AI Contract Reset Squeezes India’s IT Services Giants

Tata Consultancy Services now ties about 80% of one major services segment to outcomes, as AI pushes clients to demand more work for less money. The shift, reported by Reuters and surfaced through Google News, reaches far beyond a routine pricing adjustment. It challenges the labor-based economics that helped TCS, Infosys, Wipro, HCLTech, and Cognizant become global outsourcing leaders.

Customers are no longer satisfied with promises that AI will eventually make delivery more efficient. They want those gains written into contracts through lower prices, measurable results, shorter commitments, or shared savings. Some are also moving work in-house because AI tools let smaller internal teams handle tasks previously assigned to outside vendors.

That creates a difficult contest between enterprise buyers and service providers. Buyers want immediate savings without accepting new technology risks. Providers must invest in AI, protect service quality, and surrender part of the productivity gain before knowing whether automation will deliver consistently.

Google News Reveals a Contract Model Already in Retreat

The most important change is not that Indian IT companies use AI, but that clients now control how its economic value gets divided.

The traditional outsourcing model often linked revenue to employee numbers, billing rates, and hours worked. A provider could expand an account by adding engineers, support personnel, or business-process specialists. Automation now weakens that connection because the same output can require fewer people and fewer hours.

According to an August 21 Reuters report syndicated by MarketScreener, TCS CEO K. Krithivasan said roughly 80% of the company’s finance, human resources, and related business-services contracts use performance outcomes. A person familiar with the matter told Reuters that this share had doubled since AI became mainstream in late 2023.

Outcome-based pricing links some or all of a vendor’s compensation to an agreed business result. That result might involve processing speed, system availability, customer-service performance, operational savings, or another measurable target. The vendor therefore carries more execution risk than it would under straightforward hourly billing. A TCS analysis of generative AI and outcome-based service-desk pricing similarly notes that contracts must balance productivity incentives with service quality and clearly defined performance measures.

TCS has described a staged approach for AI projects. Management said the company begins some engagements with outcome commitments before moving to fixed-price arrangements as the work expands. Fixed-price contracts establish a payment for defined work, regardless of the exact number of labor hours required.

That model can reward efficiency when automation works as planned. If TCS delivers a defined result with fewer resources, it can retain part of the productivity benefit. However, inaccurate estimates, unexpected infrastructure expenses, or weak AI output can quickly transfer that benefit back to the client.

TCS management said in July that AI was producing productivity savings of about 10% to 15% in some work. The company was passing those benefits to customers gradually, according to a pricing update. That phased transfer can soften the immediate revenue impact while giving TCS time to adjust its delivery costs.

Other companies are testing more aggressive structures. Reuters reported that a Cognizant agreement with Daimler Truck called for AI-related savings to be shared between the companies. Neither party disclosed full contractual details publicly.

A separate HCLTech cloud-management agreement with German utility E.ON reportedly delayed payment during its first year. Payments from the second year were tied to efficiency improvements and specific business outcomes. HCLTech and E.ON declined to discuss the contract terms. Separately, HCLTech’s 2025–26 integrated annual report says clients increasingly want outcome-based models and identifies outcome- and consumption-based contracts as part of its services strategy.

These examples show why the underlying story matters. AI productivity has moved from presentation slides into clauses that determine when vendors get paid, how much they receive, and which party absorbs failure.

The shift also affects existing work. Customers can revisit a renewal and ask why an application-maintenance contract should cost the same after coding assistants reduce repetitive labor. Even stable billing rates offer limited protection when the client wants fewer workers, fewer hours, or a smaller total agreement.

Shorter contracts add another layer of pressure. They let customers reconsider vendors and technology choices sooner, limiting the time providers have to recover investments. Vendors must then spend more frequently on sales, transitions, and competitive bids.

The result is a structural change in bargaining power. Indian providers still bring industry knowledge, global delivery systems, and large pools of technical talent. Yet scale alone no longer proves that a company can deliver an AI-enabled outcome more efficiently than a smaller competitor.

Buyers Are Claiming the Productivity Dividend First

Enterprise customers increasingly treat AI efficiency as a discount they have already earned, even before vendors fully redesign delivery.

Persistent Systems CEO Sandeep Kalra told Reuters that clients were asking for the same work at prices 25% to 30% lower. Those customers also expected faster delivery and higher productivity. Their demands illustrate the central conflict inside current negotiations.

