AI Adoption ROI: Cognizant's Outcome-Driven Shift

An IT chief redraws the map of AI adoption ROI

In July 2026, Cognizant CEO Ravi Kumar S told CNBC TV18 that enterprise AI has crossed into a new phase. His framing was direct: the age of experimentation is closing, and the age of measurable return has opened.

For years, companies chased AI use cases with loose commercial goals. Kumar argues that season is ending fast, replaced by a harder question about what each deployment actually returns.

The claim carries weight because Cognizant sits inside hundreds of enterprise deployments. When the head of a major IT services firm redraws the map, boards tend to read the coordinates carefully, and they act on what they see.

From effort-based billing to outcome-driven work

The core shift Kumar describes is a change in how technology work gets priced. IT has long billed for effort: hours, seats, headcount.

He sees the industry moving toward outcomes instead. The measures that matter become productivity gained, operations optimised, and business results delivered. That reframing lands hard for anyone who buys software or services at scale.

An effort model rewards activity. An outcome model rewards result. The gap between those two ideas is where the next decade of enterprise spending will be decided, and Kumar wants Cognizant standing firmly on the outcome side of that line.

The original idea: track improvement, not revenue share

Here is the design choice worth studying. Rather than measuring AI's contribution as a share of revenue, Kumar said Cognizant tracks improvement across its work, per his interview with CNBC TV18.

That distinction sounds small. It reshapes everything downstream.

Tying AI to a revenue percentage invites vanity math and inflated claims. Tracking improvement forces a before-and-after comparison on a specific metric: cycle time, error rate, throughput. This aligns with a position this desk holds firmly: AI adoption ROI reads correctly through a before/after view on a defined number, rather than through a vague label like "significant gains."

Where fresh money is moving: BFSI

Kumar pointed to strengthening momentum in banking, financial services and insurance. He linked it to regulatory changes in the United States and rising enterprise confidence.

The telling detail sits in the source of the cash. This spending looks fresh, rather than reallocated from existing IT budgets. New budget signals conviction, while shuffled budget signals caution.

For a board reading benchmarks, that is the sentence to underline. When regulated institutions open new lines of technology spending, the perceived risk of AI deployment has dropped inside one of the most conservative sectors on the planet, and that repricing of risk travels quickly to neighbouring industries.

The friction: refined guidance and a disciplined tone

Every honest AI story carries a moment of friction, and this one does too. Cognizant refined its full-year guidance during the same period Kumar spoke.

Read that as a recalibration, rather than a retreat. Kumar said the expected exit rate points to stronger business momentum heading into the next financial year, with growth accelerating in the second half.

The willingness to adjust guidance while projecting confidence is a sign of operational maturity. Leaders who publish clean numbers with zero tension are selling a narrative. A refined forecast paired with a stated exit rate gives readers something they can actually test later, which is the clearest form of accountability an executive can offer.

A signal from the services layer, not the model makers

The voices shaping AI narratives usually come from model builders. Kumar speaks from a different seat: the services layer that installs AI inside real operations.

That vantage point matters. Model makers sell capability. Services firms live with the messy reality of integration, change management and support tickets.

When someone from that layer declares the return era open, the claim reflects thousands of implementation hours, rather than a demo on a stage. The wider Indian IT sector is reading the same shift at once, which suggests the change runs deeper than one firm's messaging.

Model selection becomes the new discipline

Kumar described the next stage of AI as disciplined and value-driven. Companies now evaluate which models suit which tasks, balancing cost, performance and efficiency.

This is the quiet engineering story underneath the headline. The winners choose the right model for the job, rather than defaulting to the largest or the loudest option on the market.

He added that organisations are building workflows where AI systems and employees complement each other. These pairings create continuous learning cycles that sharpen decisions and operations over time. The architecture of that pairing, decided early, tends to determine whether a deployment compounds or stalls.

What you can take from this

The lesson generalises well beyond one IT firm. Pick a single metric, capture the baseline, deploy, and compare the delta with rigour.

Each reader can act on a different angle:

• Founders and CEOs: price your own AI work by outcome, and track improvement on one metric first.
• CTOs and heads of product: match the model to the task, and design the human-plus-AI workflow before scaling.
• Boards and investors: treat fresh budget in regulated sectors as a shifted benchmark of the possible.
• Managers and team leads: run a before-and-after on one workflow this quarter.

The through line is discipline. Measurable return replaces open-ended experimentation, and the metric you choose becomes the contract you sign with your own results.

The open question for your organisation

Kumar's argument reduces to a single test. Can you name the metric your AI work improves, state its baseline, and show the change?

Enterprises that answer with a number and a date are operating in the return era he describes. Those answering with adjectives are living in the experimentation phase he says is closing.

So the question turns back to you. When your next budget cycle opens, will your AI spending buy activity, or will it buy a result you can prove? For more verified enterprise cases, browse our blog index.