ChatGPT charges for access to the LLM and its capabilities because it has flat subscription tiers. Because there are usage limits for the different tiers in ChatGPT, the user carries the risk if the AI is wrong or the query takes lots of tokens because this pulls from their monthly usage limit.
GitHub Copilot charges for a person given that it charges per seat at a company. The company carries the risk when AI is wrong or expensive to run because users do not have to pay more than their seat fee for additional queries and expensive tokens.
Intercom Fin charges for the outcome of successfully resolving a customer conversation. The company carries the risk when that AI is wrong because unsuccessful customer conversation resolutions do not result in a charge, so they would not receive money and would still have to pay for the expensive AI cost.
For Quicken, I think it would be best to follow ChatGPT’s model and charge flat subscription rates because initially, users who buy subscriptions are likely the people who lead meetings more often. Additionally, this way we do not need to worry about charging per seat and whether the chicken pop-ups and summaries are available to everyone in the meeting or just the Quicken user. Charging per seat could also potentially obscure the values of Quicken if multiple people in the meeting are using Quicken. Charging for an outcome would be difficult for Quicken because it is hard to measure what makes a meeting “good” or “effective,” and thus it would be difficult to only charge users for certain meetings based on a standard that is difficult to quantify and generalize.
