- ChatGPT
ChatGPT charges flat subscription tiers with $0 for Free, $8/month for Go, $20/month for Plus, and $100/month or $200/month for Pro. This charges for access. The user carries the risk when AI is wrong since they pay for a certain amount of usage, regardless of the correctness of an answer. OpenAI carries the cost of running expensive AI but has usage limits for each tier, which is one of the flat-fee fixes mentioned in Poyar’s reading.
- GitHub Copilot
GitHub Copilot has Copilot Business, which charges $19 per seat per month, including 1,900 credits per user. This charges for a person. Since GitHub charges for usage by deducting from the 1,900 credits, the company absorbs costs of expensive AI within that limit, but the user carries the risk of expensive AI beyond that threshold. The risk of wrong AI falls on the customer since it draws from their credits.
- Intercom Fin
Intercom Fin charges $0.99 per resolution with a monthly minimum of 50 resolutions, so it charges for an outcome. Fin carries the risk when AI is wrong or expensive because it only has a successful resolution when the AI completes a query without requiring human intervention. If it fails, Fin doesn’t make money but still has to pay compute costs.
Intercom Fin’s pricing model best fits our product since we can count the number of IT tickets resolved by our tool. We can also charge a platform integration fee in addition to the per-resolution charge so our revenue is more reliable.