Clients have a straightforward argument. If AI can generate code, test software, summarize incidents, and automate support tasks, the provider should need fewer billable hours. Paying the old contract value would allow the vendor to keep the entire productivity gain.

Vendors see a more complicated cost equation. AI tools require models, computing capacity, governance systems, security controls, data preparation, evaluation, and human review. Faster production does not remove the need to verify accuracy or integrate results into older enterprise systems.

Token costs also matter. A token is a small unit of text processed by an AI model, and providers commonly pay according to usage. Wipro CEO Srini Pallia said rapid deployment can increase those expenses, forcing customers and providers to examine the complete cost of ownership.

AI productivity also varies by task. Repetitive coding and testing offer clearer automation opportunities than architecture decisions, regulatory work, or complex migrations. A buyer demanding one uniform discount across every activity can overstate the savings available from the technology.

Nevertheless, broader economic conditions strengthen the customer’s position. Enterprise technology budgets have remained cautious, while discretionary projects face close scrutiny. Companies can invite several vendors to bid against each other and make AI savings a condition for consideration.

Application maintenance has become an early pressure point. Industry participants told Mint that renewal prices in that category had fallen by roughly 4% to 5% over the previous year. HCLTech’s CEO separately estimated annual AI-related revenue deflation of up to 5% for the industry.

Revenue deflation occurs when technology reduces what clients pay for an existing quantity of work. It differs from losing the contract completely, but the financial effect can accumulate across thousands of projects.

The industry has faced buyer-friendly cycles before. During the 2001 technology downturn and the 2008 financial crisis, customers pushed for price caps, savings guarantees, and outcome-based arrangements. In 2023, more than 80% of 1,600 tracked technology and business-process deals included some form of committed savings, compared with about 65% in 2019, according to an earlier Reuters deal review.

AI makes this cycle different because it attacks the unit underlying much of the industry’s revenue. Earlier downturns reduced demand while leaving the relationship between labor and output largely intact. Automation can permanently reduce the labor needed after demand returns.

This distinction explains why large contract announcements provide an incomplete picture. A multiyear deal can increase the reported order book while producing less revenue than a comparable agreement once generated. It can also contain penalties, savings commitments, or delayed payments that reduce its economic value.

Clients are additionally building internal capabilities. AI coding tools let corporate technology teams complete more work without proportionally increasing headcount. Global capability centers, which are company-owned offshore operations, can combine internal business knowledge with lower-cost engineering talent.

Insourcing does not eliminate the need for external providers. Enterprises still require help integrating models, securing data, modernizing older systems, and operating infrastructure. However, it changes what they purchase and reduces the advantage of selling large teams for routine delivery.

The immediate pressure therefore falls on providers that cannot generate new work faster than AI compresses existing revenue. TCS says new engagements have offset that downward pressure so far. Whether that balance holds will determine if AI becomes a growth engine or a continuing discount mechanism.

Smaller Rivals Turn AI Efficiency Into a Scale Advantage

AI is weakening the assumption that the largest workforce automatically makes the safest technology partner.

For decades, major Indian IT providers used scale as both a delivery system and a sales argument. Large employee bases allowed them to staff complicated programs, serve multinational customers, and replace workers when project needs changed.

AI changes that calculation by reducing the number of people needed for selected tasks. A mid-sized provider can use automation to prepare prototypes, migrate code, test applications, and analyze support records without assembling an enormous delivery team.

Smaller firms can also make decisions quickly. HFS Research CEO Phil Fersht told Reuters that many customers want rapid pilots. Mid-tier companies can compete by putting senior leaders on those projects quickly and offering flexible commercial terms.

Persistent Systems and Coforge demonstrate the competitive threat. Both recorded at least eight consecutive quarters of double-digit dollar revenue growth through the April-to-June 2026 quarter. Persistent’s revenue increased 16%, while Coforge’s sales rose by about one-third.

TCS, Infosys, Wipro, and HCLTech posted much slower growth of 1% to 3% during the same period, Reuters reported. These figures cover companies with different business mixes, so they do not prove that AI caused the entire gap. They do show that size is not producing superior growth during this transition.

Persistent’s Kalra argued that revenue alone no longer defines a scale provider. His company is benefiting on both sides of the contract reset. Customers demand lower prices, yet AI also helps Persistent compete for agreements that previously might have gone to a larger company.

