Are You Subsidizing Your Vendor’s AI Investments? (The Time & Materials Problem)

Imagine you’re Cognizant. A $21 billion software development company. Your Q3 2025 analysis shows AI tools – GitHub Copilot, Claude – have just made you 30% faster. 

Your teams are handling 46% of code generation in files where AI tools are enabled. Individual coding tasks are up to 55.8% faster.

Now comes the question: do you tell your clients? Do you charge 30% less? Or do you keep billing the same, and keep the AI productivity gains as margin? 

This is what hundreds of vendors are deciding right now. And if you have a T&M contract with a vendor in place, this should matter to you.

In this article, I’ll break down what’s really happening in that choice – and why it’s broken for both you and your vendor.

In Q3 2025, Cognizant (NASDAQ: CTSH), a $21 billion IT services firm, reported that AI productivity improvements delivered a 30% acceleration across their portfolio. 

To be fair, the gains weren’t uniform: some projects see 5% gains (legacy code rewriting), others see 60% (greenfield development). But averaged across their client base: 30%.

Here’s how this plays out in dollars: A $30,000 project that would have taken the vendor 200 hours – now takes 140 hours.

The work is identical. The outputs are the same. But if it’s a T&M contract, the vendor’s paycheck changes. 

Under traditional Time & Materials pricing, here’s what happens: they invoice you for 140 hours instead of 200. You just saved $9,000. But consequently, your vendor “lost” $9,000. They made the AI investment – but all it did was allow you to pay them less money.

So – how are vendors likely to react to this? Are they going to stay honest about their AI productivity gains, and hurt their bottom line?

Or is this system incentivizing them to be dishonest? To keep charging you for 200 hours, even though they got the work done in 140?

Note that I’m not saying here that the vendor is at fault. It’s the T&M contract that introduces perverse incentives.

The principle behind this trap is simple: optimize for billable hours, and that’s what you get.

Alan Weiss, the consulting strategist whose ‘Million Dollar Consulting‘ (1992) challenged T&M models from the start, proposed a better solution: ‘You assess fees. You create these based on value-not heads, not boxes, not time and materials, but the value you bring to the client” 

His philosophy was clear: focus on outcomes and results, not time spent. Under T&M, neither of you can do that.

Here’s the trap: your vendor spent $500K on AI tooling, training, and process optimization-created a 30% productivity boost. But under T&M, where does that value go? When AI makes work faster, your vendor faces a choice-and loses either way:

Option A – the Honest Option: Invoice actual hours (140).

• At first glance, this seems great for you as a client: you gain 30% savings, you save $9,000. 
• But the vendor realizes that they aren’t seeing any benefit from their $500K AI investment.
• So the vendor stops investing in AI. Their tools stagnate.
• If other vendors also stop investing in AI in a similar way, the result is…
• …you lose as well, in the long term. Because the next project takes longer and costs more – because under T&M the vendor doesn’t have incentive to make the project take less time and cost less.

Option B – the Dishonest Option: Stretch the project, invoice 200 hours, hide the productivity.

• You lose from the start: you pay for phantom work, for hours that were never worked. 
• Your vendor initially wins – they keep the margin gained from AI productivity gains.
• But the market eventually catches on. And when you compare notes with peers? When you see other vendors finishing similar projects in half the time? The result is simple…
• You let your vendor go – so they end up losing in the end as well. The market verifies the lie.

None of these is good for either side. Either way, the T&M model ensures someone loses. Usually, it’s both of you. The system doesn’t produce winners, but casualties.

There’s a third option – but the current model won’t allow it. We’ll get to that.

What is the difference between vendors that win vs. lose in an AI world? Let’s look at two examples.

Cognizant (The Leader): Saw this dilemma coming. Instead of focusing on T&M work, they shifted from 43% fixed-price revenue (2024) to approximately 47% (Q3 2025). So they stopped optimizing for hours. They optimized for outcomes. 

Result: adjusted operating margin expanded by 70 basis points year-over-year in Q3 2025. Organic growth (excluding Belcan acquisition) was roughly 4% in Q3 2025, with 6.5% total constant-currency growth. Translation: they realized T&M rewards the wrong behavior. So they stopped playing the game.

CEO Ravi Kumar stated on the earnings call: “Productivity has gone up 30%, which means there is more throughput, so you can actually create more throughput, share the savings, lower cost of deployment with our clients.” 

Apparently, a vendor focusing on contracts where they get paid for results instead of excuses is good for business – counterintuitive stuff!

But let’s look at a different scenario…

Endava (The Laggard): Didn’t see this change coming. 76% of revenue is still T&M. They have strong AI tools (5-10× faster in some workflows). But they didn’t change the pricing model. 

Result: Adjusted profit-before-tax margin collapsed from 20.7% in FY2023 to 5.5% in Q1 FY2026. Stock price fell significantly during this period (from an all-time high of $170.13 in Dec. 2021 to under $10 in late 2025). In short: They chose T&M. The market punished them.

CEO John Cotterell stated on the Q1 FY2026 earnings call: “As we’re delivering a much higher productivity, that is having an erosion on the revenue that’s coming through the business.” 

Endava stayed optimized for hours. Turns out, when you’re faster but charge by the hour, you lose money. Capitalism working as intended.

The economics don’t have to work this way. But they require changing the model.

Right now, under T&M in an AI world, you’re both paying for the privilege of being miserable together. Your vendor loses money when they’re good at their job. You lose trust when they pretend to be bad. 

There are better options than T&M, much better suited for the current market situation. I’ll break down each one and show you the mechanics and implementation in the next article.

Sources

• Cognizant Q3 2025 earnings call (Oct. 29, 2025) 
• GitHub Copilot productivity study (arXiv 2302.06590) – Peng, Kalliamvakou, Cihon & Demirer, “The Impact of AI on Developer Productivity: Evidence from GitHub Copilot”
• Endava Q1 FY2026 earnings call (Nov. 11, 2025)