Seat Pricing Is Fading. How to Price Your Product Now.
The pricing model that carried SaaS for a decade is breaking as AI agents replace seats. Investors and operators are debating what comes next. Here is how a first-time founder picks a pricing model that survives 2026.
Why is everyone suddenly arguing about software pricing?
Because the default just died. For fifteen years, pricing a B2B product was barely a decision: count the users, charge per seat, raise the price at renewal. Simple to sell, simple to budget, and it grew automatically as your customer hired.
Then AI agents started doing work that used to require a seat. In July, EQT's operating partner team published a piece with an unusually blunt title: seat-based pricing is fading. Manny Medina, who built Outreach and now runs the AI monetization platform Paid, put it plainly: with agents in the mix, the seat model just breaks. If your customer runs fewer people and more agents every quarter, a per-person price shrinks while the value you deliver grows.
Davis Geidt of the Alexander Group says pricing is now consistently bubbling up to the board. Translation: the people who fund and govern startups no longer treat pricing as a detail. If you're launching in 2026, your pricing model is a strategy decision on par with what you build.
What actually broke the seat model?
One assumption, quietly load-bearing: that software gets used by human employees, and that customers keep adding them.
Seat pricing was never about value. It was a proxy. More employees meant more usage, which meant more value, so charging per head roughly tracked worth. The proxy held as long as headcount grew.
That link is snapping. Development tools are collapsing into command lines and agents. Medina points out that teams now run fewer developers and product managers while output stays flat or rises. Support teams resolve more tickets with fewer agents on payroll. Marketing teams of three do what ten did.
There's a second, older problem too. Buyers spent a decade paying for sprawling seat bundles nobody opened. Finance teams got wise, and software audits became a sport. So the seat model faces pressure from both directions: customers resent paying for idle humans, and vendors lose revenue when the humans disappear. A pricing model that annoys both sides doesn't survive long.
What are the three models on the table right now?
The industry has run through three broad experiments, and no consensus has formed yet.
Seats, the incumbent. Predictable for everyone, increasingly disconnected from value, and shrinking wherever agents replace people.
Consumption, the first response. Customers pay per unit of usage, often per token of AI processing, the way AWS bills for compute or OpenAI bills its API. It tracks cost well. But it produced a new complaint: unpredictable bills with no clear link between spend and results. A CFO can't budget for a number that swings 40% month to month.
Outcomes, the current frontier. Customers pay for measurable results: a resolved support ticket, a qualified lead, a completed filing. Intercom's Fin is the loud example, charging per resolution rather than per agent seat. The appeal is obvious, because the incentive alignment is perfect on paper. The execution, as we'll see, is harder than the pitch.
Most real companies in 2026 are running hybrids of two or all three. That's not indecision. It's the honest state of the market.
Why is outcome pricing so appealing, and so hard?
Paying only for results sounds like the end of pricing debates. The operators actually running it will tell you otherwise.
Start with definition. What counts as an outcome? A support conversation can end without a clean resolution even when the AI did most of the work. Des Traynor, co-founder of Fin's team at Intercom, notes the AI can do a lot in an interaction and still get no resolution. Who pays for that effort?
Then measurement. Latané Conant at Parloa, an agentic customer service startup, flags a sharper problem: customers can get real value but structure their interactions so the formal outcome criteria never trigger, so they never pay. Gaming the meter becomes a procurement skill.
And risk cuts both ways for the vendor. An unconstrained AI conversation can burn tokens for an hour, running up your costs, and still miss the outcome that gets you paid.
Outcome pricing works best where the process is defined and success is countable: support tickets, document processing, lead qualification. If your product's value is fuzzy, forcing outcome pricing onto it will hurt.
Your AI co-founder is ready when you are.
Foundra turns everything in this article into an actual plan. Validation, customers, pricing, launch. In one place, in your voice, in an afternoon.
Start free→3-day free trial. No credit card. Cancel anytime.
What do buyers actually want from your pricing?
Predictability. That's the through-line in every operator interview, and it explains why the seat model lasted this long despite its flaws.
Katie Burke, chief operating officer at legal AI startup Harvey, frames it well: outcome pricing is likely the future, but you have to match what people are ready for. Customers don't mind paying for value. They do mind a bill their finance team can't explain or forecast.
