Per-seat pricing assumed one human equals one login. Agents do the work of many without occupying a chair, and the market has already started repricing the companies built on that assumption.
For two decades, the per-seat subscription was the most reliable revenue machine ever invented in software. You sold access to a person, you billed monthly, and growth was a matter of arithmetic: more employees at your customer meant more seats, and more seats meant more money. The model was so clean that finance teams could forecast it in their sleep.
That arithmetic just broke. An AI agent does not log in, does not need onboarding, and does not occupy a seat. It executes thousands of tasks on behalf of a workforce that may have shrunk. When one agent can compress a customer's seat count by something close to 90 percent, the meter you have been reading no longer measures the value you deliver. It measures a fiction.
The market noticed before most boardrooms did. In February 2026, a roughly 48-hour selloff wiped out about $285 billion in SaaS valuations, the largest AI-triggered software repricing on record. Traders called it the SaaSpocalypse. What they were really pricing was a single question: if your revenue is tied to headcount, what happens when your customers stop counting heads?
The seat was always a proxy, and the proxy is failing
Per-seat pricing was never about seats. It was a convenient stand-in for value. More users roughly meant more usage, more dependence, more willingness to pay. The proxy held because work was done by people, and people needed accounts.
Agents sever that link. The work still happens, sometimes more of it, but the headcount that justified your invoice evaporates. A support organization that ran on forty licensed agents might now run on four humans supervising software that resolves the rest. Your customer is getting more value and paying you less. No vendor survives that gap for long.
The first reflex, raising per-seat prices to recover lost volume, fails on contact. It asks buyers to pay more for fewer seats at the exact moment they can watch the seats disappear. The structure has to change, not the number on it.
Where the money is actually moving
The numbers tell a consistent story. Pure per-seat pricing slipped from 21 percent of SaaS companies in 2025 to 15 percent in 2026. Hybrid models, a base subscription plus usage overage, have become the most common structure at roughly 41 percent adoption, and the firms running them report about 38 percent higher revenue growth and net revenue retention than peers. Hybrid is winning because it hedges: predictable for the CFO, elastic enough to capture upside when an agent does more work.
Outcome-based pricing is the louder story. In April 2026, HubSpot dropped its Customer Agent to $0.50 per resolved conversation, charging only when the software does the job. Roughly 47 percent of SaaS companies are now exploring or piloting outcome-based models. Gartner projects that by 2030, at least 40 percent of enterprise SaaS spend will shift to usage, agent, or outcome pricing. That is not a tweak to the model. That is a different business.
The seat measured who could use your software. The new question is what your software actually did, and almost no contract written before 2025 knows how to answer it.
Figure 1 — The pricing models, compared
Model | What you charge for | Strength | Weakness |
|---|---|---|---|
Per-seat | Named users with login access | Predictable, simple to forecast and sell | Collapses when agents replace logged-in users |
Usage | Volume consumed (calls, tasks, compute) | Scales with value delivered, agent-friendly | Revenue gets lumpy and hard to forecast |
Outcome | Defined results (resolved tickets, closed cases) | Aligns price to value buyers actually feel | Disputes over what counts as a "result" |
Hybrid | Base subscription plus usage overage | Predictable floor, elastic upside; highest growth | More complex to price, bill, and explain |
How to read it: Move down the rows and you trade forecasting comfort for value alignment. Per-seat is the easiest to run and the most exposed to agents; outcome is the best aligned and the hardest to govern. Hybrid sits in the middle, which is why adoption and growth are clustering there.
The contrarian catch nobody puts in the deck
Outcome-based pricing is sold as the buyer-friendly answer: you only pay for results, so risk shifts to the vendor. That framing is half the truth. Pay-per-result only works if both sides agree what a result is, and most contracts do not define it with enough precision to survive a dispute.
Consider HubSpot's resolved conversation. Is a conversation resolved when the agent closes it, or when the customer stays gone for thirty days, or when no human ever has to touch it again? Each definition produces a different invoice, and each gives one party an incentive to argue. Outcome pricing does not remove risk. It relocates it into measurement, and measurement disputes are slow, expensive, and corrosive to the relationship. The vendors that win here will be the ones who write the definition of "done" before the lawyers do.
What this means for leaders
Audit your revenue for seat dependency now. Pull the share of recurring revenue tied directly to named-user counts and model what happens when those counts fall by half. If a 50 percent seat reduction guts your forecast, you are not running a pricing question. You are running a survival question, and the time to find out is before renewal season, not during it.
Move to hybrid before you reach for outcome. The data favors hybrid for a reason: it captures agent-driven upside without betting the company on contested metrics. A base plus usage overage gives your CFO a floor and gives you a meter that grows when agents do more work. Treat outcome pricing as a deliberate later step, taken only where you can define and instrument the result cleanly.
Make measurement a contract discipline, not an afterthought. If you sell outcomes, the single most valuable asset you own is an airtight, mutually accepted definition of the result and the system that proves it. Invest in instrumentation and dispute language with the same seriousness you once gave to seat counting. The vendor who can prove what their software did will win the argument the vendor who merely claims it will lose.
The per-seat era did not end because the model was bad. It ended because the thing it measured, a human at a keyboard, stopped being where the value lived. Whoever can price the work itself, cleanly enough that nobody fights about the bill, inherits the next decade of software revenue. The $285 billion that vanished in February was not a verdict on AI. It was a verdict on every company still selling chairs in a room that is emptying out.
A BusinessInfomatics original. Figures drawn from 2026 SaaS-pricing analyses by MindStudio, Monetizely, and Improvado, and contemporaneous "SaaSpocalypse" market reporting.
