The AI Pricing Dilemma: Access, Seats, or Outcomes?
There are many different ways to price your AI:
-
ChatGPT (Flat Subscription): You are charging for access. These flat-fee models are simple and predictable, but the vendor carries the financial risk. If a user leaves the AI running constantly, the vendor is responsible for the compute costs, creating “low/negative margin customers”.
-
GitHub Copilot (Per Seat): You are charging for a person. Seat-based pricing is increasingly viewed as a “dinosaur” because the value of AI doesn’t scale linearly with headcount. The customer carries the risk here, because you pay the same monthly fee whether the AI writes brilliant code or generates useless errors that waste your time.
-
Intercom Fin (Per Resolution): You are charging for an outcome. This is often seen as the “ultimate win-win” because buyers only pay for tangible business impact. The vendor carries the risk of the AI being wrong because if the AI hallucinates or fails to resolve the customer’s issue, the vendor doesn’t get paid.
Which model fits our product?
For our product, perhaps we should adopt a hybrid outcome model. While pure outcome pricing sounds ideal, it is often impossible in practice because customers will dispute with us what technically qualifies as a successful resolution.
Instead, we should implement a base platform fee combined with a smaller outcome-based bonus. This structure guarantees our baseline compute costs are covered if the AI is expensive to run, while still aligning our revenue upside with the actual value we deliver to the user.
