Why most companies are pricing AI wrong
Maarten Laruelle Back in February, I argued that outcome-based pricing fails when customers invest heavily in achieving those outcomes, providing people, data, processes, and expertise.
The article resonated. But it also sparked a question I have been getting weekly: “What about AI? Surely that is different?”
Let me save you six months of pricing experiments: it is not different. It is the same framework, just with newer technology. A lot of founders are scrambling to price their AI features. They are making the same mistake I wrote about with outcome-based pricing 7 months ago.
The problem: treating all AI the same or not understanding the value chain.
After recently working with several scale-ups adding AI to their products, here is what I have learned.
Not all AI is created equal
Three distinct categories for pricing.
Autonomous AI delivers outcomes independently. The AI runs, delivers results, done. No hand-holding required. Examples: AI that automatically identifies and fixes security vulnerabilities, or an agent that handles support tickets without human intervention. Price this on true value. The customer gets clear outcomes without effort.
Collaborative AI needs input but multiplies value. It requires data, prompts, and validation, but the value multiplication is clear. Examples: AI that generates marketing copy based on your brand guidelines, resume templating that smartly transforms resumes from one format to another, or niche apps enriched with prompt history, licensed content, and expert insights. Price based on effectiveness. How much does the AI accelerate the outcome?
Efficiency AI is just faster infrastructure. It makes existing processes quicker, like upgrading from dial-up to broadband. Examples: AI-powered search that returns results faster, or an agent that serves your knowledge base in a smart way. Traditional pricing works here. It is a better tool, not a value creator.
The reality check
Most AI features fall into category 2 or 3. Yet everyone wants to price like category 1.
If your customer provides the data, the prompts, the validation, the implementation, and the quality control, then you cannot price like you delivered the entire outcome.
What this means for your pricing
Stop asking “What is the AI worth?” Start asking “How much work does the customer still do?” Envision their way of working and the value chain.
The more your AI depends on customer input, the more your pricing should anchor to tangible elements: data processed, reports generated, time saved. Not the final business outcome.
The parallel to outcome-based pricing
This builds on my earlier article about outcome-based pricing. Same principle: customers resist paying for outcomes they helped create.
Whether it is AI or any other tool, if they are doing significant work, they want pricing that reflects that.
Practical example from a recent client
They had AI that could predict customer churn. Sounds like autonomous AI, right?
But the customer had to clean and format their data, define what “churn” meant for their business, validate predictions against their knowledge, and implement retention actions.
We moved from outcome pricing (per churn prevented) to usage pricing (per prediction). Revenue went up because adoption went up.
The test
Before you price your AI feature, ask: could a customer achieve success if they just pressed “start” and walked away?
If yes, price the outcome. If no, price the elements.
Want to figure out the right pricing approach for your AI features? Book a conversation.