Investment Gems

Seat model breaks, pricing shifts to usage and outcomes, and the winners are the ones who control the new meter

The “SaaSpocalypse” take is seductive because it’s clean. AI agents show up, humans do less clicking, and suddenly all those per seat subscriptions look like dead weight. It sounds like the same story we’ve seen before: a new layer arrives and eats the old one.

But if you’ve spent any time inside an enterprise, you know the enterprise doesn’t work that way. Nobody wakes up one morning and uninstalls Salesforce, ServiceNow, SAP, Workday, Microsoft, because an agent got smarter. Those systems are not “tools,” they’re the rails. They are where data lives, where permissions live, where audit trails live, where compliance lives, where approvals live. They are the system of record and the system of action.

So no, SaaS isn’t dying. What’s dying is the idea that the right unit to bill for is “a human seat.”

And that is a much bigger change than people realize.

Because the whole SaaS model of the last decade quietly assumed a stable relationship between headcount and economic output. More employees meant more seats. Growing company meant rising ARR. Even if the product wasn’t loved, it could still ride the headcount curve. Seat based pricing was a tax on organizational growth.

AI breaks that relationship. Not overnight, not everywhere, but enough to matter. If a company can run flatter, if one agent can do the repetitive work of a team, you end up with something that feels weird at first: output grows, headcount doesn’t. Or headcount shrinks.

And once headcount becomes a variable you can reduce without destroying output, every CFO looks at per seat SaaS and starts asking a question they didn’t used to ask out loud: why am I paying a headcount tax?

That’s the re-metering event.

The fight isn’t about whether software tools remain. They will. The fight is about what the vendor gets to charge for when the human seat is no longer the natural unit.

You can already see the shape of it. Vendors are trying to move the meter from “how many employees do you have” to “how much work gets done on my platform.” Some will do it gracefully. Some will do it clumsily. Some won’t be able to do it at all.

The new meters are pretty obvious once you say them out loud:

How many workflows executed.
How many tickets resolved.
How many calls handled.
How many sales interactions initiated.
How many documents processed.
How many API events, queries, jobs, runs.
How many agent identities operating in the system.
How many outcomes delivered.

It’s the cloud playbook all over again, just applied to business software. Cloud didn’t win because compute got more expensive. It won because the unit shifted. You stopped buying servers and started buying consumption. Unit price fell over time, but total spend exploded because usage exploded.

Agents push SaaS into the same transition. Seat counts can flatten, but activity can explode. In many orgs, the whole point of agents is to do more work, faster, around the clock. That creates more events, more workflows, more monitoring, more governance, more everything. So the total addressable market can expand even if the seat denominator shrinks.

That’s why the naive “headcount down 20%, SaaS revenue down 20%” math is too linear. It’s a decent instinct, but it misses the vendor response. Vendors don’t sit there and accept a 20% haircut. They redesign packaging, add AI and automation SKUs, introduce usage based pricing, bundle platform fees, and add minimum commits. They try to keep the revenue line attached to value, not to seat count.

So where does the repricing war actually get won?

It gets won wherever the vendor has enough leverage to change the meter without getting disintermediated.

And leverage in enterprise software comes from three things that don’t show up in a demo:

Data gravity, meaning the data is already there and it’s painful to move.
Workflow lock in, meaning real business processes run through the system.
Governance, meaning permissions, auditability, compliance, security live there.

If you own those, you can tell a customer “sure, reduce seats, but you’re going to pay for the automated work and the agent layer now.” And they will, because they can’t rip you out without breaking the business.

If you don’t own those, you’re in trouble. The shallow point solution that’s basically a UI for humans to do repeatable work is exactly what agents eat first. Not because the tool is bad, but because the tool’s value was the human doing the work inside the UI. Once the work moves to an agent, the UI stops being the product.

That’s the split. Systems of record and workflow platforms tend to survive and reprice. UI wrappers and seat tax tools tend to get squeezed.

There’s another uncomfortable wrinkle here, and it matters for investors: the shift to usage and outcomes changes the quality of revenue.

Seat based subscriptions are predictable. That’s why SaaS got valued like a bond for a decade. Usage based revenue can be bigger, but it’s more cyclical. If customers can dial down usage, they will. It creates more variability. It also forces vendors to talk about unit economics in a way the market hasn’t demanded from pure seat SaaS.

So part of the “re-metering” transition is that some SaaS names will look less like stable subscriptions and more like consumption businesses. That doesn’t mean worse, it just means different. And the market will have to relearn how to value it.

Now, if you want to be practical about this, you watch for a handful of tells. You don’t need to guess from vibes.

First, net revenue retention. If seats are compressing and the company can’t replace them with usage or outcome based monetization, NRR rolls over. If they can reprice successfully, NRR holds up even if seat counts flatten.

Second, what happens at renewal. Seat rationalization shows up in step changes at contract renewal, not as a smooth curve. That means the damage, or the repricing success, comes in waves.

Third, the structure of AI and agent SKUs. Are customers paying incremental dollars for AI features, or are vendors giving them away to defend seats? If it’s the latter, margins can get ugly because inference costs are real.

Fourth, gross margin trends. Re-metering often pulls vendors into consumption models with real cost of goods sold. If gross margins degrade and pricing doesn’t adjust, you’ve got a problem.

Fifth, contract language. Minimum commits, platform bundles, agent identity pricing, outcome pricing. Those tell you whether the vendor has the negotiating power to set the new meter.

And then there’s the hidden second order effect that connects back to the infrastructure trade you’ve been focused on. If SaaS gets re-metered to “work performed,” and agents start doing more work than humans ever did, that means more inference, more compute cycles, more monitoring, more security, more orchestration. It pushes demand down the stack. Even if seats shrink, the machine economy expands. That is why the “SaaS is dying” conclusion doesn’t follow from “seat model breaks.” The spend just moves and the profit pools reshuffle.

So the way I’d summarize the whole thing is simple:

SaaS is not getting deleted. It’s getting repriced.
The unit is shifting from humans usage to work done.
The winners will be the platforms that own the workflow rails and can charge for automated activity.
The losers will be the ones whose product was basically a seat based interface for work that agents can now do without them.

And if you’re underwriting a specific name, the questions become much sharper than “do they have an AI strategy.”

Can they move the meter without losing the customer?
Can they charge for agents as first class identities?
Can they attach usage and outcomes to the core contract?
Can they do it without gross margin getting eaten by inference costs?
Do they control the system of record, or are they a layer on top that can be bypassed?

That’s the repricing war. It won’t be announced in a press release. It will show up quietly in renewals, in retention, in margin structure, and in whether the vendor is still the one collecting the toll when the work is no longer done by humans.