Tuesday, June 23, 2026

In 2008, Salesforce was charging $65 per user per month.

It felt revolutionary. Software you didn't install, didn't maintain, didn't pay a six-figure upfront license for — just a clean monthly fee per person who used it. The model was so elegant and so obviously correct that the entire industry reorganized around it. Per-seat pricing became the default architecture of SaaS for the next fifteen years.

It wasn't a coincidence that it worked so well. The per-seat model was a genuinely good approximation of value delivered. More users meant more people collaborating in the tool. More people collaborating meant more value generated. The proxy metric — number of seats — tracked well with the real metric — how much the software was actually helping the business.

That alignment is breaking down. And the reason it's breaking down is AI.

Because here's the thing nobody at Salesforce in 2008 could have anticipated: what happens to per-seat pricing when the most active user of your software isn't a person?

The Agent Problem

AI agents don't have seats.

That sentence sounds obvious until you sit with the implications. A significant and growing percentage of the work happening inside SaaS platforms today is being done by software — autonomous agents that query databases, generate reports, draft responses, monitor conditions, and trigger actions. Not humans using software. Software using software.

Under the per-seat model, this work is essentially free. The human paid for a seat. The agent that does most of the actual work in that seat generates no additional revenue. As agents get more capable and take on more of the execution work, the per-seat model creates a growing disconnect: the product is delivering more value per account than ever before, but the pricing mechanism has no way to capture that additional value.

This is not a theoretical problem. It's showing up in the financial results of public SaaS companies right now — in gross margin discussions, in net revenue retention conversations, in investor questions about whether per-seat pricing can survive the transition to agentic AI. The companies that have built their entire GTM motion around selling seats are increasingly selling something that doesn't map to how their products are being used.

And they know it. The conversations are happening in every board room and product strategy session that matters. The per-seat model is being held together right now by contract inertia, sales team comfort, and the difficulty of changing pricing at scale. None of those forces are permanent.

Why Per-Seat Was Always a Proxy

To understand what's replacing per-seat pricing, it helps to understand why per-seat pricing worked in the first place — because it wasn't actually measuring the right thing, even in its prime.

What customers are really paying for when they buy SaaS is outcomes. A CRM doesn't sell seats. It sells pipeline visibility, shortened sales cycles, fewer missed follow-ups, and higher close rates. A project management tool doesn't sell seats. It sells shipped projects, team coordination, reduced miscommunication, and accountability structures that keep work on track.

Seats were a proxy for outcomes because outcomes were hard to measure. You couldn't instrument whether a deal closed because the CRM was good or because the salesperson was good or because the market conditions were favorable. So you counted the people using the tool and assumed a relationship between usage and value, which was close enough to correct for long enough that nobody had to solve the harder measurement problem.

AI changes the measurement problem in two ways.

First, AI makes it possible to instrument outcomes much more precisely than before. You can now build products that track not just whether a user opened a feature, but whether the action the user took led to a downstream result. Did the draft email get sent? Did the sent email get a reply? Did the reply convert to a meeting? Did the meeting produce a deal? The causal chain from product action to business outcome is increasingly traceable in ways it simply wasn't before.

Second, AI shifts value creation in products from enabling actions to producing outcomes directly. When a human uses a CRM, the CRM enables the human to do things that (hopefully) produce value. When an AI agent uses a CRM, the agent is often producing the outcome itself — generating the follow-up, scheduling the call, updating the pipeline stage. The middleman between product action and business outcome gets removed, which makes the value delivered by the product much more legible.

Legible value changes pricing. When you can see clearly how much value a product is producing, the pressure to price against that value rather than a proxy metric becomes overwhelming — for customers who want to pay proportionally to what they get, and for vendors who want to capture the value they create.

The Three Models Fighting to Replace Per-Seat

Right now, three pricing architectures are competing to become the new default for AI SaaS. Each has a coherent thesis and real companies proving the model out. The winner — or winners — will define the next decade of SaaS economics.

Usage-based pricing (UBP)

The simplest departure from per-seat: you pay for what you consume. API calls, words generated, automations run, reports created. Usage-based pricing has been around for a while — AWS built its empire on it — but it's gaining significant momentum in SaaS as AI features make usage the natural unit of value.

The appeal is intuitive alignment. A company that uses the product a lot pays more because they're getting more. A company that barely uses it pays almost nothing and stays in the ecosystem without resentment. There's no ghost-seat problem — those five licenses that were bought and forgotten don't exist under UBP.

The challenge is predictability. CFOs deeply dislike billing that varies month to month without a ceiling. Sales teams have a harder time with UBP because they can't quote a clean annual contract number and close the deal. Customer success gets complicated because high bills can surprise and upset customers even when the high usage indicates high value.

The companies navigating UBP well are the ones building sophisticated usage dashboards, offering committed-use discounts that give customers budget predictability while still capturing growth, and training their sales teams to sell outcomes ("here's the ROI on your expected usage level") rather than licenses.

Outcome-based pricing

The most philosophically aligned model with AI SaaS, and the hardest to operationalize. You pay for what gets done. The hiring platform charges when a candidate is hired. The legal SaaS charges when a contract is executed. The revenue intelligence tool charges when a deal closes.

