It's a Tuesday morning in March 2026, and a CFO at a 32-person B2B company is writing an email to her CEO and head of revenue.
The subject line is "HubSpot bill — we need to talk."
The body is short. The company hired three account executives in Q1, two of them in the last six weeks. The HubSpot seat count went from 19 to 24. The annual contract, originally negotiated at a comfortable $2,900/mo, just re-quoted at $4,800/mo after seat true-up and a quiet 12% price-per-seat increase that landed in the renewal email at the same time. The CFO is not surprised. She is not even angry. She is tired.
At the bottom of the email, she writes one sentence that she has been thinking about for a year: "Every hire we made this quarter cost us $380 a month in software, before the person did any work."
This is what per-seat pricing feels like in 2026. It used to feel like growth. It started to feel, somewhere in 2023, like a tax on success. By now it feels like a structural mistake — a pricing model that's misaligned with how businesses actually run anymore.
This essay is about why per-seat pricing in SaaS is quietly ending, what's replacing it, and why most small and mid-market software buyers — and a growing share of mid-market ones — are going to look back on the per-seat era the way we look back on per-minute long-distance phone billing.
How per-seat became the default
To understand why per-seat is dying, you have to understand why it ever won.
In the early SaaS era — call it 2008 to 2015 — software was sold into a specific kind of customer: a growing team. The whole pitch of cloud software was that it scaled with you. You didn't have to buy a perpetual license for 50 seats up front. You bought five seats now, added five next year, added ten the year after. The vendor's revenue grew as the customer succeeded.
It was a beautiful loop. The product earned its place. The buyer didn't pay for unused capacity. The vendor's metrics — net dollar retention, expansion revenue, net revenue retention above 120% at the best companies — all encoded a story where customer growth and vendor growth were the same line on a graph.
Per-seat also made enterprise sales clean. A sales rep at Salesforce or HubSpot or Atlassian could point at the org chart and quote the deal. You're 80 seats today, you'll be 140 by next year, the contract reflects that. Procurement understood it. Finance understood it. The product team didn't have to invent a complicated metering scheme. The model was simple, fair-feeling, and it scaled the vendor's revenue with the customer's success.
For about a decade, this was the smartest thing in software.
What broke the loop in 2024-2026
The loop broke for a reason nobody planned for: AI made it possible to do more with the same team.
Not dramatically, not all at once, but consistently across function after function. Sales prospecting copilots replaced two SDR hires. AI-assisted customer support handled the first-tier ticket load that would have justified another two CS reps. Engineering teams shipped at the pace of teams 30% larger. Marketing teams generated 8x the content with the same headcount. By late 2025, the median fast-growing company was growing revenue 40-60% year over year while growing headcount only 8-15%.
That sounds like a triumph, and it is. But for per-seat SaaS vendors, it's a balance-sheet disaster. Their expansion revenue was built on the assumption that customers would keep adding seats. When customers stopped adding seats, the vendors had two choices: accept lower growth (which the public market would punish), or raise the price per seat on the seats that already existed.
They raised the price.
You can see it in the renewal emails. Asana raised prices twice in 18 months. Notion did. Linear did. Atlassian's whole "Cloud Premium" repricing of 2024 was, depending on the seat count, a 30-60% effective increase on existing customers. HubSpot's quiet seat repricing in Q4 2025 added an average $840/month to a typical 25-seat Marketing Hub Professional account, with most customers finding out at renewal.
The buyer's experience of this is unambiguous. The product didn't get 60% better. The team didn't grow. The bill grew because the vendor needed it to. The relationship stopped feeling like a partnership and started feeling like a meter.
That's the loop that broke. The per-seat model only works when expansion comes from customer growth, not from vendor extraction. Once it became the second thing, the model became the enemy of the customer relationship.
The four signals that per-seat is dying
You can argue about whether per-seat is "dying" or just "changing," but here are the four signals that point the same direction.
Signal one: net dollar retention is sliding at public SaaS
The headline metric of SaaS health for the last decade has been net dollar retention. In 2021, the median public SaaS company had NDR around 118%. By 2023, that median was 110%. By the end of 2025, the median had drifted to roughly 102-104%, with several formerly best-in-class names dipping below 100% for the first time in a decade.
