Agents broke per-seat pricing. Outcome pricing is a trap for subjective products.

When your AI does the work, you can't charge per seat anymore — but the fashionable fix, charging per outcome, only works if your outcome is something you can count. Here's how I think about pricing when the thing you sell is subjective.

There's a pricing question sitting under every AI product right now, and most makers are answering it by copying whoever's biggest in their category. That's how you end up mispriced.

Here's the shift, as plainly as I can put it. Per-seat pricing is quietly falling apart, and the reason is structural, not a fad. Seat pricing assumes software is a tool a human uses — more humans, more tools, more money. But an agent doesn't hand a person a faster tool. It does the work instead of the person. So when your product replaces five hours of someone's week, charging by the seat punishes your customer exactly for getting the efficiency they paid for, and punishes you for delivering it. The better your agent gets, the fewer seats they need. You've built a business that gets worse as your product gets better.

The industry's answer this year is outcome-based pricing: don't charge for access, charge for the result. The cleanest live example is Intercom's Fin, the customer-support agent that charges $0.99 per resolution — you pay when it actually closes a ticket, not for a login. It's a genuinely elegant model, and if your product produces a discrete, countable, agreed-upon result, you should look hard at it. "You pay when it works" is the easiest value story a founder ever gets to tell.

But here's the part I want to flag for the makers building the weirder, softer products, because I'm one of you. Outcome pricing has a hidden requirement: the outcome has to be legible. Both sides have to agree, cheaply and instantly, that the thing happened. A resolved ticket is legible. A closed deal is legible. A qualified lead is legible. That's why support and sales tools went there first — the outcome practically timestamps itself.

Now try it on a product whose value is subjective. I build Murror, an AI companion for understanding your own emotions. What's the outcome I'd charge for? A journal entry written? People write when they're spiraling and skip when they're fine, so charging per entry would mean my incentives are best served by my users being at their worst. An "insight delivered"? There's no click that proves a reflection landed. Sometimes the most valuable session is the one where someone writes for ten minutes, closes the app, and feels quietly clearer — and there is no event I could bill for that. The moment I attach a price to a countable proxy, I start optimizing the proxy, and the proxy is never the thing.

This is the trap I want to name: outcome pricing doesn't just charge for value, it defines value as whatever you can measure, and for a lot of the most human products, the measurable part and the valuable part are different things. Charge per outcome and you'll slowly bend the product toward manufacturing the billable event. Users feel that bend before they can name it. It's the pricing equivalent of a company optimizing engagement and accidentally rewarding anxiety.

So where does that leave a solo maker who can't just cargo-cult Intercom? A few things that hold up for me.

Price the access, not the outcome, when the outcome is a feeling. A calm, predictable subscription is honest about what you actually provide: a space, a tool, a companion that's there when they need it — not a slot machine that pays out billable moments.

Never let your pricing metric become a metric you'd be ashamed to optimize. Run the test before you ship a plan: if this number went up because users were suffering more, would my revenue go up too? If yes, throw the model out. That single question would have saved me from two ideas I thought were clever.

Make your pricing match your product's honesty, not the category's default. Support tools bill per resolution because that's genuinely what they do. If your product's value is quieter and less countable, forcing it into a per-outcome frame doesn't make you look modern. It makes you look like you don't understand your own product.

Pricing feels like a spreadsheet decision and it's actually a values decision in disguise. You're telling the model — and your users — what you think "value" means. The makers who get this right in the agent era won't be the ones who copied the pricing page that was winning last quarter. They'll be the ones who priced the thing they actually do, especially when the thing they do can't be reduced to a number.