ChatGPT runs on fixed monthly tiers, and charges for higher capability tiers, unlimited prompts, and advanced aspects, like unlimited upgrades. Because of this, if the user maximizes their limits, OpenAI pays the price. On the flip side, if the AI is wrong, the user still pays the flat fee. In a different lane, Github Copilot relies on a monthly fee/developer seat + AI tokens and charges for a person in the loop and compute consumption. If something goes wrong or the user goes past base completions, the customer pays through credit overages. For the last one, Intercom Fin, it relies on a usage-based fee for each resolved customer conversation, and charges for the outcome. Because of that, Intercom pays for the tokens and steps and absorbs the computer cost if it has to escalate to a human.
I think a usage-based pricing model makes the most sense for us. Companies would be charged for each simulation they run, while still having access to all the information generated and the databases at no additional cost. While we would have costs for the tokens and processing steps behind each simulation, individual contracts with companies would allow us to account for those expenses and maintain a sustainable and profitable model.
