in organizations that rush transformation without adequate reskilling is real. He called it a potential "train-wreck" if handled badly.
Who Else Is in This Race
Infosys is not alone in targeting this market. The enterprise AI services land grab is genuinely competitive.
Accenture has reported nearly $1 billion in generative AI bookings in a single quarter, with full-year AI bookings reaching approximately $3 billion. They're positioned at the strategy end of the market, targeting Fortune 500 companies with multi-year transformation engagements.
IBM is leading with WatsonX, its enterprise-grade generative AI platform, emphasizing governance, security, and regulated industry expertise. IBM Consulting has 21,000 data and AI professionals with pre-built accelerators for specific industries.
TCS and Wipro are competing on execution efficiency: structured modernization frameworks that deliver at scale without the premium price tags of the top-tier consultancies.
Every one of these firms is doing the same thing: building a proprietary framework around existing client relationships and selling it as the definitive path from experimentation to scale. The frameworks have different names and slightly different emphases, but the strategic intent is identical. Become the essential intermediary between AI technology and business outcomes before a client develops the internal capability to do it themselves. That's simply how IT services markets work. The question for enterprise buyers is whether they enter these engagements with clear eyes.
What This Means If You're Evaluating AI Services Partners
The structured frameworks now being deployed by every major IT services firm create real risks for buyers who don't approach them strategically.
Lock-in is the default outcome, not the exception. A 2026 survey found that 94% of IT leaders fear vendor lock-in as AI services scale. That fear is well-founded. Once your data pipelines, AI governance, and modernization programs are embedded in a vendor's platform (Topaz Fabric, WatsonX, or anything comparable), switching costs compound fast. Six-pillar approaches and multi-year roadmaps are not designed for easy exits. And unlike cloud migrations of the past decade, where you could often lift-and-shift workloads between providers, AI service engagements involve organizational change, trained models, rewritten processes, and custom agent architectures that are genuinely difficult to replicate with a different partner.
The "deployment gap" is real, but treat it as a diagnostic, not a verdict. Nilekani is right that most enterprises haven't deployed AI effectively. But the conclusion vendors draw ("therefore you need us, end-to-end, for seven years") deserves scrutiny. Some organizations genuinely need a full-service partner to close the gap. Others need focused, targeted help in specific areas. Know which one you are before signing anything.
Agentic legacy modernization is where the money and the risk concentrate. Every major services firm is promising AI-driven modernization of legacy systems. This is the highest-value pool in Infosys's framework, and the one most likely to create deep dependencies. Once a partner's AI agents have reverse-engineered your mainframe code and rebuilt it in their cloud environment, you cannot swap partners without serious operational disruption. The switching cost is operational as much as financial.
"AI Trust" deserves independent scrutiny. Governance and compliance frameworks built by your implementation partner create inherent conflicts of interest. Infosys auditing the risks of Infosys-built AI systems is like asking your contractor to inspect their own work.
Questions Worth Asking Before You Commit
What does portability actually look like? This is the most important question on the list, because the answer changes the negotiating dynamic before anything is signed. Ask explicitly: if we end the engagement in three years, what do we walk away with? Can our AI models, training data, and rebuilt processes run independently of your platform? Push past the vague language in the initial proposal and get specifics in writing.
How are costs structured as we scale? A framework that looks affordable at the pilot stage can become expensive once AI agents are running across core business processes. Understand the pricing model at 10x current usage. Hidden dependencies that inflate costs at scale are a common issue in large technology engagements and they are rarely disclosed upfront.
Who owns AI governance, and who audits it? If the answer is "we do, as part of the AI Trust pillar," probe further. What audit rights do you retain? What happens when the AI system makes a high-stakes error in a regulated context? For most enterprises in finance, healthcare, or infrastructure, this cannot be left as a footnote in the engagement framework.
What's the reskilling plan for our workforce? Ask your prospective partner for specifics on how they've managed workforce transitions in comparable engagements. Not a vague mention of "change management," but actual examples and outcomes. If they don't have a concrete answer ready, that's a signal worth taking seriously before the contracts are signed.
The Strategic Takeaway
The vendors have already decided what the enterprise AI market looks like. The question is whether buyers show up with their own answer.
Nilekani is right that this AI transition requires deeper organizational surgery than previous technology waves. The six-pool structure of the AI First Value Framework addresses real problems that enterprises face as they try to scale AI beyond pilots. But "well-constructed" is not the same as "right for you." The firms racing to capture this market are building frameworks designed to maximize their own revenue and retention.
The enterprises that navigate this well won't necessarily be the ones that found the best vendor. They'll be the ones that came to the table knowing exactly what they needed, where they were willing to build deep partnerships, and where they were drawing the line on dependency. Most organizations don't show up knowing those things. That's the actual gap the vendors are counting on.