the end of per-seat

When I set the price for fikst, I picked one flat monthly fee per organization. Not per user. Why? Because I want a customer to see their bill coming. And, funny thing, that’s the model half the market’s reaching for right now anyway. AI agents are pulling per-seat apart, and vendors are swapping it out fast for prices per use, or per result. And there’s something uncomfortable tucked in there for you, the buyer. Fewer users on the invoice doesn’t always mean a lower bill. Not really. Because the meter just slides off the number of people and onto whatever the software’s up to. The one everybody quotes? SaaStr. The SaaS community that cut 80 percent of its Salesforce seats and then watched the bill climb 83 percent anyway. So. Renewing a software contract in 2026? Then mostly you read the usage terms. The credits, the caps, and what “resolved” even means.

80% fewer seats, an 83% higher bill

Jason Lemkin, the guy who founded SaaStr, went and threw his own Salesforce invoice online back in April 2026. Have a look. Last year, twelve grand for a good ten human users. This year? Around twenty-two. And that’s for two human seats and one API seat. So more money, fewer people. Up 83 percent, 80 percent fewer users (SaaStr, 2026). And what’s sitting in between? The twenty-plus AI agents the company wired into Salesforce. They lean on the platform, by Lemkin’s own count, roughly a hundred times harder than the people ever did. And that usage gets billed as consumption.

Now, one caveat, right up front. This is one company. Three people, twenty-plus agents. No market average, mind you. But the mechanism underneath? That one holds wider. Lemkin puts it well in a follow-up post: seats crash to consumption, not to zero (SaaStr, 2026). You automate the work away and the bill just sits there. For as long as the vendor slides the meter over a notch, anyway.

what is ai changing about software pricing?

Everywhere you look, same move. Vendors going from a price per user to a price per use, or per result. And that new meter mostly just lands on top. On top of the subscription you already had anyway. Bain & Company looked at a good thirty established vendors and saw about 65 percent of them running a hybrid model: an AI usage meter parked right next to the existing seat price (Bain & Company, 2025). And a survey of 240 software companies points the exact same way. Per-seat as the main model slid from 21 to 15 percent in a single year, and hybrids climbed from 27 to 41 (Growth Unhinged, 2025).

And it’s moving fast, right? Across the five hundred biggest SaaS and AI companies, one index clocked more than 1,800 price changes in 2025. That’s 3.6 per company. Let that land (Growth Unhinged, 2026).

Let’s get concrete, because here’s how it shows up on the price list. Intercom charges $0.99 per resolved conversation for its Fin agent (Macha, 2026). HubSpot flipped over in April 2026 to $0.50 per resolved conversation for its Breeze agent, with the story that you only pay once the task’s done (HubSpot, 2026). And Salesforce? Takes the cake. Hangs three pricing models on Agentforce at once: per conversation, per action, and a flat license per user for unlimited internal use (The Pricing Conundrum, 2026). The market leader running three meters side by side. That tells you pretty much everything. Nobody knows yet what AI work is worth. The seller doesn’t either. And meanwhile Agentforce keeps thundering on, up 169 percent, toward some $800 million a year (The Pricing Conundrum, 2026).

is outcome-based pricing fairer for the customer?

Outcome-based pricing, paying per result then, sounds fairer than per seat. Only it rarely takes the guesswork off the buyer. Not really. And it drags in problems of its own.

For starters, it stacks. At most vendors that result meter just lands on top of the seat price, exactly like that Bain research shows. So your old bill stays put and a variable line gets bolted on next to it. Lovely.

Then the question of who actually decides what “resolved” is. HubSpot says in its own announcement that the Breeze agent resolves an average of 65 percent of conversations across more than 8,000 customers. But sure, that’s the vendor’s own research. And it’s the vendor keeping the tally too (HubSpot, 2026). For as long as they define it, measure it and bill it, you walk into every disagreement already a step behind.

And then the default setting. Meanest of the bunch, if you ask me. As of May 2026 Zendesk folded AI agents into every Suite subscription, with a handful of free resolutions per employee a month. Sounds decent. Only everything above that goes per unit, and that overage’s been switched on with no ceiling, by default, since January 2026 (Richpanel, 2026). So “included AI”, in practice, is just a new meter with no roof on it. And people feel it. In a survey by SaaS management platform Zylo of 218 IT leaders, 78 percent hit unexpected costs from AI features or usage pricing over the past twelve months (Zylo via Advisable, 2026).

what does this mean for dutch small business?

