AI consultants who only charge if it works: how outcome-based pricing works

Does anyone actually price this way

The question arrives in a skeptical tone, and the skepticism is earned. Most "pay for results" offers in AI consulting are marketing with a refund clause nobody has read. The model itself is real, and it has moved upmarket fast. In December 2025, Hunt Scanlon reported that McKinsey now generates about a quarter of its global fees from outcomes-based pricing, with clients tying compensation directly to whether the firm delivers the result.

The same shift is visible at the other end of the market. Outcome-based and fixed-fee models appear in most 2026 rate guides beside hourly rates, and Layer3 Labs lists a fixed discovery price and fixed project bands rather than an hourly menu. The reason is the same at every size: AI compresses the hours a piece of work takes, and a fee built on hours shrinks with it.

Why the hour is losing

An hourly fee pays the consultant for time, and time is the one thing the client never wanted. The client wanted a process fixed. When the work takes forty hours, the hourly model charges forty. When better tools cut it to ten, the model charges ten for the same result, and the consultant has a reason to keep it at forty.

A fixed project price fixes that. An outcome price goes one step further and makes the consultant carry the risk that the fix does not work. That is a fair trade only when the consultant can actually carry it, which is what the four conditions are about.

The four conditions a guarantee needs

Strip away the contract language and every outcome-based fee that holds up rests on four fixed points. If any one is missing, the guarantee is a slogan.

1. A baseline. The number before the work: hours a week on the task, fees a month, errors a quarter. Measured the same way the result will be measured. Without it, nobody can say the outcome happened.
2. A window. When the result gets measured and for how long. A real window is short enough to test and long enough to see a normal week. Thirty to ninety days is the range most outcome contracts use, and a single good day is not a result.
3. Control. The consultant has to control enough of the work to be accountable for its outcome. Consultancy.uk put it plainly in August 2026: expecting a consultancy to own an outcome without giving it decision-making authority is like asking someone to open a door while you keep the key. If the outcome depends on your team adopting a tool, the contract needs to say what adoption means and who drives it.
4. A definition of "works." The result in one sentence a customer could check. "Five hours back across the team in a normal week" is one. "Improved efficiency" is not.

The failure research says the same thing from the other side. RAND's first root cause of failed AI projects is that the business and the builders never agree on the problem. An outcome fee forces that agreement before any money moves, which is most of its value.

How guarantees go wrong

• The outcome is undefined. "Results" or "value" with no number. The dispute at the end is about what was promised, and the consultant wrote the promise.
• The baseline is soft. Nobody measured the before, so the after is compared to a guess. A guess can be moved.
• Attribution is unbounded. A success fee on "verified gains" where verification is the consultant's job and the gains include things your team did on its own.
• The scope grows and the outcome does not. The measured result stays fixed while the work needed to reach it doubles. In an hourly contract that is your problem. In an outcome contract it is the consultant's, which is the point, so the consultant will try to reopen the scope.
• The refund has conditions nobody read. A guarantee that requires your team to complete a training course, adopt three tools, and log usage daily is a guarantee against your team, not for it.

None of these is exotic. MIT's 2025 report found that about 5 percent of generative AI pilots produced rapid revenue gains and the rest stalled, and that buying from a specialized vendor or partner succeeded about 67 percent of the time against a third of that for internal builds. An outcome fee is one way to make the partner's success rate your problem to verify and their problem to deliver.

How to read one before you sign

Six questions, and each has a one-sentence right answer.

1. What is the number today, and who measured it?
2. What is the number that counts as "works," and over what period?
3. What do you control, and what do you need from my team, in writing?
4. What happens if we get halfway there?
5. What is excluded from the outcome: licenses, data work, maintenance?
6. When is the money paid, and when is it returned?

Our terms, in full

Buy Your Time Back is our entry engagement, and it is the offer this article's title describes. The terms, as published on that page, are these.

