Pricing AI based on outcomes is a non-starter

In partnership with

Pricing By Elimination

Every enterprise wants AI but neither the seller nor the enterprise knows how to value it. So sellers are pricing AI by elimination. The fundamentally different value of AI applications is that it does not need humans to sit in front of the screen and push work to the next human. Pricing per seat (per human) does not make sense. We are eliminating that option. So what then should get priced? Work. Work?

Work gets done when a sequence of tasks get done. Some get done in the background by AI. Some get done by humans after AI has done the first pass. We can price the “runs” that AI makes - how many documents read, how many errors found, how many meetings scheduled based on some conditions, etc. We can even differentiate between background runs (autopilot) and assistive runs (copilot). But the first rule of pricing is to not complicate. People need certainty. So the callout was to price not the work but the outcome.

Pricing an outcome seems like the right way to align price to value. It makes the sales guy look fair and the buyer feel taken care of. This is what the vendors of AI applications believe in. But outcome based pricing has not taken off.

The Economic Value of An Outcome

What is the economic value of an outcome? It depends on whom you ask. As a seller of AI applications, I’d think that I really moved the needle for you by processing all your work and auto-generating those invoices or drafting your contracts or making sure your goods reached the shores with all paperwork done along the way.

As a buyer though, I don’t care as much about the economic value of an outcome as I do about the cost of getting that outcome and the alternatives I have for getting that outcome.

Imagine this:

If there is a lifeguard at the swimming pool, it is nearly impossible for any one to die due to drowning. What is the cost of one live saved, as an outcome? Immeasurable. Have you seen a millionaire lifeguard? No. That’s because its easy to become one and there are many vying for that job.

While the value of life saved is immeasurable, the cost to deliver the value is kept very low due to alternatives competing for that job.

AI is deflationary

The first definitive proof of AI being deflationary is in how AI-based coding assistants have made it easy to unlock a software developer’s productivity. The cost per line of code is on a precipitous fall, even as value to the business increases. More AI software will fill the gaps where traditional software could not serve. The world will have more software. The world pay less for each of those software applications.

The outcome based pricing model did not take off because, sellers looked at the value they created while ignoring alternatives. The alternatives have a deflationary impact on pricing. Value-based pricing moves price closer to the economic value. Labor replacement or software replacement takes the price closer to cost.

Outcome based pricing was a defense mechanism

I started the essay saying that AI is getting priced by eliminating approaches like seat-based pricing. But there is another angle to elimination. AI eliminates parts of work or the necessity for a worker to do that work or part of that work. AI applications are pricing ingredients (tokens it takes to do a job). Each discrete part of work is being broken down to tasks and each task has a token-consumption-linked pricing. There are more complex variations based on which AI model is being used, how fast work gets done, etc. But the trend points to consumption pricing and not outcome pricing. The price per consumption unit is not neatly evaluated against the value that is created. In practice, the price is valued against value created, software and/or labor replaced.

But herein lies the problem for software vendors. Cost as a reference as against value as a reference leads to cost-plus pricing. All AI companies are wrappers around intelligence. But that intelligence is now getting far ahead to obliterate various traditional value levels. Screens are replaced by chat. Standard protocols like MCPs make it hard for traditional software companies to defend their turf with gated, private APIs. Its getting easier by the day assemble software. Traditional software interfaces and APIs won’t go away but for the first time there are alternatives and they are not bad for several use cases.

Everyone can build products. But they generate reliable outcomes? This is the defense for outcome based pricing. The truth is that the reliability of AI-driven workflow outcomes is yet not a given and the alternatives are all equally good or bad. There are no clear winners and the daring customers build on their own. The alternatives are priced low and compete against each other to deflate the price.

Outcome-based pricing would work well if you are the only one doing it and your competitors didn’t opt for cost-plus approach, but that is exactly what is happening. AI is, indeed, deflationary.

The #1 AI Newsletter for Business Leaders

Join 400,000+ executives and professionals who trust The AI Report for daily, practical AI updates.

Built for business—not engineers—this newsletter delivers expert prompts, real-world use cases, and decision-ready insights.

No hype. No jargon. Just results.

Subscribe free—trusted by leaders