This model has a compelling story: the vendor's incentives are perfectly aligned with the customer's. You only pay when the thing you bought the software for actually happens. No results, no payment. The pitch practically writes itself.

The execution challenges are significant. Attribution is hard. Did the deal close because of the software, or because the salesperson was exceptional, or because the prospect was already sold before the first call? What happens when the customer gets value from the product but chooses not to pay by structuring the outcome in a way that doesn't trigger the pricing event? What happens to vendor revenue predictability when outcomes are volatile by nature?

The companies that are making outcome-based pricing work are doing it in narrow verticals with highly legible, easily attributable outcomes — where the causal chain between product action and business result is short and unambiguous. The broader the attribution challenge, the harder outcome-based pricing is to defend.

Value-tiered flat pricing

Less revolutionary than the other two, but more battle-tested: structure pricing around capability tiers defined by the value they deliver, not the number of users who access them. A team of five paying for the enterprise tier gets every capability the product has. A solo founder on the starter tier gets a defined capability set. Price reflects the comprehensiveness of the solution, not the headcount.

This sidesteps the agent problem neatly — agents use capabilities, not seats, so capability-based pricing naturally captures agent-generated value alongside human-generated value. It preserves the simplicity and predictability that both buyers and sellers want. And it gives vendors a natural growth lever — as you build more valuable capabilities, you can add tiers that capture that value without repricing the entire customer base.

The weakness is that value-tiered pricing still doesn't capture variable consumption. The customer who uses the product intensively every day pays the same as the customer who logs in twice a week. That's fine in categories where usage intensity doesn't correlate strongly with value received. In AI-heavy products, where the product is doing substantial computational work on the customer's behalf, that disconnect can become a real margin problem.

What This Means for Builders Right Now

If you're building a SaaS product today, the pricing architecture decision is more consequential than it's ever been — because you're building during the transition, not after it.

The instinct to default to per-seat is understandable. It's familiar to buyers, it's easy to sell, it produces predictable ARR that investors understand and value. The playbook is written. The objections are known. For early-stage companies that need to close quickly and establish revenue patterns that attract follow-on capital, per-seat has genuine short-term appeal.

But the long-term risk of building on per-seat in an AI-native product is real. You're pricing a capability-oriented product on a headcount metric that doesn't capture the most valuable things your product will do. As your AI capabilities improve, the value you deliver grows while your ability to capture it stays flat. That's a margin compression problem waiting to happen.

The founders who will be best positioned in three years are the ones who make a deliberate decision about pricing architecture now — who choose a model, build the instrumentation to support it, train their customers to understand it, and iterate on it based on real usage data rather than inheriting a legacy model and defending it as the product evolves.

The Customer Side of the Equation

It's worth pausing on what this shift means from the buyer's seat, because buyers are not passive in this transition. Some of them are pushing for outcome-based arrangements more aggressively than vendors are offering them.

The most sophisticated buyers of SaaS today — the operations leaders and CFOs who have been burned by shelfware and underutilized enterprise contracts — are increasingly asking vendors to put skin in the game. They want pricing that goes down if the product doesn't deliver and up when it does. They want alignment, not just promises.

This is actually an opportunity for AI-native SaaS companies willing to take it. A vendor who is confident in their product's ability to generate value can offer outcome-sensitive pricing as a differentiator — a signal that they believe in their product strongly enough to tie their revenue to its performance. In a market full of vendors charging confidently for features that may or may not get used, that's a real point of distinction.

The vendors who will win the next wave of enterprise SaaS sales aren't just the ones with the best products. They're the ones who can look a CFO in the eye and say: we have a clear model for how our product creates value for your business, we price in alignment with that model, and we're prepared to demonstrate the ROI before you sign the contract. That posture requires pricing architecture that supports it — and per-seat, with its tenuous link to actual outcomes, makes that posture very hard to hold.

The Next Default

Per-seat won't disappear overnight. SaaS pricing changes slowly because contracts are long, buyer habits are sticky, and sales motions take years to rebuild. For the next few years, per-seat will remain the dominant model in most categories simply because the installed base runs on it and nobody wants to reprice mid-contract at scale.

But the next generation of SaaS companies — the ones being founded today, built on AI-native architecture, designed from the ground up to deliver and measure outcomes rather than just enable activity — won't default to per-seat. They'll default to whatever model best captures the value their product delivers.

The pricing landscape five years from now will be more fragmented and more category-specific than the era of per-seat monoculture that preceded it. Infrastructure and consumption-heavy AI products will cluster around usage-based models. High-stakes vertical SaaS with legible outcomes will push toward outcome-based arrangements. Horizontal productivity tools will iterate toward capability tiers that capture the full value of AI agents working alongside humans.

What unites all of these successors to per-seat is a common thesis: price should follow value, not headcount. The tools that deliver more should charge more. The tools that deliver less should charge less. And increasingly, AI makes it possible to know the difference.

That's not just a pricing change. It's a different relationship between software vendors and their customers — one based on demonstrated outcomes rather than purchased potential. For customers, it's long overdue. For vendors willing to build toward it, it's the clearest competitive signal in SaaS right now.

The seat is empty. The agent is working. Time to price accordingly.

Built for founders who think in systems, not just features. Our AI SaaS template includes the architecture to support flexible, outcome-oriented product models from day one. [Explore the template →]