This is the core reversal facing the industry. Automation should theoretically favor large providers because they possess extensive data, expertise, and investment capacity. In practice, it can remove the staffing advantage that protected them from smaller rivals.

The shift does not mean every challenger will win. Large providers retain deep relationships in banking, insurance, manufacturing, healthcare, and government. They also operate security, compliance, and delivery systems that a new entrant cannot reproduce quickly.

However, incumbency becomes less valuable when a client breaks one large transformation into several shorter projects. Each smaller engagement creates another competitive opening. It also lets customers compare results before making a broader commitment.

AI-native delivery can strengthen those challengers further. A provider that designs a workflow around automation from the beginning avoids the difficulty of reworking a labor-intensive process. An incumbent may need to reduce its own revenue before winning the replacement engagement.

This self-cannibalization problem affects internal incentives. Sales teams are accustomed to expanding account value, while delivery managers often plan around staffing and utilization. An AI proposal that shrinks both can appear unattractive even when it protects the relationship.

Large providers are responding by building expertise closer to customers. TCS is expanding its group of forward-deployed engineers, specialists who work directly with client teams to turn AI prototypes into operational systems. The company is also seeking acquisitions that can add AI capabilities. Its broader effort to move clients from experiments to operational outcomes is reflected in the company’s fiscal 2026 results and product announcements, including a platform designed to accelerate AI deployments into real-world business workflows.

Wipro sees demand moving beyond experiments. Pallia said boards and chief executives were asking for returns on AI investments during 2026, after many companies focused on proofs of concept in 2025. He estimated AI-assisted software development could cost about 25% less because of coding and testing gains.

That opportunity does not automatically belong to the lowest bidder. Clients need providers that can connect models with data, security policies, enterprise applications, and business processes. A failed deployment can cost more than the initial contract savings.

The competitive dividing line will therefore be credible execution, not company size by itself. Vendors must show that they can produce measurable benefits while controlling model errors, infrastructure costs, and operational risk.

Aggressive AI Guarantees Can Turn Savings Into Losses

The contract reset rewards real productivity, but it can punish providers that promise improvements their technology cannot reliably deliver.

Tech Mahindra CEO Mohit Joshi has warned that some competitors are pricing deals around projected productivity improvements of 70% to 80%. Those agreements can extend for five to seven years, even though the savings require major client-side process or system changes.

Tech Mahindra says it has avoided taking that risk. Joshi characterized parts of the market as irrationally competitive during a July analyst call. His concern exposes the largest uncertainty behind outcome-based pricing.

A vendor can automate its own delivery process, but it does not control every condition that determines a business outcome. A client may have fragmented data, outdated applications, slow approval procedures, or employees who resist a new workflow.

Outcome contracts must allocate those dependencies clearly. Otherwise, the provider can miss a target because the customer did not complete a required system change. Disputes then shift from hours recorded to responsibility for the failed result.

Long agreements create another forecasting problem. Models, chips, energy costs, regulations, and security requirements can change before the contract ends. A fixed commitment based on current assumptions may become uneconomic several years later.

Infosys demonstrated that providers retain one defensive option. The company told analysts in July that it had left contracts that no longer made financial sense. Walking away protects margins, but it can reduce revenue and weaken a longstanding client relationship.

The financial evidence is mixed. Wipro’s operating margin fell 130 basis points sequentially to 16% in the April-to-June quarter. Tech Mahindra’s margin increased 60 basis points to 14.4% during the same period, according to a recent pricing analysis.

A basis point equals one-hundredth of a percentage point. These quarterly moves reflect many factors beyond AI contracts, including workforce utilization, subcontracting, currency movements, and project mix. They should not be treated as a clean test of pricing strategy.

Employment changes offer another signal, though their causes also require caution. TCS announced reductions affecting more than 12,000 positions during 2025. The company described the move as part of broader strategic initiatives, including investment in newer technologies.

Former Infosys CFO V. Balakrishnan told Reuters that coding agents reduce the need for large groups of entry-level programmers. If that assessment proves accurate, the industry’s staffing pyramid faces a lasting redesign.

The pyramid model places many junior employees beneath smaller groups of experienced managers and specialists. It supports margins when companies can bill clients for large teams while keeping average labor costs controlled.

