For ChatGPT it charges for access, Github charges per seat, and Intercom Fin charges per outcome. For ChatGPT, the company itself (OpenAI) is carrying the risk of AI costs, and the accuracy of their models. Github’s costs are shared between the provider and the customer through extra credit usage and the limited allotted allowance. For Intercom Fin, the provider would absorb the costs per outcome that are not successful. For our product, I believe that relying on a frontier model such as ChatGPT, that absorbs the AI costs within plan limits, so the user can actually have a lot of bandwidth to interact with the model without needing extra usage credits, for everyone in the company.
