AI-native companies face significant challenges in determining the right pricing model for their AI features. Traditional pricing models like flat subscriptions or usage-based pricing do not adequately address the unique cost structures and customer concerns associated with AI services.
This issue affects AI companies and their customers, leading to unpredictable costs, customer dissatisfaction, and potential loss of margins for the companies.
Pain Points
- Customers dislike token-based pricing.
- Flat subscriptions can lead to margin loss with heavy users.
- Usage-based pricing causes anxiety as usage meters climb.
- Complex AI workflows can be significantly more expensive than simple queries.
- Difficulty in defining and implementing outcome-based pricing.
I work at an AI-native company and pricing has been the thing keeping me up at night more than any technical problem we've faced. Our customers hate token-based pricing. So we've been bouncing between models and nothing feels quite right: \- Flat subscription? Great until your heaviest user is burning 40x what everyone else uses and your margins disappear. \- Usage-based with caps? People still get that pit-in-their-stomach feeling when they see a usage meter climbing. The part that makes this uniquely painful for AI companies: a simple query might cost us fractions of a cent, but a complex agentic workflow can run $0.50+. Would love to hear from anyone who's been through this: \- What model did you land on and how many times did you change it before it stuck? \- How do you talk about cost to customers? \- Enterprise folks - how do you sign annual contracts when your own costs aren't predictable? \- Has anyone actually made outcome-based pricing work? We keep talking about it but can never define "success" cleanly enough to bill against it.