ChatGPT has a flat subscription model that features different tiers with varying levels of AI models and features. This means users pay for access, regardless of how frequently they are using the product. OpenAI takes on this risk of expensive computational costs, though they have tried to manage this through usage limits. Users also carry some risk in the event that the AI is wrong (for example when they are using a lower tier model) since they may end up spending additional time verifying or correcting inaccurate outputs.
With GitHub Copilot, they opt for a per seat pricing model where an organization pays for each developers’ access to the coding tool. This means that the costs will be easy to calculate for companies, though if a developer uses more than the average, there may be higher costs. This is similar to ChatGPT since customers (in this case the organization) pays regardless of whether the AI code is useful. Developers are also responsible for verifying the accuracy of generated code, meaning the risk of incorrect outputs largely falls on the user.
Finally, Intercom Fin takes the most different approach by charging per successful outcome, for example through resolved customer conversation. This aligns pricing with the value delivered, meaning the financial risk is on Intercom Fin when conversations are unsuccessful. Additionally, Intercom takes on the risk of expensive computational costs for conversations that require multiple interactions but ultimately go unresolved.
For our product, which helps PMs gain visibility into what other teams are launching, I think a per-seat model would be most appropriate. Similar to GitHub Copilot, our product aims to improve existing workflows rather than automate a single measurable outcome. Since the value comes from continuous visibility and better cross-team coordination, it would be difficult to define a successful “resolution.”
