CI&T grew sales 15% with no more billable engineers, and now charges for results on 40% of new deals

When AI makes programmers faster, who keeps the savings: the client or the outsourcing firm? CI&T, a Brazilian software services company listed in New York, gave one of the clearest answers yet on Wednesday. Increasingly, the firm does.

CI&T grew third-quarter revenue 14.9% to $146.3 million. It did that with roughly the same number of billable engineers it had a year earlier. The company's securities filing puts the quarterly average at 6,786 of what it calls "AI builders."

The lever is pricing. Chief financial officer Stanley Rodrigues told analysts that 40% of new engagements signed in the quarter were "value-based", up from 30% in the first half. Value-based means the client pays for an agreed result, not for hours worked. For brand-new clients, chief executive Cesar Gon said the share reaches 75%.

Gon also put a number on why the firm prefers it. He said those contracts earn, on average, 8 to 10 percentage points more margin than traditional hourly billing. Under hourly billing, an AI tool that halves the hours also halves the vendor's revenue, and the client pockets the difference. Under a fixed-result price, the vendor keeps much of that gain.

What the numbers show

Revenue per billable engineer reached $82,500 over the past 12 months, up about 10% from a year earlier. That growth rate sped up from 7% the previous quarter. Over the same period, overall headcount rose just 1.8%, the CFO said.

The cost of delivering work grew 10.9%, slower than revenue. Adjusted gross margin, the share of revenue left after paying for the people and tools doing the work, rose 2.7 points from a year earlier to 36.6%. Management credited what it calls AI monetization, meaning pricing that lets the firm profit from its own productivity gains.

CI&T says it runs several AI models through its own delivery platform rather than betting on one supplier. It reported more than 50 major projects using Anthropic's Claude. It also said it reached a preferred tier in Anthropic's partner program within three months of joining. Management conceded that partnerships with the big AI labs were not yet a meaningful source of revenue.

Why investors were not convinced

The shares fell about 7% in after-hours trading, to around $3.10, according to Investing.com. Adjusted earnings of 9 cents a share missed analysts' 10-cent consensus, and net profit fell 14.4% to $7.6 million.

Winning these deals costs money. Selling expenses rose to 14% of revenue, from 11.9% in the second quarter. Before the report, Earnings Whispers noted the company had already cut its 2026 profit-margin target, citing heavier sales spending and a stronger Brazilian real. Third-quarter adjusted EBITDA margin, a common gauge of operating profit, was 15.5%, inside the lowered full-year target of 15% to 17%.

Key questions remain open. CI&T did not say how it measures the "value" in value-based deals, how many hit their targets, or what happens to fees when they miss. The productivity figures are the company's own. Concentration is a further risk: the largest client brought in about 12% of quarterly revenue, and the top 10 about 43%.

Guidance was wide. CI&T expects fourth-quarter revenue of $142.3 million to $152.0 million. Gon said the low end reflects economic and currency risk, and the high end depends on turning a sales pipeline that Rodrigues said is 50% bigger than a year ago into signed work.

What it means for buyers

The firm is not alone in repackaging AI work. On Tuesday, Cognizant launched a 25,000-person group devoted to Claude projects. It also introduced a managed service, in which it runs AI systems for clients, and said the service would make AI spending more predictable. In May, Anthropic announced a new services company, with Blackstone, Hellman & Friedman and Goldman Sachs as founding partners, to install Claude at mid-sized firms.

Asked whether the AI labs' growing in-house teams threaten firms like his, Gon was skeptical. "I don't believe the FTE levels these labs are promoting are sustainable," he said, referring to full-time staff.

For a chief financial officer or procurement head, the lesson is about contract design. Fixed-result deals move delivery risk onto the vendor, which has real value. But they also hand the vendor most of the savings from AI. CI&T's own 8-to-10-point margin premium is a rough gauge of what clients pay for that certainty.

Buyers signing such deals can ask for three things: an hourly-cost baseline to compare against, a clear definition of the outcome being paid for, and a clause to revisit prices as AI tools improve. Without those, the productivity gains that AI promises may show up mainly in the supplier's margins.