psp – romir jain
One thing I find interesting about AI companies is how differently they charge customers, even though they are all selling AI products. ChatGPT, GitHub Copilot, and Intercom Fin are three good examples. ChatGPT uses a subscription model, where users pay a flat monthly fee to access certain features and models. This makes costs predictable for customers, but OpenAI takes on the risk if someone uses the product a lot and costs more to serve. If ChatGPT gives a wrong answer, the user still pays the same amount. GitHub Copilot charges per seat, meaning companies pay for each developer using the product. This makes sense because Copilot helps individual developers write code faster. However, companies still pay even if developers barely use it or the code is incorrect. GitHub also has to manage the cost of running the AI. Intercom Fin takes a different approach by charging per resolution. Instead of paying for access or users, companies pay when Fin successfully resolves a customer conversation. I think this is interesting because the pricing is directly connected to the value delivered. Intercom takes on more risk because it doesn’t get paid for unsuccessful conversations, even though those still cost money to process. For our team’s product, I would lean toward an outcome-based model like Fin, assuming we can clearly measure success. Customers would be more willing to pay if they knew they were getting results. The challenge is defining what counts as a successful outcome.
