The future of AI pricing

🌈 Abstract

The article discusses the pricing strategies for AI-based productivity tools, highlighting the limitations of seat-based pricing and the advantages of work-based pricing.

🙋 Q&A

[01] Pricing AI Products

1. Why is seat-based pricing relatively undesirable for AI tools?

• AI-based productivity tools do not follow the same pattern as traditional work productivity tools, where the value is derived from having the whole company integrated.
• The value of AI tools is in the automation and work they provide, which can vary significantly among users. Charging based on seats rather than work done is not an accurate reflection of the value provided.

2. What is the key principle behind work-based pricing for AI products?

• The value of AI tools is in the automation and work they provide, so pricing should be based on the amount of work done rather than the number of users.
• Examples include RunLLM pricing based on the number of questions answered, and AI-based SDRs pricing based on meetings booked.

3. What are the challenges in implementing work-based pricing for AI products?

• Defining what constitutes "work done" can be complex, with different options like meetings booked, meetings held, or meetings converted.
• Earning customer trust that the AI will behave as expected and not make mistakes is important, as mistakes by an AI are scrutinized more closely than human errors.
• Aligning enterprise budgets with consumption-based pricing can be tricky, though a tiered usage-based model can help address this.
• Determining the appropriate pricing for the value of the work done, as it is a high-margin, low-volume business compared to cloud infrastructure pricing.

[02] Exceptions to Work-Based Pricing

1. Why do ChatGPT and GitHub Copilot use seat-based pricing instead of work-based pricing?

• Predicting usage of these tools is extremely difficult, so a work-based pricing model could create a negative incentive for users.
• Quantifying the "work" done by these tools, such as determining whether a task was completed successfully, is more challenging compared to other AI use cases.
• The relatively low cost and generic nature of the tasks in these cases means that seat-based pricing is likely to continue working in the short term.

2. How might these exceptions evolve in the future?

• As the market matures, products like Copilot may move towards more holistic task completion and charge accordingly by work done, rather than a fixed seat-based price.

[03] The Future of AI Pricing

1. What is the overall direction for AI product pricing?

• The article suggests that work-based pricing is the direction that AI is headed, particularly for enterprise use cases, and potentially for consumer technology as well.
• This shift towards work-based pricing may enable ubiquitous micro-transactions on the internet, as AI becomes more capable of completing various tasks.

2. What are the next steps in the article's exploration of AI pricing?

• The article mentions that there is a sub-topic around implementing usage-based billing and the services available to automate this process, which the authors have not yet formulated their opinions on.
• This is an area they plan to explore further, as it is not specific to just AI products.