Week 7 PSP: Three Ways to Price AI

Kyle Poyar suggests that conventional pricing schemes are often flawed, so the question is how do particularly successful companies hold up? ChatGPT, GitHub Copilot, and Intercom/Fin are three examples of varied approaches to pricing.

For ChatGPT, the charge is for access, as a Pro subscription is a flat payment on a recurring monthly basis. The company carries the risk, as they have to absorb the compute and usage for power users (Altman has admitted that OpenAI spent double the company’s revenue running inference in 2025 [3]). However, the company mitigates this risk with incremental and longer-term usage limits per account, so that upper-limit usage caps can be adjusted by the company.

GitHub Copilot charges corporations on a per-person basis for each team member [1]. Employers or team leaders therefore bear a lot of the risk and overhead, since they are the ones in charge of managing and allocating resources. A downside is that highly active users and inactive users cost the company the same. Some risk mitigation exists in the control of pooled resources and organization-level policies of who gets access.

Fin is quite distinct as the company charges $0.99 for an outcome, since its prices are based on the amount of resolutions achieved [2]. A lot of the risk is carried by the company, since they have to deliver outcomes as opposed to a service or process. Costs to the customer are mitigated as the number of resolutions is capped per instance, but the company’s monetary benefits heavily rely on outcome success.

For our team’s product (a decision-tracing tool for PMs), I believe that a per-seat basis makes sense if we consider company onboarding. However, we recognize that some people might need certain parts of the tool for individual-level work, we can also provide a hybrid option. Decision-tracing isn’t very outcome-oriented, so we would not want to rely on outcomes. Key considerations for us include the customers’ range of usage and the stakeholders we want downstream financial impacts to fall on. 

 

[1] https://github.com/features/copilot/plans 

[2] https://www.intercom.com/learning-center/intercom-pricing 

[3] https://www.koncentrik.co/p/the-broken-economics-of-ai

Avatar

About the author

Leave a Reply