ChatGPT, Github Copilot, Intercom Fin have similar pricing models with one small difference.
For individual customers, they first charge based on flat subscription tiers that give access to their LLM with usage limits. For businesses, they first charge per seat with usage limits that increase as you pay for the higher tiers, so they are charging per person. Intercom Fin only charges per seat and to use as an individual you would buy one seat.
The small difference is what happens when you hit your usage limit. ChatGPT and GitHub Copilot allow you to buy more credits to use. On the other hand, Intercom Fin charges per resolved customer resolution, paying for a specific outcome. In the Intercom Fin model, you only pay when you type in that the desired outcome was achieved, which is once per conversation even if multiple questions are answered.
When the AI is wrong or is expensive to run, users of ChatGPT and GitHub Copilot carry the risk of being charged more because they pay more for better models and asking more questions, while Intercom Fin doesn’t charge based on the model or asking more questions in one conversation.
For our team’s product, per seat with usage limits and the option to buy more credits would make the most sense since our ICP is large tech businesses, so the pricing should scale with more requests to the AI. Following the suggestions of the reading, this means as a per seat price model we should add a “lite user” seat and require a min number of users (Poyar). The reading also suggests that as a hybrid pricing model, we should introduce a unified credit model, continually pare down the actions that cost credits, and translate pricing into segment-specific offers.
