While ChatGPT, Copilot and Fin are all AI products, they have different pricing plans. ChatGPT is basically charging for access. Individuals pick a flat tier like Plus or Pro, and companies pay per seat and also share a pool of credits that everyone pulls from once they go over their included usage. They also recently added a $100 Pro tier and paused the $200 one for new users, which I think is kind of a gamble to get heavy users to move up from Plus. This makes sense with the reading, since most of the token cost comes from a small group of power users, so a flat fee means OpenAI eats the cost while the user still pays the same even if the answer is wrong.
Copilot is charging for a person, $19 a seat for Business, but since June each seat comes with credits and anything past that is billed by usage, so GitHub kept seats but slowly moved the cost risk onto the customer.
Fin is the only one charging for an outcome, $0.99 per resolved conversation, so Intercom takes on more of the risk of the AI being wrong. But even then, if a customer just leaves the chat it still counts as resolved, which goes back to the reading’s point that outcomes are hard to actually measure.
For our product, Rooted, we market it per seat right now since a company only has a fixed number of PMs, but AI usage can be really different from one PM to another, so I think we should do something closer to Copilot and add AI credits on top of the seat, while keeping it to one simple credit system so it doesn’t get confusing.