That's not a small movement. That's the whole expansion-revenue story of SaaS unwinding. NDR slid because seat expansion slowed and because price increases generated churn that offset the expansion. The per-seat growth engine is no longer reliable, even at the companies that built the playbook.
Signal two: "annual seat true-up" became a hated phrase
Go read SaaS-buyer forums right now. The phrase that comes up over and over in complaints is some version of "annual true-up" or "seat reconciliation." It's the moment in the contract where the vendor counts how many seats you ended up using and back-bills you for the overage. Customers hate it. It feels like the meter ran while they weren't looking.
When a pricing mechanic generates this much language and this much frustration, it doesn't survive forever. Either the vendor changes the mechanic or the customer leaves. Vendors have been losing both ways for two years now.
Signal three: YC companies are increasingly NOT per-seat
If you want a leading indicator of how SaaS will be priced in five years, look at how it's being priced by the founders building today. The 2025 YC batches have an obvious skew: a large share of the B2B software companies are using flat-fee tiers, usage-based metering, or outcome-based pricing — and a small and shrinking share are defaulting to per-seat.
That's not because the founders are ideological. It's because per-seat is now considered an inferior model for new B2B software. It's been re-categorized, in the heads of the people who fund and build this stuff, from "default sensible model" to "model you choose if you have a specific enterprise sales motion that requires it."
Signal four: the AI agent question
The cleanest sign that per-seat is broken is that nobody knows what to do about AI agents.
If a sales team has five humans and three AI agents doing prospecting work, does each AI agent cost a seat? What if the agent does the work of three SDRs? What if it does the work of 0.4 SDRs? What if it's a multi-agent system that flexes from 1 to 20 instances depending on workload?
Most per-seat vendors don't have a clean answer to this. They've shipped half-answers — "AI Studio adds $X/user/mo," "Agent seats cost the same as user seats," "Contact sales for AI workload pricing." None of those resolve the underlying contradiction, which is that the seat metaphor breaks the moment the seated entity isn't a human.
A model that can't price its own future doesn't survive its own future.
What replaces per-seat
Three pricing models are absorbing the share that per-seat is losing. They have different shapes and different tradeoffs.
- Flat-fee tier pricing. The model used by Basecamp, by Mewayz, by a growing list of SMB-and-up platforms. The customer picks a tier ($19/mo, $29/mo, $99/mo) and gets the features at that tier with unlimited or generous seats. Predictable. Decoupled from hiring. Wins on simplicity.
- Usage-based pricing. API calls, GB processed, events ingested, messages sent, runs executed. Snowflake popularized it for data, Twilio for communications, Vercel for hosting, Stripe for payments. Aligns vendor revenue with actual customer value extracted, not with seat count.
- Outcome-based pricing. A percentage of revenue, of savings, or of some other measurable business result. Affirm takes a cut of merchant revenue. Salesforce now sells "Agentforce" partly as a percent-of-conversion model. This is the rarest of the three, partly because it requires deep instrumentation and trust, but it's growing fastest in absolute terms.
Each model has a use case. Usage-based is right for infrastructure and developer tools where consumption varies enormously between customers. Outcome-based is right where the vendor can credibly take risk on results. Flat-fee is right for almost everything in the SMB and mid-market range — because the buyer at that scale wants to know what software costs, not to be metered on it.
Why flat-fee wins for SMB and most mid-market
The buyer's number-one job, when they're buying software for a small or mid-sized business, is to stop thinking about it.
They want to pay a known number once a month, get the tool, and go run their company. They do not want to plan around seat counts. They do not want to renegotiate when they hire. They do not want to spend a Tuesday afternoon doing the math on whether the AI agent should be a seat.
Flat-fee pricing decouples software cost from team size. A 5-person business and a 20-person business pay the same. A team that hires three people next quarter doesn't owe the vendor anything extra. A team that fires three people next quarter doesn't see savings, but they also don't experience the bill as a constraint on hiring.
That last point matters. Per-seat pricing isn't just expensive — it's behavior-shaping. It puts a small but real tax on every hiring decision. A CFO who's already nervous about a hire now has one more reason to delay it: "the software cost goes up, too." Multiply that across thousands of mid-market companies and you get a small but real drag on hiring decisions, which is exactly the thing AI was supposed to allow companies to do more confidently.
Flat-fee pricing also makes the sales motion cleaner for the vendor. No annual seat re-quote dance. No procurement negotiation about whether contractors count as seats. No awkward conversation when the customer's growth slowed and they want to drop seats. The vendor commits to a tier, the customer commits to the tier, both sides move on.