• Price. $995, fixed. You pay only after we confirm the business is a fit, through a payment link sent before discovery begins.
• Baseline. A 45-minute session on how the work moves in a normal week, which systems the team uses, and where repeated effort accumulates.
• The work. Within about a week, a path built from existing tools that fit your constraints, and help getting the team to use them. No custom build.
• The outcome. Five hours back each week across the team in a normal week, counted after any review, correction, maintenance, oversight, and exception work. Not five hours per employee.
• The window. The agreed test, plus one adjustment if the first path falls short.
• The guarantee. If the team does not get five hours back after the agreed test and one adjustment, we refund every dollar.
• Fit. Owner-led service businesses with roughly 5 to 25 employees. Not a fit: regulated or high-stakes workflows, work that needs a custom build, or work that cannot be observed repeatedly.

Held against the four conditions: the baseline is the 45-minute session, the window is the agreed test plus one adjustment, we control the tool selection and the training, and "works" is five hours in a normal week. The fit rules exist because outcome pricing fails without control, and we do not control a regulated workflow or a build that has not been scoped.

When outcome pricing is the wrong tool

It is the wrong tool for a custom build whose scope nobody has seen yet, because the fourth condition cannot be met before the work is understood. Our custom AI systems are priced after we have seen the process, the software it touches, the data boundary, and the people who will own it, and you receive the written scope, schedule, and price before work starts. The dental payment integration is the shape of that: the outcome was a system running in production, and we do not publish a savings figure for it because the practice has not shared one we can stand behind.

It is also the wrong tool for regulated or high-stakes work, where the correct outcome is often that the system stops and hands the case to a person. You cannot put a refund on judgment. What you can do is put the fee on a scope and the stop conditions in writing, which is what we do there.

A guarantee is a statement about control. Read it as one, and it tells you what the consultant actually does.

Frequently asked questions

Questions owners ask us

Is there an AI consultant who only charges if the project actually saves time?

Yes. Buy Your Time Back, our entry engagement, costs $995, payable after we confirm the business is a fit, and is refunded in full if the team does not get five hours back in a normal week after the agreed test and one adjustment. Larger firms use outcome-based fees too: McKinsey reported about a quarter of its global fees are tied to outcomes.

What is outcome-based pricing in consulting?

A fee tied to a measured result rather than to hours or deliverables. It works when the baseline, the measurement window, what the consultant controls, and the definition of the result are all fixed in writing before work starts. Without those four, it is a fixed fee with a marketing line attached.

What does "five hours back" actually mean?

Five hours a week across the whole team in a normal week, counted after any review, correction, maintenance, oversight, and exception work the new tools add. It does not mean five hours for every employee. We verify it against the baseline from the first session.

What happens if the guarantee is not met?

First, one adjustment: we change the path and test again. If the team still does not have five hours back after that, we refund every dollar you paid. There is no course to complete or usage log to keep as a condition.

Why pay before the work instead of after?

Payment happens after we confirm fit and before discovery, because the guarantee is the refund, not a deferred invoice. Confirming fit first is how we keep the refund rare: businesses with regulated workflows or work that needs a custom build are told so on the fit call and not charged.

Can a custom AI build be priced on outcomes?

Rarely, and not before it is scoped. A custom build crosses systems and depends on data nobody has examined yet, so the definition of "works" cannot be fixed up front. We price builds after seeing the process and put the scope, schedule, price, and stop conditions in writing before work starts.

What this is based on

Sources

The market figures are quoted from published reporting; our terms are quoted from the offer page. Checked September 5, 2026.

1. Hunt Scanlon Media, "McKinsey Continues to Deliver Value; It Just Charges Differently for it Now". December 8, 2025. The quarter-of-fees figure, as reported to Business Insider.
2. Consultancy.uk, "What it takes to make outcome-based pricing work in consulting". August 27, 2026. Control and accountability as the condition for outcome fees.
3. RAND Corporation, "The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed". Misalignment on the problem as the first root cause.
4. Fortune on MIT NANDA, "The GenAI Divide: State of AI in Business 2025". August 18, 2025. Pilot success rates and vendor-versus-internal outcomes.
5. Layer3 Labs, "AI Consulting Rates and Pricing in 2026". Updated June 17, 2026. Fixed-fee bands beside hourly rates.
6. Buy Your Time Back. Our terms, as published in September 2026.