This is why the emerging playbook wraps new models in old comfort. Minimum commitments. Usage thresholds with caps. Quarterly true-ups instead of monthly surprises. The customer gets a number they can put in a budget; you get revenue that scales with the value you create above that floor.
Medina's take is worth pinning above your desk: seats can still be a bridge, but in most cases they won't be the engine of growth anymore. Use familiar structures to get customers comfortable, and build toward the model that actually reflects what you deliver.
How should a first-time founder pick a starting model?
Work through three questions in order, and write your answers down before talking to a single customer.
What's your value metric? Name the unit that grows when your customer wins: tickets resolved, invoices processed, candidates screened, reports generated. If usage of that unit rises and your customer gets happier, it can carry your price. If you can't name one, you're not ready to price, you're ready to do more discovery.
Who signs, and what budget does this come from? A product bought by a team lead on a card needs a simple monthly number. A product bought by a VP replacing labor cost can support outcome pricing, because you're competing with a salary line, not a software line.
What can you measure without arguments? Only price on numbers both sides see the same way, in a dashboard, from day one. Every ambiguous metric becomes a billing dispute later.
Then start hybrid: a modest platform fee for predictability, plus a usage or outcome component tied to your value metric. Floors protect you. Caps protect them.
How do you test a price without wrecking your pipeline?
Treat pricing like a product experiment, not a tattoo. You'll get it wrong the first time; the goal is to make being wrong cheap.
Run the willingness-to-pay conversation before publishing anything. Ask five prospects: what would make this a no-brainer at twice the price? Their answers tell you which value metric they already believe in.
Test on new deals only, and grandfather existing customers loudly. Nothing burns early goodwill like surprise repricing on the people who bet on you first.
Model the revenue math before you commit. Take your three most recent customers and project each candidate model over twelve months: floor revenue, expected usage, worst case if usage stalls. This is exactly the kind of scenario work that's worth doing in a structured template rather than a napkin. A spreadsheet works; planning tools like Foundra include financial projection templates built for founders running this comparison for the first time.
Give every experiment a review date. Ninety days, then you look at win rate, expansion, and billing disputes, and you keep, tweak, or kill.
What are the warning signs your pricing is wrong?
Your customers and your metrics will tell you, if you're watching the right places.
Discounting is the loudest signal. If every deal closes below list, your list price is fiction and your value story isn't landing. Conversely, if nobody ever pushes back, you're underpriced; a healthy price gets questioned in about a third of negotiations.
Watch the usage-to-bill ratio per account. Customers paying a lot and using little are churn scheduled for later. Customers using heavily and paying little are your expansion pipeline, and evidence your value metric is miscalibrated.
Listen for budget language in sales calls. If buyers keep asking "which line does this come from," your pricing doesn't map to how they think about money. That's fixable with packaging, not discounts.
And track time-to-invoice-dispute. Outcome and usage models fail quietly at the billing stage first. One disputed invoice is noise. The same dispute from three accounts means your meter, not your customer, is the problem.
Frequently asked questions
Is seat-based pricing completely dead? No. It still fits products where a human really does sit in the tool all day, like design software or an IDE. It's fading as a default, especially for AI products where agents do the work. If seats stopped tracking your value, stop using them as your meter.
Should I copy Fin's per-resolution pricing? Only if your outcome is as countable as a closed support ticket. Fin works because resolution is defined, logged, and hard to fake at scale. If your outcome needs a meeting to agree on, you're not ready for pure outcome pricing.
What's a reasonable starting point for a brand-new AI product? A hybrid: small monthly platform fee plus a metered component tied to your value metric, with a cap for the customer's comfort. It's boring, and it keeps every future model open while you learn.
How often can I change pricing without annoying customers? On new deals, as often as you learn something. For existing customers, once a year at most, with notice and grandfathering. Churn from a botched migration costs more than the uplift.
Do investors care which model I choose? They care that you chose deliberately. Expect diligence questions about net revenue retention under your model, gross margin at scale, and what happens to revenue if customer headcount halves. Have the scenario math ready.
You just read the theory. Ready to build the thing?
Foundra is your AI co-founder. It turns an idea into a validated business plan, a go-to-market, and your first 10 customers. In an afternoon, not a semester.
3 day free trial. No credit card. Works in 20 languages.