The AI services market is splitting. On one side are the headcount sellers, and on the other are the outcome engineers moving clients to outcome-based pricing. For 25+ years, IT services ran on one simple formula: rate x hours x people. The client had a problem, we sized a team, and the invoice followed the timesheet. That model built an industry. It also taught both sides a habit we now need to break: we measured effort and called it value.
AI is ending that, or at least starting that journey. It isn’t happening all at once, and some firms are still in denial, but the change is already underway.
What the headcount seller looks like in 2026
You’ve seen the pitch. “We have 2,000 certified GenAI engineers.” “We can ramp a 40-person pod in three weeks.” “Our AI Center of Excellence has trained 10,000 associates.”
Every one of those lines is about supply, and none of them is about results.
The problem goes deeper than marketing. The headcount model has a conflict of interest built into it. If AI makes a team of 20 do the work of 60, a firm paid by the seat has every reason to keep 60 seats on the invoice. So you get “AI-augmented delivery” that somehow never shrinks the team, and copilots rolled out across a program while the rate card stays the same.
Clients are catching on. CFOs are asking a blunt question in renewal meetings: “If your people are using AI, why is my bill the same?”
A headcount seller has no good answer to that.
What the outcome engineer looks like: outcome-based pricing in practice
The outcome engineer starts the conversation in a different place. The first question isn’t how many people the client needs. It’s what must be true in 90 days for this to count as a win.
In practice that means:
They price the result, not the resource. Claims processed per hour. Days taken out of the underwriting cycle. Defect leakage cut in half. Cost per test case. The fee is tied to a number the client’s business already tracks.
They treat AI as the delivery engine, not a side offering. Agents, automation and data pipelines do the repeatable work. People own the judgment, the edge cases, the change management and the accountability.
They get smaller over time on purpose. A well-run outcome engagement should need fewer people in year two than in year one. That isn’t lost revenue. It’s proof the thing works, and it earns the next scope.
They take on risk. That might be gain-share, fixed-fee milestones or a guaranteed SLA. They have skin in the game because they understand the system well enough to stand behind it.
Why this split is happening now
Three things came together.
First, the cost of producing things collapsed. Code, test scripts, documentation and data mappings, work that used to fill a staffing plan, can now be generated, reviewed and refined in a fraction of the time. When producing the work is cheap, charging for the time it takes stops making sense.
Second, buyers grew up. Two years of AI pilots taught enterprise leaders the difference between a demo and a production system. They’ve been burned by proofs of concept that went nowhere, and they now want someone accountable for getting from pilot to P&L impact. The shift is visible at the top of the market too: McKinsey says about a quarter of its global fees now come from outcome-based pricing.
Third, the talent itself changed. A strong engineer with the right AI tooling isn’t 20% more productive. On the right problem, they can be several times more productive. You can’t price that on an hourly rate card without either underpaying your best people or overcharging your clients.
Frost & Sullivan Best Practices 2026
Global Company of the Year for Digital CX
11 pages of independent analysis. Free PDF, one short form.
Download the report →What this means if you’re buying AI services
In banking and financial services, insurance, retail and consumer goods, the sectors I spend my time in, the smartest buyers are changing how they write RFPs. They’re asking vendors to commit to business metrics, not resumes. They’re asking how the team will shrink as automation matures. They want to see the vendor’s own AI stack in action, not a slide about it.
If your partner can’t tell you what outcome they’re accountable for, you’re buying headcount with an AI label on it.
What this means if you’re selling
I’ll be candid about my own side of the table: this shift is uncomfortable. Outcome-based models mean harder scoping, deeper discovery, real risk on our balance sheet and sales cycles that need both commercial and technical depth. You can’t sell an outcome without understanding the client’s operations better than the last firm that sent them a rate card.
But the alternative is worse. The headcount model is heading toward a price war against automation, and automation wins that war.
The firms that make it through this won’t be the ones with the biggest bench. They’ll be the ones that can walk into a claims operation, a data platform or a QA function and say: here’s the number we’ll move, here’s how, and here’s what we’re putting on the line.
The question I’d leave you with
Look at your top three services partners. Ask each one to describe their engagement with you in terms of the outcomes they own, not the people they’ve deployed.
Their answers will tell you which side of the split they’re on.
Which side is your organization buying from, or building toward?
Frequently asked questions
What is outcome-based pricing in AI services?
Outcome-based pricing ties a services partner’s fee to a business result the client already measures, such as claims processed per hour, days removed from an underwriting cycle or defect leakage reduced. The partner is paid for the result, not for the hours or headcount it supplies.
How is outcome-based pricing different from time and materials?
Time and materials bills for effort: rate x hours x people. Outcome-based pricing bills for results, often through gain-share, fixed-fee milestones or a guaranteed SLA. The partner carries delivery risk, and both sides benefit when AI lets a smaller team do the work.
How can I tell if my AI services partner is a headcount seller?
Ask them to describe the engagement in terms of the outcomes they own. A headcount seller answers with team size, certifications and ramp-up speed. An outcome engineer answers with the metric they will move, how they will move it and what they put at risk if they miss.
Want a partner that owns the outcome?
Compunnel’s Applied AI practice scopes every engagement around a business metric you already track, then builds the agents, automation and data pipelines to move it, with governance built in.
Explore Applied AI →
