AI is pushing IT services away from hours and toward owned outcomes_

AI is changing something more consequential than developer productivity: the unit buyers use to purchase technology services. Reuters reported on August 20 that clients of major Indian IT firms are pushing for faster delivery and lower prices as AI reduces the effort behind repeatable work. Persistent Systems CEO Sandeep Kalra said some clients are asking for the same work at 25–30% lower prices.

The contract structure is moving with that pressure. TCS CEO K. Krithivasan told Reuters that roughly 80% of contracts in its finance, HR and other business-services segment are now priced around outcomes rather than hours. The shift is not universal across IT services, but it is concrete enough to change how engineering firms should package work.

Sell the bottleneck, not the bench_

A customer rarely needs six developers as an end state. They need quotations produced faster, a support backlog reduced, data synchronized, a product launched, or a manual process removed. When the engagement is framed around that bottleneck, the buyer can compare the cost of delivery against the cost of leaving the problem in place.

• Define the baseline before promising the improvement.
• Price a bounded operational outcome, not an unlimited transformation.
• Separate implementation from infrastructure, model/API consumption and ongoing operations.
• Tie milestones to usable capability: a reconciled data flow, an automated workflow, a production release, a measured reduction in handling time.
• Write assumptions, dependencies and failure boundaries into the commercial model. Outcome pricing without these controls is simply unbounded fixed-price risk.

AI makes measurement more important, not less_

The useful part of AI-assisted delivery is not that fewer hours can be billed. It is that repetitive engineering work can become cheaper while the customer still values the solved problem. That margin only exists when both sides can measure what was delivered.

This is where smaller engineering firms can compete differently. They do not need to imitate a large outsourcing bench. They need enough technical depth to own a narrow result end to end, enough instrumentation to prove it, and enough discipline to decline outcomes they cannot control.

A practical starting point_

Take one recurring workflow and write down its current cycle time, failure rate, manual touches and operating cost. Then define the smallest production change that moves one of those numbers. That is a better starting point for an AI engagement than a catalogue of models, agents or features.