For decades, the typical contract between an Indian IT services firm and a large Western corporation followed a simple script: billable man-days, a pyramid of junior consultants that swelled with each renewal, and a supplier that grew more competitive the more heads it could line up.
That grammar is being rewritten right before our eyes.
A Reuters investigation shows that TCS, Infosys, Wipro, HCLTech and Cognizant are rebuilding their business models, tying more of their fees to measurable outcomes rather than hours worked.
Industry leaders are frank about losing mandates outright.
Indeed, clients are re-insourcing certain tasks thanks to AI, while a backdrop of uncertainty is shortening contract durations.
A market under pressure on the provider side
According to Jimit Arora, head of the Everest Group analyst firm, the balance of power has decisively shifted in favor of clients.
The shift is reflected in the numbers finance chiefs are beginning to disclose, albeit cautiously.
According to TCS chief executive K. Krithivasan, roughly 80% of the contracts in his finance, HR and other business services segment are now based on performance indicators, versus about half before the widespread adoption of generative AI at the end of 2023.
At Cognizant, the pivot is just as clear. Fixed-price and transactional contracts—where the provider bears the execution risk and shares productivity gains with the client—now account for more than 50% of the group’s revenue, up from 41% three years ago.
Saurabh Gupta, president of HfS Research, encapsulates the mechanism taking hold for new deals: a subscription backbone for the platform component, a portion of consumption billed on hours for human labor and in tokens for AI agent time, plus a performance component.
Case studies that are turning into jurisprudence
Three recent deals illustrate this contractual shift in concrete terms.
At Cognizant, the February agreement with Daimler Truck to modernize its global workplaces through AI and automation stipulates that AI-driven savings will be shared between the provider and the client.
Another case, even more radical: the multi-year cloud-management agreement signed in June 2025 between HCLTech and the German energy giant E.ON provides that the Indian provider will not be paid in the first year, payments starting in the second year and remaining indexed to efficiency gains and the achievement of specific business objectives.
The third file, more spectacular in monetary value, is emblematic of a scale shift.
On July 3, HCLTech announced a $1.14 billion contract over 5.5 years with a European Fortune Global 50 group, identified by several Indian media outlets as Mercedes-Benz.
The agreement, which runs through December 2031 with a five-year extension option, covers the building and operation of an AI-driven operating model for the group’s entire digital-workspace environment and corporate networks.
The deal is not inconsequential from a competitive standpoint: it replaces a $3.2 billion eight-year mandate Infosys held since 2020 with Daimler before its split into Mercedes-Benz Group and Daimler Trucks.
Where the 2020 agreement read as traditional outsourcing, the new version rests on AI agents taking charge of IT support, network anomaly detection and automated first-level resolution—a mandate intrinsically anchored in AI execution, no longer tied to dedicated staff billed by the hour.
The paradox of full-order books
This price reconfiguration yields a perplexing statistical effect for investors: orders surge, but the translation into revenue slows.
At Infosys, the major contracts signed in the first quarter of fiscal 2027 reached $3.6 billion in total value, with 61% of new business, yet the company trimmed its annual growth guidance to a range of 1.5% to 3%.
The management attributed this disconnect to the very nature of these deals.
Vendor consolidation programs typically translate into revenue faster, as the provider takes on existing operations, whereas transformation programs take longer to unfold as clients complete transitions, refine their processes and validate subsequent phases.
In other words, the total contract value disclosed in press releases says little about ramp speed or the behind-the-scenes productivity commitments negotiated, precisely the parameters that are becoming central to contractual negotiations.
Not all players are moving at the same pace
This transition is neither uniform nor without internal frictions.
Tech Mahindra’s latest health-care contract illustrates the growing sophistication of contractual metrics. The business model is tied to measurable outcomes, including around 40% fewer support tickets, a 20% reduction in average resolution time, and a 30–35% drop in technical debt over the contract’s duration.
By contrast, Infosys acknowledges, via its CEO Salil Parekh, that it has not yet observed a broad wave of such contracts. Mayank Verma, Xebia’s global head of data and AI, sees this as a transitional, experimental phase expected to intensify. He argues the sector must move from a pure helpdesk model toward a “pod” model that combines AI agents with human staff.
Contracts rewritten, in both directions
The shift is not limited to the billing method; it also touches the very structure of commitments.
Outsourcing lawyers note, on the buyers’ side, a desire to secure five risk areas before signing. They are: transparency about the AI tools actually used, an explicit prohibition on using client data to train models, expansion of intellectual-property clauses to AI-generated outputs, a liability framework for errors or “hallucinations” by the system, and productivity-sharing clauses that prevent the provider from pocketing all the automation gains.
On the vendor side, the logic is sometimes the reverse. Some push for longer engagement terms (five, seven or even ten years) to amortize their technology investments, which forces IT management to lock in contract protections up front, before the economic balance of the deal shifts.
Engineers are being redeployed
This shift has a direct corollary in the workforce. At Wipro, AI initiatives have yielded productivity gains equivalent to the output of 20,000 employees, redeployed internally rather than laid off.
Of a total staff of roughly 243,000 as of June, more than 100,000 employees have already undergone AI training or earned advanced AI certifications. A single engineer can oversee several AI agents, be reassigned to other projects, or upskill into a new role. That does not necessarily mean a one-for-one replacement by an agent.
This redeployment is accompanied by investments in a cadre of engineers stationed directly at client sites to accelerate AI adoption.
TCS plans to build a team of up to 8,900 of these “forward-deployed engineers” (FDE); Infosys aims for about 6,000 in the coming years, and Wipro is also expanding this pool.
For IT professionals negotiating with these providers today, the day rate has not disappeared, but it is no longer the only currency of exchange.
Hybrid pricing grids are now on the table, the precise mechanics still being hammered out deal by deal, between a client who wants to pay for a result and a provider that must continue to fund the transition toward a model in which the number of engineers involved matters less.