AI can automate work that traditionally trained entry-level engineers, such as basic coding, testing, documentation, and support. That creates an operational benefit but also removes part of the career ladder used to develop experienced technical leaders.

Companies must replace that learning pathway rather than simply remove junior positions. They need employees who can inspect AI output, understand client systems, manage security, and exercise judgment when automated results fail.

The same caution applies to optimistic revenue claims. AI creates projects involving data modernization, model deployment, governance, and workflow redesign. Yet new demand must exceed the continuing compression of maintenance, development, and business-process revenue.

TCS says it has managed that balance so far. Krithivasan acknowledged that future growth depends on staying ahead of revenue deflation. That is a more useful measure than the number of pilots or employees trained on AI.

Readers encountering this story through Google News should also distinguish reported contract structures from independently audited performance. Several important examples rely on people familiar with private agreements. The companies involved declined to disclose complete terms.

The direction of travel is clear, but the long-term profitability remains unsettled. Outcome pricing can preserve margins for efficient providers, or it can convert ambitious productivity forecasts into contractual liabilities.

Three Signals Will Show Who Controls the Next Deal Cycle

The next phase will be decided by contract economics, relative growth, and workforce composition, not by the volume of AI announcements.

The first signal is how major providers describe revenue deflation in their next earnings reports. Investors should watch whether TCS continues to offset lower revenue from existing work with new AI engagements.

A widening gap would weaken the claim that implementation demand can replace lost billable hours. Stable growth and margins would support the view that phased savings and outcome pricing can protect the business.

Infosys and Wipro deserve similar attention. Infosys reduced the upper end of its full-year outlook after management cited stronger productivity demands and aggressive pricing. Wipro has acknowledged both compressed delivery economics and opportunities from new AI work.

The useful figures are constant-currency revenue growth, operating margin, large-account performance, and the value of completed work. A large order book matters less if contracts are shorter, cheaper, or dependent on difficult performance targets.

The second signal is whether mid-sized competitors sustain their growth advantage. Persistent and Coforge have expanded much faster than the largest providers, but maintaining that pace will test their delivery systems.

Continued double-digit growth would reinforce the conclusion that AI has lowered the scale barrier. Slower growth, troubled projects, or margin deterioration would suggest that large providers still hold meaningful advantages once pilots become complex production systems.

Contract announcements should be examined for more than headline value. Buyers should look for the project duration, payment structure, savings allocation, and responsibility for model or infrastructure costs. Those details reveal which party truly controls the economics.

The third signal is how companies redesign entry-level employment. TCS’s workforce reduction and industry comments about coding automation suggest that providers expect lasting changes. Hiring plans will show whether they are removing capacity or creating a different talent model.

A successful transition would combine smaller junior cohorts with structured training in AI evaluation, enterprise architecture, security, and client operations. A simple collapse in recruitment would create short-term savings but risk a future shortage of experienced staff.

Enterprise buyers should care about that distinction. A lower contract price offers limited value if the provider lacks people who can manage failures, understand older systems, or preserve institutional knowledge.

Knowledge workers also face a practical consequence. When contracts reward outcomes instead of hours, employees gain value by connecting AI output to business decisions. Producing more code or documents matters less than proving that the output improves a measurable result.

Teams adopting AI should preserve evidence behind those decisions. A searchable AI knowledge base can help employees retain project context, compare model output, and document why a workflow changed. Those records become especially important when payment depends on results.

The broader market has already registered concern. Reuters valued India’s IT services industry at about $315 billion in annual revenue. The Nifty IT index had fallen roughly 20% during 2026, erasing a combined $73 billion in market value across its 10 members by August 21.

Markets can overshoot, and those losses do not settle the industry’s future. Indian providers still manage complex systems that enterprises cannot replace with a chatbot or coding agent. Their challenge is proving that expertise has value after routine labor becomes cheaper.

The next contract cycle will clarify which companies can make that case. If providers maintain margins while accepting outcome terms, AI will have changed their operating model without destroying it. If prices fall faster than delivery costs, customers will capture most of the benefit.

That is why this development deserves attention beyond India’s stock market. It offers an early view of what happens when AI productivity becomes a commercial demand rather than a technical promise.

Watch the next earnings calls for measurable revenue replacement, sustained mid-tier growth, and a credible workforce redesign. Those three signals will reveal whether India’s IT leaders are selling better outcomes or merely accepting smaller contracts.