Outcome-based pricing for AI-run services only works if buyer and provider agree the outcome definition, the baseline and how each result is verified. This guide helps operations, finance and procurement leaders settle those terms and choose a first process before signing.
This paper establishes that the services market is moving from pricing by headcount to pricing by result, but the contracts and proof have not kept up. An outcome price only works if both sides agree what the outcome is, what the starting point was, and how each result is checked. Most contracts answer none of the three.
How far has the market moved towards outcome pricing?
One HFS Research study with Firstsource expects contracts priced on headcount to fall from 42% of engagements to 28% within three years, while outcome-based and shared-revenue models rise from 20% to 39%. In HFS research with Cognizant, 101 US business and functional leaders ranked a base fee plus outcome payment first (56%) and time and materials last (18%). Multiple answers were allowed.
Why do outcome contracts stall?
In the HFS and EY survey of 304 Global 2000 executives, 72% said they are comfortable with outcome pricing when AI does much of the work, but only 39% have used it in the last two years and 73% frequently use time and materials for AI-led work. Only 13% evaluate AI-led proposals against an enterprise-wide framework. The paper names three gaps: no agreed definition, no trusted baseline, and no way to check each billed outcome.
What should be settled before signing?
The paper sets out five questions that need written answers in the contract:
- What exactly counts as an outcome, written as a test a third party could run, with exclusions.
- What is the baseline, and who measured it from the buyer's own systems.
- How is each outcome verified, and from what record.
- What is not billed, such as agent time, retries and failed items.
- Which decisions stay with the buyer.
It also gives a cost-per-outcome formula, a specification for an evidence record, a comparison of four commercial models, a twelve-point contract checklist and six tests for choosing the first process.
Key takeaways
- HFS Research expects headcount-priced contracts to fall from 42% to 28% of engagements in three years, while outcome-based and shared-revenue models rise from 20% to 39%.
- 72% of surveyed Global 2000 executives are comfortable with outcome pricing, yet only 39% have used it in the last two years and 73% frequently use time and materials.
- An outcome price moves risk to the provider only if the buyer can check the outcome; otherwise the buyer swaps a rate card it understood for a count it cannot audit.
- Cost per outcome should be divided by outcomes that meet the definition, not items touched: if 88 of 100 touched items pass first time, the cost belongs to the 88.
- A floor plus verified outcomes is the paper's preferred structure, because an outcome-only provider with no floor has reason to define success loosely or price the risk in.
- Every billed line should link to an evidence record that is written as the work happens, append-only, hash-chained and owned by the buyer, who can export it.
What is in the full paper
- The market: Services are being repriced around results
- The gap: Why outcome contracts stall
- Before you sign: Five questions to settle first
- The record: What an evidence record has to hold
- The operating model: Agents, an exception desk, and your approvals
- The commercial model: A floor, plus verified outcomes
- Where to start: Choosing the first process
- Where agent-run services fit today
- A practical path
The full paper has the figures, the sources and the working.
Download the PDF
