ChatGPT generally charges a flat monthly subscription price, but if you’re on a team under a business plan it charges per seat. This model charges for access, particularly access to more compute. If the AI is wrong or expensive, ChatGPT carries the risk. This is especially true for people on the flat monthly subscription. There are some subscription offerings where you pay on a per token basis. Here, the risk falls directly on the user as they are paying more as usage increases.
GitHub Copilot charges per seat, with an additional pool of credits. This model charges for access, since users have to pay to use the platform. This means that individual users are billed more if they go over their pool of credits, or will just be blocked by admin. This places more cost on the individual or business.
Intercom Fin is a per customer conversation model. This means that you pay money as a user when Fin resolves a customer’s issue. This means that the individual is paying partially for a happy customer (or person) and for an outcome. This means the vendor is taking more risks, since if the AI is incapable of resolving customers’ issues they won’t make money.
The model that fits our team’s product the best is a per seat basis, since we will use more compute for bigger teams than we will for smaller teams. I think it’s too difficult to build our financial model around an outcome since the outcomes are hard to set firm boundaries around and quantify.
