AI Is Adding Revenue and Removing It: The Double Edge Facing Indian IT

The Headline Paradox: Expanding AI Inflows Alongside Deflationary Anxiety

TCS reported on 8 October 2026 annualised AI revenue of US$3.1 billion, above 10% of its total revenue and up from US$2.6 billion in the previous quarter, according to Business Standard. Yet at the same time, a Kotak Neo preview of Wipro's Q2 FY27 results listed, among key concerns for the sector, "AI-led deflation", alongside seasonal weakness, higher employee furloughs, weak discretionary spending and whether large deals convert into revenue.

What "AI-led deflation" means: In IT services, clients traditionally pay for effort: people, hours and time. If AI tools let a team write code, test software or answer support tickets in half the time, a client may expect to pay less. Contracts increasingly include productivity commitments, where the vendor passes some savings to the client. Deflation here means the same work earns less revenue. The Kotak Neo preview used the term as a concern; we are describing the mechanism in general terms, not reporting a figure from any company.

This dynamic strikes directly at the economic foundation of Indian technology exports. Over four decades, the Indian IT services delivery model grew into a US$250 billion export engine on the arithmetic of linear headcount expansion: more software developers billing more billable hours equalled higher top-line revenue. Generative code assistants, synthetic test data generation, and automated incident triage disrupt that linear equation by decoupling output volume from human hours billed.

Brokerage Projections for Wipro Q2 FY27: A Sector Barometer

Because Wipro reported on 15 October 2026, the preview is dated just before the results, and we did not find a reliable report of the actual numbers. The preview gave consensus estimates, which are brokerages' forecasts and not results:

• Year-on-year revenue growth of about 9% and profit growth of about 2%. • A decline of 0.5% to 1.4% quarter on quarter in IT services revenue in constant currency; Emkay expected a 1% fall in dollar terms and Nuvama a 1% fall in both measures. • Margins improving: Emkay expected IT services EBIT margin up 50 basis points sequentially, Motilal Oswal expected a rise of 40 basis points to 16.4%, Jefferies 25 basis points and Nuvama 20. • Third-quarter guidance in a range of minus 2% to flat (Emkay and Nuvama) and minus 1.5% to plus 0.5% (Jefferies).

These are forecasts. Readers should consult the company's actual announcement. What these brokerages highlight, however, is a common industry reality: even as operational margins slightly improve through internal automation and offshore cost discipline, sequential constant-currency top-line expansion remains stagnant across multiple quarters.

“When the machine does the hours, the industry has to sell the outcome.”

How AI Can Both Help and Hurt: A Five-Pillar Structural Breakdown

1. New work: Clients need help to build, secure and run AI systems: data pipelines, model integration, governance. This is the revenue that firms count in AI run-rate metrics.

2. Cheaper old work: Routine coding, testing, documentation and support are the tasks where AI saves the most time, and these have been the base of IT services revenue.

3. Timing: New AI revenue arrives in pilots that may take months to scale, while price cuts on old contracts can come at renewal. A period of net pressure is therefore plausible even if the long-run effect is positive. This is our reading, not a reported fact.

4. Pricing models are shifting: Vendors are exploring outcome-based pricing, such as paying per resolved ticket. That moves some risk to the vendor and rewards efficiency.

5. Measurement is loose: Because "AI revenue" is a company-defined term, comparing the US$3.1 billion at TCS with another firm's figure requires care. Ask for definitions.

When evaluating corporate claims of "AI revenue", enterprise leaders must recognize that vendors classify projects differently. For instance, migrating a legacy SQL database to a modern cloud data lake may be classified as "AI enablement" revenue because the cloud data will eventually feed an enterprise LLM. True generative AI engineering—fine-tuning proprietary models, deploying RAG pipelines, and building autonomous agent workflows—remains a smaller, albeit rapidly expanding, slice of the aggregate total.

Sharpening the Analytical Picture and Key Unanswered Questions

What share of the AI run-rate is net new, and what share is relabelled from existing work? What fraction of large deals include productivity give-backs? How much of the fresh-graduate pipeline is being cut because tools do the entry-level work? What is the revenue per employee trend? None of these was in the sources we saw.

Furthermore, the speed at which productivity give-backs are institutionalized creates structural margin volatility. If a multi-year master services agreement mandates a 15% annual productivity reduction, the service provider must continually deliver 15% more scope or discover new project workloads merely to prevent contractual revenue from shrinking. If client demand remains flat, billable billing rates suffer steady downward compression.

Actionable Guidance for Clients, Vendors and Early-Career Engineers

What a client should do: If you buy IT services, ask your vendor to show how it uses AI tools in your work, and to share the productivity gains in price or scope. Define outcomes, set service levels and review contracts at renewal. Ask about data security: where does your data go when tools are used? Enterprise IT buyers should insert explicit AI productivity clauses into statement-of-work (SOW) renewals, establishing clear benchmarks for defect rates and delivery cycle velocity.

What a vendor's customer-facing staff should do: Prepare to explain value in outcomes rather than hours. Teams that can show faster delivery, fewer defects or lower run costs will fare better than teams that sell headcount. Solution architects must shift client conversations from "how many engineers are needed" to "what business metric is being transformed".

What a student or early-career engineer should do: Learn the fundamentals and also how to work with AI tools, because the entry-level tasks are changing. Build projects that show you can design, test and explain a solution.

A note on timing: Because results seasons move quickly, numbers in this piece may be overtaken within days. Check the latest company announcements before relying on any figure.