Skip to content
47% off: the complete course for $79 (reg. $149)Order now →
AI Builder
Module 12Full course11 min read

Pricing AI Products: Seats, Usage, and Credits

Pricing models for products with real per-use costs — seats, usage-based, credits, outcome pricing, and hybrids — and how to choose one with margins in mind.

Traditional software pricing rests on a convenient fact: once the product is built, serving one more user costs almost nothing. AI features break that. Every summary, draft, and answer has a real cost, and your heaviest customer can easily use many times what your median customer does. Price an AI feature like regular SaaS and you can end up with your best customers being your least profitable ones, sometimes losing money on every month they stay.

This lesson gives you a method for pricing a product with real per-use costs. You'll calculate cost per unit of value, see how seats, usage, credits, outcome pricing, and hybrids each handle cost variance, design a credit system that survives model changes, and check your plan's margins against your heaviest customers before they find the gap for you.

Why AI breaks the usual pricing playbook

Three properties of AI features make pricing harder than it looks.

Real marginal cost. Each use of the feature spends money on model tokens. The amount per call is small, often around a cent, but it scales directly with usage. The key business number here is gross margin: revenue minus the direct cost of serving customers, as a share of revenue. Software businesses are valued partly on high gross margins, and AI costs come straight out of that number.

High variance between customers. Usage of AI features is rarely uniform. One Deskmate customer reads summaries on every ticket; another never opens the panel. One runs a bulk job to tag three years of archived tickets. Average cost per customer is a misleading number when the distribution has a long tail, and Unit Economics of AI Features showed how a feature's success can push more customers into that tail.

Moving costs. Model prices for a given capability level have tended to fall over time, while what customers ask the product to do tends to grow: longer documents, more steps, better models. Your cost per unit will change in both directions, and your pricing should survive that without a public repricing every quarter.

A good pricing model answers one question honestly: who carries the cost variance, you or the customer? Every model below is a different answer.