What performance-based AI consulting actually means
How an engagement works when the consultant is paid from verified results instead of billed hours.
Austin Vu · Managing Partner, Qonnex
Most consulting is sold on confidence and billed on time. The proposal describes an outcome; the invoice describes hours. Whatever happens in between, the meter runs.
Performance-based consulting inverts that. The firm's revenue depends on a measured result, so the entire engagement is built backwards from one question: what will we be able to verify?
The mechanics, in plain terms
A performance-based engagement has four load-bearing parts.
A baseline, agreed before work starts. You cannot verify improvement against a moving target. Serious performance-based firms lock the "before" picture in writing: current costs, current cycle times, current volumes. If a firm resists defining the baseline precisely, that is a warning sign, because a vague baseline is what lets bad actors claim credit for savings that would have happened anyway.
A written definition of savings. Labor hours reclaimed count, or they don't. Software licenses cancelled count, or they don't. The definition is negotiated once, up front, while incentives are still neutral. This is the piece that protects both sides.
A measurement window. Results are compared over a fixed period, typically one to three months, long enough to smooth out seasonality and one-off effects. Shorter than that and both sides are arguing about noise.
Settlement from verified results only. The consultant's payment is calculated from the verified difference between baseline and outcome. Not projections, not estimates presented in a slide, the verified difference.
Traditional consulting
- Proposal
- Invoice by hours
- Delivery
- Outcome unknown
Performance-based
- Free analysis
- Agreement
- Implementation
- Verification
- Settlement
Why firms take this risk
The honest answer: selection. A firm paid from results cannot afford engagements that produce none, so it invests heavily in diagnosis before commitment. That is why performance-based firms tend to lead with analysis rather than sales calls. The diagnostic is not a marketing gesture; it is the firm protecting itself from bad engagements, and the client benefits from the same filter.
It also changes behavior during delivery. When revenue depends on the measured outcome, scope discussions get shorter and implementation gets more pragmatic. Nobody paid from results builds a system nobody uses.
Where the model fits, and where it doesn't
The model works where results can be measured in money within a reasonable window: operational cost reduction, process automation, error reduction, response-time improvements that convert to retained revenue. It struggles where outcomes are diffuse or long-horizon, like brand building, or where the client cannot share the data needed for verification.
That boundary is worth taking seriously. A firm offering performance pricing on something unmeasurable is either planning to argue about attribution later or planning to lose money. Neither is a good partner.
Questions worth asking any performance-based firm
Before signing anything, ask four things. How exactly is the baseline established, and do I see it before work starts? What is the written definition of savings for my case? Who runs the measurement, and do I review results before anything is settled? What happens if no savings are found?
The answers should be specific and boring. In this model, boring is what protection looks like.