AI Productivity in Indian IT: Who Keeps the Savings?

AI productivity can benefit an IT customer without producing an equal increase in the supplier’s revenue or profit. Who keeps the saving depends on the contract, the price of the work and the cost of delivering a reliable result.

That distinction matters for Indian IT services. Faster coding, testing or support is an operational claim. Revenue growth and cash generation are commercial outcomes that still need their own evidence.

Public evidence reviewed: 26 October 2026. The TCS and Wipro observations below concern the quarter ended 30 September 2026. They are reported company disclosures, not Altys forecasts, audited AI-attribution studies or securities recommendations.

What the latest disclosures actually tell us

In its 8 October 2026 results release, TCS reported annualised AI revenue of about US$3.1 billion for Q2 FY27. It also reported group revenue of about US$7.6 billion for the quarter, sequential constant-currency growth of 0.5% and an operating margin of 24.0%.

In its 15 October 2026 release, Wipro reported flat sequential constant-currency IT services revenue and an IT services operating margin of 16.5%, up 0.5 percentage points sequentially. Management described AI as a long-term source of transformation demand.

The observations support a narrower conclusion than either “AI has solved growth” or “AI is destroying the industry.” AI activity, total revenue and margin can move differently. Neither release by itself isolates AI’s causal contribution to the quarter’s profit.

Do not compare TCS’s AI run rate with Wipro’s total services revenue as though they are the same measure. Do not add TCS’s annualised figure to its reported quarterly sales. An annualised metric is not automatically twelve months of recognised revenue or a forecast of the next twelve months.

The contract determines the first financial effect

Under a time-and-materials arrangement, a customer may pay for hours or people. If the same work needs fewer billable hours and the rate is unchanged, the invoice can fall. The supplier may need additional work or a different commercial model to offset that reduction.

Under a fixed-price arrangement, an efficiency improvement can initially leave more of the fee with the supplier. But renewal negotiations, competitive bids and delivery obligations can redistribute the benefit later. Faster completion may also free capacity for other paid work, provided that demand exists.

Under an outcome-based arrangement, payment depends on the defined result. The analysis then turns to measurement, acceptance, service quality, liability and the cost of achieving that result. Changing the contract label does not remove those questions.

This is why an industry-wide assumption such as “twice the productivity means twice the profit” is not a financial model. It skips the price negotiation.

A ₹100 fee becomes a different business at ₹85

Consider an invented project with a fixed fee of ₹100 lakh and delivery costs of ₹80 lakh. Its project contribution is ₹20 lakh before central overheads, depreciation, interest and tax. This is not a listed company’s operating margin.

Assume AI reduces the original delivery cost by ₹20 lakh, but adds ₹8 lakh for software usage and human review. The new delivery cost is ₹68 lakh. With the same ₹100 lakh fee, contribution rises to ₹32 lakh.

Now suppose the customer negotiates the next comparable project down to ₹85 lakh. With the same ₹68 lakh cost, contribution is ₹17 lakh. The work is still cheaper to deliver than before, but the supplier earns less contribution in absolute rupees.

Invented comparable-project scenarios, not TCS or Wipro data. Contribution = project fee minus stated delivery costs, before central overheads, depreciation, interest and tax. Costs fall from 80 to 68; fees are 100, 100 and 85. No volume change or cash-timing assumption.

The contribution percentages are 20%, 32% and 20%, respectively. The repriced project has the same percentage as the original, but ₹3 lakh less contribution. A stable margin percentage can therefore hide a smaller rupee contribution when the revenue base shrinks.

To restore the original ₹20 lakh contribution at ₹17 lakh per project, the simplified business would need roughly 18% more comparable projects: 20 ÷ 17 − 1. That is a break-even arithmetic sensitivity, not a demand forecast. It assumes costs scale exactly with projects and that capacity, quality and overheads do not constrain delivery.

Faster output is not automatically an equal labour saving

A coding step might become faster while requirements, integration, security review and customer acceptance still take time. Measuring the isolated step and applying that percentage to the whole project overstates the improvement.

Similarly, staff time saved does not immediately remove a salary cost. People may be redeployed, retained for new demand or required to review the new output. Transition costs can arrive before commercial benefits.

A sound productivity measure therefore states the task, baseline, quality threshold, review requirement and period. “More lines of code” and “less time to an accepted working system” are not identical outcomes. A company-level financial claim requires the bridge from that operational measure to billed work and actual costs.

Our IT services results guide provides the broader growth, deal-conversion and utilisation context. The constant-currency explainer separates operating growth from currency translation.

What deserves monitoring next quarter

Track changes in comparable revenue and pricing first, not merely how often AI appears in a presentation. Retain the company’s definition of AI-related revenue and note any change in scope. A reclassification of existing work is different from additional customer spending.

Then examine deal ramp-up, renewal pricing, utilisation, subcontracting and the costs of delivery. An improving margin does not identify its own cause: currency, staff mix, acquisitions and restructuring can also matter.

Finally, check collections and cash conversion over a suitable window. A project can be commercially successful while payment timing makes one quarter unusually strong or weak. The quarterly model-update workflow helps keep period changes, assumptions and financial statements connected.

Altys’s view: record the commercial bridge

The durable research question is not “Does this company use AI?” It is “How does changed delivery economics reach the customer’s bill, the supplier’s costs and the resulting cash?”

Altys’s source-linked research, guidance history and company monitoring help teams return to the original statement when later results arrive. Use available Excel exports to review the evidence and calculations rather than treating a polished AI summary as the conclusion. Request access to Altys.

The AI bubble discussion asks whether investment economics can justify the spending. For Indian IT services, the immediate test is more specific: after customers and delivery costs take their share, how much of the productivity gain remains?

General business education only. All project arithmetic is hypothetical. No securities recommendations, price targets, return expectations or allocations. Altys Labs is not a SEBI-registered Research Analyst or Investment Adviser.