Emineon

Commercial

AI is changing the economics of consulting. Hourly billing rewards effort, and buyers want to pay for results. This is how we structure that fairly for both sides.

Published
Reading time
4 min read
By
Emineon · Strategy practice
Abstract cover: a balanced torus of white dots, its two halves in equilibrium, on a grainy ember and steel-blue gradient.
On this page
  1. 01Three layers
  2. 02When outcome pricing works
  3. 03How it fails
  4. 04How it changes the work
  5. 05What to ask before you agree

Professional services have long been sold by the hour. That made sense when the work was mainly human effort and the client could not easily judge the output. It makes less sense now. A small team using AI can produce in a week what used to take a month. If hours fall and the price stays tied to them, the firm’s revenue falls with its own productivity. If the price stays the same while hours fall, the client pays for effort that no longer happens.

Outcome-based pricing is a natural answer, and large firms are moving towards it. It is also easy to get wrong. This article explains how we structure it, when it works and what a buyer should ask before agreeing to it.

Three layers

No single outcome-pricing model fits each stage of an AI programme. We use three layers, each matched to how much is known at that point.

Fixed fee for diagnosis

At the start, the outcome is not yet known, so it cannot be priced. The work of finding out can be priced. Interviews, baselines, business cases, a policy and a roadmap. A fixed fee for a fixed scope removes the risk for the buyer and keeps the firm disciplined. It also produces what later pricing depends on, a measured baseline.

Milestones and workflows live

For the build, we price per milestone or per workflow in production. Each milestone has a test anyone can run. The eval set reaches the agreed pass rate. The system is live in the production environment. The team is trained and operates it. This is not yet payment for business results. It is payment for progress you can check.

A share of the measured result

Where the baseline is clean and the result is attributable, part of the fee can depend on the outcome. Hours saved on a process, shorter cycle time or fewer errors reaching the client. Incentives align most here, and this is also where the most care is needed.

When outcome pricing works

Outcome pricing works when four conditions are met. If one is missing, it leads to disputes.

  1. The baseline is measured before the work starts, with data both sides accept.
  2. The result can be attributed to the system, and not to market conditions, a reorganisation or a seasonal dip.
  3. The client controls the conditions for success. Access to data, people and systems, and the decision to roll out.
  4. The measurement period is long enough to mean something and short enough to settle.

Back-office workflows usually meet all four. Handling time per confirmation, per KYC review or per regulatory draft is measurable, attributable and controlled. Revenue outcomes, such as more deals closed or higher client retention, usually fail the attribution test. We do not price on them.

How it fails

We have seen outcome-based arrangements go wrong in a few recurring ways.

  • The baseline is estimated instead of measured. Months later, nobody agrees what “before” looked like.
  • The metric can be gamed. If the fee depends on tickets closed, tickets will be closed. Pair volume metrics with quality checks.
  • Adoption is outside the firm’s control. A system that works but is not rolled out shows no measured saving. The contract must say what happens then.
  • There is no cap. Uncapped success fees cause friction when the result is large. Agree a ceiling in advance.
  • The whole fee is at risk. A firm in that position will pick only the safest work. A sensible mix keeps a fixed part for the effort and a variable part for the result.
Outcome pricing is an agreement on what both sides want to achieve, written down before the work begins.

How it changes the work

The main effect of outcome pricing is on behaviour. A firm paid by the hour has little reason to finish early or to build something the client can run alone. A firm paid per workflow live has good reasons to choose a feasible scope and write the eval set early. It will also design for exceptions and train the client team. These are the steps that get a system into production.

It also changes the shape of the team. Outcome pricing suits small senior teams that decide quickly, supported by AI and by engineering capacity that grows with the work. Large teams of juniors learning on the client’s time fit it poorly. This changes how consulting is organised, and it is one reason we built Emineon this way.

What to ask before you agree

If you buy AI work on an outcome basis, ask these questions before signing.

  • How will the baseline be measured, by whom, and over what period?
  • Which metric will the fee depend on, and what quality check sits next to it?
  • What is fixed and what is variable, and what is the cap on the variable part?
  • What happens if we decide not to roll out a system that meets its acceptance criteria?
  • Who owns the code, the prompts and the evaluation sets at the end?
  • How are model and infrastructure costs handled, passed through at cost or bundled?

A firm with experience should have clear answers to each of them. The answers also show whether the firm has put systems into production before.

Field notes

One short note a week on putting AI into production. We do not use tracking pixels.

One email per new article. You confirm first and can leave in one click.

PrivacyRSS feed