The hardest part for the vendor — and the most honest part — is that flat-fee tiers force you to be clear about who each tier is for. You can't hide product positioning behind seat math. The $19 tier has to be obviously right for one kind of buyer, the $29 tier for another, the $99 tier for the third. If your tiers don't have that clarity, customers feel it immediately and the model breaks.
We priced our $19/mo Personal and $29/mo Business plans flat for exactly these reasons. The plans don't get more expensive because you grew. The relationship doesn't reset every Q4. Every module is included at every tier, and the tier you pick is about company size and feature depth — not about how many people are logging in this month.
When per-seat still wins
It would be dishonest to write this essay without acknowledging where per-seat still makes sense.
Enterprise sales with deep account management. When a vendor is putting a dedicated account team on a customer, doing custom integrations, and committing to an SLA, per-seat pricing aligns the relationship cleanly. The customer is buying a managed relationship, not just software, and seats are a reasonable proxy for the scope of that relationship. Salesforce, ServiceNow, Workday — these companies aren't going flat-fee anytime soon, and they shouldn't.
Products with highly variable customer sizes. If your customer base ranges from 3-person startups to 30,000-person enterprises, you need a meter, because a single flat fee can't be both fair to the startup and meaningful to the enterprise. Per-seat is one such meter — usage-based is another. The mistake is using per-seat for SMB when the customer-size variance is actually low.
Network-effect products where each seat creates more value. Slack, Notion, Linear when used as a team — each additional seat literally makes the product more valuable to the existing seats. Per-seat pricing captures part of that compounding value. Even here, though, the model has been weakened by the rise of AI agents and the willingness of teams to use these tools sparingly, and the "viral seat expansion" story isn't what it was in 2019.
The honest summary: per-seat is going to remain the right model for ~20% of the software market — the enterprise wedge, the highly-variable-customer-size wedge, and the strongest network-effect products. For the other 80%, it's becoming the wrong model, and the vendors that stay attached to it past 2027 are going to find themselves competing for shrinking share.
What this means for buyers right now
If you're shopping software in 2026, "per-seat or flat-fee?" is now the most important question after "does this product do what we need?"
Three years ago, this was a secondary question. The default was per-seat, and the only people who pushed back were finance teams at the moment of contract signature. Today, it's a leading question. It shapes how the relationship will evolve over the three-to-five years you'll be on the product.
Some practical things to ask any SaaS vendor in 2026:
- Will my bill change if I hire three more people? If the answer is "yes, by a lot," weigh that as part of the cost.
- What's your AI agent pricing? A clean answer is a sign the vendor has thought about the future of their model. A muddled answer is a sign they haven't.
- Can you move me to a flat-fee tier? Even per-seat vendors will sometimes negotiate this for mid-market customers. If they refuse outright, that tells you something about how flexible the relationship will be.
- What's your net dollar retention? If they share it, look at the trend. If it's below 105% and falling, the vendor will be under pressure to extract more from existing customers — which usually means raising your bill.
If a vendor refuses to move off per-seat, that's a signal about the relationship — not necessarily a disqualifying one, but a signal. They are committing to a pricing model that is structurally misaligned with how a 2026 business runs. Decide whether that's the right model for the work you're trying to do.
The quiet end
The unbundling era ended because the math caught up with it. The per-seat era is ending the same way — not with a sudden collapse, but with a slow accumulation of customers who finally figured out that "successful hiring shouldn't make my software bill go up."
The new default is flat-fee for most of the SMB and mid-market world, usage-based for infrastructure and developer tools, outcome-based for the narrow set of products that can credibly take risk on results. Per-seat retreats to the enterprise wedge where it belongs, and stops being treated as the universal answer.
The CFO who wrote the email at the top of this essay isn't going to keep writing those emails forever. She's going to switch vendors. She's going to switch to something where the bill doesn't punish the company for growing. Multiply her by every CFO at every mid-market company in 2026, and you have the actual market signal that's reshaping SaaS pricing in real time.
If you're tired of per-seat math, the alternative is here.
See our flat-fee pricing — $19/mo Personal, $29/mo Business, every module included, no seat surprises — or start a free account and find out for yourself what running a business looks like when the software bill stops being a hiring tax.