Announcing outcome-based pricing takes an afternoon. Operating it takes a quarter, and most of that quarter is spent on questions you did not know existed when you wrote the press release.
Here is the order to answer them in.
Decision One: What Counts as an Outcome
This is 80% of the work and where nearly every dispute originates. Your definition must be specific enough that a customer and your finance team reading it separately reach the same number.
The reference implementation is Intercom Fin, where one resolution equals one outcome per conversation, regardless of how many messages or actions it takes. Notice what that definition does. It removes the argument about effort entirely, and it means a hard case and an easy case cost the same. Simple beats fair here, every time.
Decision Two: Who Decides It Happened
Three options, in descending order of how much your customers will trust them.
- The customer confirms. Highest trust, lowest revenue capture, because people forget to confirm.
- A measurable system event. The sweet spot. A ticket closed and not reopened within seven days is observable and hard to dispute.
- Your system decides. Cheapest to build, and the one that generates suspicion at every renewal.
Pick the middle option and publish the rule. Transparency about a mediocre rule beats silence about a good one.
Decision Three: What Happens on Failure
If the AI tries and fails, do you charge? The defensible answer is no, and saying so plainly is one of the strongest trust signals available in this pricing model.
Then define failure. An escalation to a human is usually a failure of the outcome even though work happened. Accept that. Charging for handovers is how you teach customers to distrust the meter.
The Remaining Nine, Briefly
| Decision | Common answer |
|---|---|
| Floor or platform fee | Yes. Pure consumption is unforecastable for both sides. |
| Volume tiers | Yes, with declining unit price |
| Ceiling or cap | Offer it. Cheap for you, valuable to them. |
| Overage handling | Notify before billing, always |
| Trial structure | Free outcomes, not free days |
| Attribution when a human helps | Split or waive. Do not fight over it. |
| Disputes process | Published, with a credit path |
| Reporting cadence | Real time. Monthly surprises kill renewals. |
| Contract minimums | Modest, or you have rebuilt the seat |
Where It Goes Wrong
Three failure modes repeat. Customers cannot forecast their bill, so procurement blocks the deal. The outcome definition is loose, so every invoice starts a negotiation. Or the vendor instruments outcomes badly and cannot prove what it charged for, which turns a routine audit into an existential one.
All three are solved before launch, not after. Once customers distrust the meter, no amount of reporting rebuilds it.
Conclusion
Define the outcome so tightly a stranger could count it, tie it to an observable system event, charge nothing on failure, and give customers a live view of the meter plus a cap they can choose. Add a platform floor so neither side is forecasting from zero. Do those five things and outcome pricing becomes an advantage. Skip any of them and it becomes a monthly argument with your best customers.
Frequently Asked Questions
Can we run outcome and seat pricing together?
Yes, and most successful transitions do. Keep the platform on seats, price the AI capability on outcomes, and let customers migrate at renewal rather than forcing it.
How do we handle a customer whose outcomes drop?
Treat it as a churn warning rather than a revenue problem. Falling outcomes usually mean falling usage, and that conversation should happen months before the renewal.
What if competitors undercut our per-outcome price?
Compete on resolution quality and reopen rate rather than headline price. A cheaper outcome that comes back next week is more expensive, and sophisticated buyers already measure this.