All three services offer AI but the interesting part about all of them is how differently they bill their customers.
First of all, ChatGPT employs flat monthly subscription plans and the customers essentially pay for access. Irrelevant of whether the service is used once a month or daily, people tend to pay similar amounts within their plans. OpenAI takes on the risk of heavy users consuming expensive computing resources, while customers bear the risk of getting inaccurate or unhelpful answers.
On the other hand, GitHub Copilot charges businesses per developer seat. Hence, companies essentially pay per seat for the right to use the product, which makes the revenue predictable for GitHub. At the same time, businesses will continue to pay even if they rarely use GitHub Copilot, while developers will continue being responsible for reviewing potentially incorrect code. Moreover, any additional usage could lead to further costs.
Finally, the most interesting billing model of Intercom Fin is per outcome, which means that customers are charged whenever a customer conversation is successfully closed. On one hand, the company itself bears the risk of having high costs of computing power or failure of the conversation to close. On the other hand, there is an issue with defining the conditions of successful conversation closure.
As far as the billing model for our own product is concerned, we could go for per-seat pricing with usage limits, as it is not easy to define whether the AI suggestion helped to make the decision. A seat-based model would be simpler for both us and our customers, while usage limits would help protect our margins.
