Half the Code Is AI-Assisted. So Why Are Engineering Teams Still Growing?

Note: This story was written with AI assistance and edited by me.

Cognizant reported its third quarter 2026 results on October 29. Revenue was $5.7 billion, up 4.6% year over year, or 4.8% in constant currency. CEO Ravi Kumar S said the company is moving through the AI shift with "over 50% of our code now AI-assisted" and "continued improvement in revenue and adjusted operating income per employee." Total headcount was 359,900 at the end of September, up 10,100 from a year earlier and up 3,200 from June. In one client example, Cognizant said a seven-person team used Devin, Cognition's AI software engineer, on a legacy conversion for Odyssey Logistics. Delivery throughput rose by roughly a third, and Odyssey recorded a 37 percent net cost saving against the conventional approach. Cognizant also narrowed its full-year constant currency growth guidance to 4.0% to 4.4%, from 4% to 5.5% earlier.

Those are the facts as reported. Everything below is my opinion.

My view: AI is writing more of the code, but it is not shrinking the need for engineers. It is changing what a good team looks like. The winners will run small, senior teams that own the result and treat AI output as a first draft, not a finished product.

Half the code, more people

The usual story says that when AI writes half the code, you need half the coders. Cognizant's numbers do not fit that story. More than half its code is AI-assisted, revenue per employee is improving, and it still has 10,100 more people than a year ago.

That tells me the work did not disappear. It moved. Less time goes into typing boilerplate, and more goes into deciding what to build, checking what the machine produced and making it work inside a messy client system.

The bottleneck moved to review

AI-assisted code is fast to write and easy to get wrong. Someone has to read it, test it, catch the security gap and say no when it solves the wrong problem. That someone is almost always an experienced engineer.

So the scarce skill is no longer output. It is judgment. A junior developer with an AI assistant can produce a lot of code. A senior developer with the same assistant can produce a lot of code that actually ships and does not break in production.

What buyers should ask their vendors

If you buy software delivery, the Odyssey example is the one to study. Seven people, a hard legacy conversion, a third more throughput and a 37 percent net saving. The point is not the tool. The point is a small team that knew what done looked like.

Ask how big the team really needs to be. A smaller senior team with good tooling can beat a large pyramid.

Ask who reviews the AI output. If the answer is vague, the risk is yours.

Ask what result you are paying for. Headcount on an invoice is not the same as working software.

A fair counterpoint

To be fair, AI is not only adding jobs at Cognizant. Its Project Leap program expects $150 million to $200 million in employee severance and other personnel costs, and third quarter bookings declined 7% year over year. The company also plans to hire 1,500 U.S. college graduates, which suggests it still sees a real role for junior talent. Headcount growth is not the whole story. My view is only that AI is reshaping teams faster than it is shrinking them.

What I learned running small senior teams

I have staffed projects both ways. The big team looks safer on paper because there are more names on the plan. In practice, the small senior team usually ships sooner, because fewer people need to agree and everyone on it can spot a bad decision early.

AI makes that gap wider. When a tool can draft code in seconds, the slow part is knowing whether the draft is right. Experienced engineers answer that question quickly. Less experienced ones often cannot tell, and the mistakes show up later, when they cost more to fix.

It also changed how I hire. I stopped asking how fast someone could write code and started asking how well they could read it, question it and explain the tradeoffs to a client. Those people are harder to find, and they are worth it.

Takeaway

AI is taking over the typing, not the thinking. That is good news for anyone who can define a result and judge the work that gets you there. I took a firm from $6M to $60M in under three years, and the turning point was building teams around senior judgment instead of headcount.

If you run an engineering or services team, look at one project this quarter and ask a simple question: who checks what the AI writes, and do they have the experience to say no? The answer will tell you more than the team size.