── ── AI & agents
Outcome-Based AI Pricing
When AI replaces human work, the customer needs fewer seats — so seat pricing moves in the opposite direction from the value delivered. The more the product works, the less it earns. The charge unit has to move to a measure of value the customer recognizes: a usage unit close to the work, or the outcome itself.
Run Outcome-Based AI Pricing on a real problem
Bring something you're actually deciding — free, in the browser.
How it works
1. Choose the value metric. Candidates: seat · usage (calls, tasks, documents) · outcome (resolved tickets, qualified leads, hours saved). Score each 1–5 on four tests: measurable, attributable, grows with value, budgetable by the customer. Gate: no candidate scores ≥3 on all four → use fixed + usage hybrid → stop. 2. Define success and attribution. Write down what counts as one success, whose data decides it, the dispute window, and exclusions (results the customer caused). Gate: both sides cannot reconcile against the same data source → fall back to usage pricing. 3. Anchor the price. Use the customer's current replacement cost (staff or outsourcing per unit) as the upper anchor and take a share of it. Never price as a markup on inference cost. 4. Check gross margin. Per unit: inference + human review + infrastructure, computed for P50 and P90 tasks. Gate: P90 unit margin is negative → cap the unit, tier it, or change the unit. 5. Share the risk. Monthly minimum commitment (floor) · spend cap (ceiling) · tiered overage price · guaranteed minimum during pilots. 6. Plan the migration (if coming from seats). New customers first → existing customers opt in → switch at renewal. Provide a bill simulator and transition price protection. Gate: the simulation shows most existing customers' bills jumping → adjust; do not force it. 7. Stop-rule: attribution disputes stay high → return to usage pricing and keep the outcome as a reported metric only.
Output template: charge unit · success definition + deciding data source · unit price · floor / cap · P50 / P90 unit margin · replacement-cost anchor · migration pace · review date
→ Method in Action: Intercom Fin's $0.99 per resolution (2023–2026)
When to use it
- user says 'how should we price our AI product?', 'per seat or per usage / per outcome?', 'how do we design pay-for-results pricing?', 'inference costs are eating our margin', 'price per resolution', or AI is shrinking the number of people a customer needs, so seat pricing has started to cannibalize itself
When not to use it
When the decision is routine and reversible, applying a formal method costs more than it returns.
Worked example
Intercom Fin's $0.99 per Resolution (2023–2026)
Intercom sells customer-service software priced per seat. In March 2023 it launched Fin, an AI agent that answers customers directly — the exact product that shrinks the number of support seats a customer needs. Instead of folding Fin into the seat price, Intercom charged per resolution. The definition has been refined publicly ever since, and the current billing rules read like a worked answer to this skill's steps 2 and 5. This is the most-cited public example of outcome pricing for an AI agent.
Install this skill (free, MIT)
npx skills add deciqAI/knowledge-skillsUseful? Star the repo — stars help other builders find it.
Related mental models
In the AI era, models and compute can be bought.
Answer engines — featured snippets, "People Also Ask" boxes, voice assistants, AI answer panels — lift one passage and present it as the answer.
Teams usually decide how much to hand to an agent by watching a demo.
An agent that is live without logs, quality metrics or alerts fails quietly.