And for Dutch small business, this lands at a rotten moment. Because right here is where everyone’s piling in right now. Research by the software company Wolters Kluwer, across a good thousand European SMBs, found 84 percent of Dutch companies want to invest more in AI over the next three years. The highest of any country they surveyed (Wolters Kluwer, 2026). And that money? Mostly lands with exactly the vendors rebuilding their pricing model right now.

Add to that: a small business like that misses the negotiating power a player like SaaStr does have. Buy through a self-serve subscription or a standard contract, and you just get the list price and the default settings. Uncapped overage included. And a procurement department of your own, picking through the fine print on credits and “resolution windows”? Usually not there either. And here’s what really jumped out at me: across two rounds of research I couldn’t find a single Dutch analysis reading this shift for SMBs. The only Dutch-language piece that came up was a reactive explainer of one HubSpot price change from 2024 (Bright Digital, 2024). And that quiet, that’s not safety, mind you. It’s playing out here just as hard. Only the warning doesn’t come with it.

It’s the same gap I wrote about before, the one between pilot and production. The pilot phase is small and easy to keep an eye on, and it’s only at production scale that you see what a thing really costs (see the production gap).

And that it can go differently, I know because I do it myself. For fikst I charge one fixed, transparent monthly fee. Not per user, because a small business has to be able to build on what it can predict. Honestly, I want away from the per-user thing altogether. Too much fuss. And yeah, sure, if you’re a giant running 500 seats, you were never my target group. But the pilot price I’ve kept deliberately low, because I want a real partnership with these companies, and a low entry cost buys mutual trust. And it makes sense for me too: I want to focus on three live customers, no more. It’s a collaboration. I expect a lot back from them, I really want them working the system, because that’s when you run into things, not that the process stalls, but things I can then fix really fast. That’s worth a lot, because you can’t test everything yourself, and you get real business input. I’m on my own for now, so my own costs are low, and I pass that straight into the price. It’s not much more than a preference, mind you. Mostly it just shows that a vendor can make that choice. If they want to.

how to prepare your contract renewal in 2026

Look, you really don’t have to turn into a pricing expert. And don’t go bolting to another vendor in a panic either, because per-seat’s alive and kicking for now, just as part of these hybrid models. What you do need? Half an hour, your contract in front of you, and a few things.

First, write down which meters are already in there right now. Seats, credits, conversations, resolutions, API calls. Usually there are more of them than you’d think, honestly. Then ask which pricing models run side by side, and have every one of them worked out on your own volume. At Salesforce there are three at once, remember. So pick one on purpose. And get that billing unit explained down to the bone. What counts as a “resolved conversation”, or an “action”? Who measures it? Can you check that measurement yourself? Keep pushing.

Then the overage. Find out what happens by default, and demand a cap. Or a hard stop. I learned this one the hard way: I let an agent test autonomously against Google Maps routing with no spend cap, and walked into a thousand-euro bill. Always set your API limits. It’s a serious trap. And at Zendesk that uncapped overage has just been on by default since January 2026. Ask too how agent traffic gets billed the second you wire AI into the system yourself. As a seat? As API usage? As consumption? Because this is the exact spot where that SaaStr bill went off the rails.

One more, and I hold myself to this one from the vendor side: pin down the maintenance. How it gets handled, in what format, and how fast they answer. A mutual agreement, on paper. I’m solo, so a short line to my customers is everything. I’ve got all the knowledge in one head, so I can switch gears incredibly fast. I’m basically available, I put in the hours. That’s exactly what lets me run cost-efficiently and still give the service that’s needed. So ask your vendor for the same clarity: who fixes what, and by when.

And demand a usage report every month, with an automatic little nudge at, say, 80 percent of your budget. Then, last thing, run a growth scenario. What does this contract cost if your AI usage goes up fivefold? Can’t the vendor lay that out on a single page? Then you’ve pretty much got your answer already.

And where I stand? Straight, no detours. Outcome-based pricing, in its current form, is above all a revenue model for the vendor. And for a small business, being able to predict the bill is worth more than some theoretical fairness per result. So go with vendors who can explain their meter in one sentence. Cap every variable cost line. And treat every price change in 2026 as your opening to renegotiate. Because you’ve got something they badly need. A customer who stays.