Why Per-Seat Pricing Is Dying (and What Replaces It)

Picture the best thing your team could ship this year: an AI agent that quietly does the work of eight of your ten support reps. It clears tickets at 2 a.m., stops escalating the easy stuff, and your customer simply never backfills the roles people leave. You built something genuinely great.

Now look at the invoice. If you charge per seat, the eight humans who used to log in are gone — and so is most of your revenue. You built a product so good it cancels your own contract. That's the trap at the center of why per-seat pricing is dying, and why how you charge has quietly become one of the most consequential product decisions you'll make all year.

Per-seat pricing is dying, and the data isn't subtle

This isn't a vibe. Per-seat pricing adoption among SaaS companies fell from 21% to 15% in a single twelve-month stretch, according to The SaaS CFO's Ben Murray — a steep slide for what used to be the default way to sell software. Bessemer's investors put it flatly in their AI pricing playbook: "AI-native companies are abandoning seat-based SaaS pricing in favor of usage-, output-, and outcome-based models."

And nobody has settled on the replacement. Kyle Poyar, who tracks this obsessively, counted more than 1,800 pricing changes across the top 500 SaaS and AI companies in 2025 alone — 3.6 changes per company, and called it the year "seemingly everybody lost confidence in their pricing." When an entire industry reprices itself three times in twelve months, that's not optimization. That's panic with a spreadsheet.

The real problem: your value metric stopped matching your value

Per-seat pricing was never really about seats. It was a proxy. For thirty years, "number of people who log in" was a decent stand-in for "how much value this software creates," because software needed humans to operate it. More users, more work done, more value. Fair enough.

AI breaks the proxy. The entire promise of an agent is that it does the work without a human in the seat. So the meter you've been reading — logins — now points in exactly the wrong direction. As Poyar put it on Metronome's blog, the shift is away from "subscriptions times seats" toward pricing that actually tracks what the customer consumes. Seats measure the one thing your product is trying to eliminate.

It gets worse on the cost side. AI features run at gross margins of 50–60%, versus 80–90% for classic SaaS, because every inference call costs real money. So you have a pricing model that falls as your product succeeds, sitting on top of a cost structure that rises every time someone uses it. That's not a pricing tweak waiting to happen. That's a leak below the waterline.

Pricing is a product decision, not a finance chore

Here's the part most teams get wrong, and the reason I keep dragging this conversation out of the finance meeting and into the product review: your pricing metric is part of your product. It tells the customer what you think is valuable, and it tells your own team what to optimize.

Charge per seat and you're quietly telling everyone that success means "more logins." That's the same logic that produces shelfware — software people buy and never open. I've written before about how most of what teams ship goes unused, and per-seat pricing actively hides that rot, because you get paid whether or not anyone gets value. A usage- or outcome-aligned price can't hide it. If nobody's using the thing, the revenue evaporates and you find out fast. Painful, yes — but that's a feature, not a bug. It points your roadmap at value instead of at renewals.

The menu, and the honest tradeoffs

There's no clean winner yet, so don't let any blog post — this one included — sell you a silver bullet. The real options:

Usage / credits. Charge for tokens, actions, or credits consumed. Honest, and it scales with your costs — but it can punish exploration. If customers feel a meter running every time they try something new, they use it less, which is the opposite of what you want early in adoption.
Hybrid: base plus overage. A subscription floor with usage layered on top. This is fast becoming the default — by one Bain analysis, 65% of vendors have added AI usage charges on top of an existing base structure. It hedges nicely: predictable revenue for you, room for the price to grow as value does.
Outcome-based. Charge per resolved ticket, per qualified meeting, per dollar collected. Incentives align perfectly — you win only when the customer wins — and it's the model everyone romanticizes. It's also brutally hard to attribute and still rare. Bloomberg's forecast has outcome-based pricing climbing from 10% to 60% of SaaS over the next decade, which tells you it's the destination, not the current address.

Pick based on where your value actually shows up, not on what's trending on LinkedIn this quarter.

Choose your value metric like you'd choose a feature

The best pricing move I've seen isn't a model — it's a question: what goes up for the customer when we do our job well? Tickets resolved. Hours saved. Pipeline booked. That's your value metric. Charge for that, and your incentives and theirs finally point the same direction.

Look at one case The SaaS CFO flagged: after a team deployed AI agents, their Salesforce bill climbed 83% — from $12K to $22K a year — even as they cut human seats from more than ten down to two, because the agents hammered the platform a hundred times harder than people ever did. Fewer seats, far bigger bill. The vendor got paid for the work done, not the chairs filled. That's the whole game in one invoice.

So who owns this?

If pricing is a product decision, someone in product has to own it — and increasingly that's the PM. Pricing, packaging, and the activation funnel are exactly the go-to-market turf I argued product managers are being pushed to own as the role splits in two. The PM who can name the value metric and defend it across the table from finance is worth a great deal more than one who can only groom a backlog.

The takeaway is uncomfortable and simple: if your revenue goes down when your product gets better, you haven't been disrupted — you've mispriced. Per-seat pricing is dying because it counts the wrong thing in a world where the software, not the user, does the work. Find the number that rises when your customer wins, and charge for that. Everything else is rearranging chairs nobody's sitting in anymore.

Per-Seat Pricing Is Dying — and It's a Product Problem