Saas Platform Seats Are Not Dying. The Margin Is

The Seat Is Not Dying. The Margin Is.

Testing the consensus view on SaaS competitive dynamics against the quarter just reported

The assumption

Look at nearly any software investor note, board presentation, or conference panel from the last year and a half, and you’ll see the same idea: per-seat SaaS is over. The reasoning makes sense. Software pricing has usually followed headcount, and now agents are taking over tasks people used to do. That means the billing base for seat-priced platforms should shrink, putting incumbents at risk, while AI-native challengers without seat-based business models are expected to come out ahead.

This argument is strong, and it can be tested. The most recent quarter, covering April to June 2026 and reported between late May and mid-August, gives us the first clear look. That’s because it’s the first time the big platforms had agent products widely available and in use, with new pricing models actually in the market instead of just being announced.

We checked this idea by looking at twenty public platform and application software companies. We compared each company’s latest reported quarter to the same quarter last year, and we also read management’s comments to understand what was really happening behind the numbers.

The result is that the argument doesn’t just fall apart at the edges, but right at the core.

What the evidence says

Let’s start with the overall numbers, since that’s where we’d expect to see the effect first. If agents were really hurting seat-priced software, this group should be slowing down. But that’s not happening. The median year-over-year revenue growth for these twenty companies was 22.0% in the latest quarter, up from 19.5% a year ago.

If you leave out Palantir, which grew revenue by 92.8% to $1.94 billion from $1.00 billion, the median was 21.9% compared to 19.4%. The standard deviation of growth rates was 7.6 points versus 7.8. In other words, there’s no real statistical difference. The group didn’t slow down or become more uneven.

It’s important to understand why this idea became so widely accepted. The reasoning wasn’t careless. Seat pricing really does track with headcount, and agents really do take over tasks people used to do.

There’s also real history here. Like when software moved from perpetual licenses to subscriptions. Where a pricing model faded faster than the companies relying on it. The argument didn’t fail because of bad logic, but because of timing.

It assumed customers would replace people with agents quickly enough to shrink seat counts before vendors could add a new revenue stream. The latest quarter shows that vendors moved faster.

What about the seats?

Now, let’s look at the specific claim about seats. Three of the biggest seat-priced enterprise software platforms were asked about seat compression, and all three answered with actual data, not just talk.

ServiceNow grew revenue 24.0%, to $3.99bn from $3.22bn, and management took the question head-on:

“Are we worried about seat compression? Not at all. Our addressable user base is growing and 50%, 5-0 of our net new business is already non-seat-based.”

It went further: “A lot of people are like, are seats going away? No, actually, active seats going up.” Seat pricing is being kept deliberately, “because customers prefer it for predictability.”

Salesforce, the classic seat business, reported that “our largest applications, sales and service saw year-over-year seat growth with humans and agents both expanding on the platform.”

Management was blunter still on its largest deals: “7 of the top 10 deals added seats, new seats. This is the new way that we have to monetize AI.” monday.com, mid-market and seat-native, reported it is “still seeing a double-digit seat growth year-over-year in enterprise.”

Growing not strinking

That’s not a shrinking billing base. It’s a growing one.

The companies that use consumption-based pricing show the same trend, just from a different angle. Datadog reported, “Q2 revenue was $1.12 billion, up 36% year-over-year,” with “quarter-over-quarter revenue growth… the highest since Q2 2022” and a record $115 million added in one quarter. Importantly, growth “accelerated with our broad base of customers, excluding AI customers to the high 20s.”

That detail matters because some might say consumption vendors just serve AI labs, but even without those, their enterprise customers are still growing.

Atlassian also reported, “Total revenue grew 28% to $1.8 billion. Cloud revenue surged to $1.2 billion… up 31% year-over-year. RPO grew 44% year-over-year to $4.8 billion.” Having a backlog that grows much faster than revenue is the opposite of what you’d expect if seat deflation was happening.

What is actually happening

The most interesting part isn’t just that the assumption is wrong, but what the evidence shows instead. Last quarter’s reports reveal three things, and none of them is a decline in seats.

1. Seats are no longer the main growth driver.

They’re now the baseline, with a second meter added on top. monday.com launched a combined seat-and-credit model in May and explained it clearly: “Before the changes in the product and the pricing, the only way customers could expand was to add more people and more seats.

This is the first time… we see customers expand not only on the seats for humans, but on AI consumption.” ServiceNow now uses a clear “hybrid with the seat and the consumption.” HubSpot has gone even further, letting customers “swap seats for credits and manage their budget.” So, it’s not about replacing seats with consumption. Instead, seats are the steady base, and usage-based pricing is layered on top—this dual-meter approach is where the growth comes from.

2. Even more importantly, the key issue on the income statement has shifted.

For years, the main competition in SaaS was about revenue—who owned the workflow, the seat, or the budget. But last quarter, the real pressure showed up in the cost of revenue instead.

ServiceNow said the pressure came from a faster-than-expected ramp-up with hyperscalers and higher AI usage, noting that “even though we had a little pressure on gross margin, we held the operating margin flat.”

That’s a fair explanation, but it’s important to point out that, on a standardised basis, the margin compression is 680 basis points—much more than “a little.” The difference between how it’s described and what actually happened is part of the story.

What is the tell?

The exceptions are the tell. Atlassian and Palantir both expanded margin. Salesforce held flat despite what an analyst on its call characterised as “surging token demand.” So the cost of intelligence is real, it is arriving now, and it is distributed extremely unevenly.

That dispersion — not seat counts — is the new competitive variable. The vendors absorbing the damage are those pushing inference into a large installed base on their own COGS, ahead of the pricing that recovers it. Those expanding margin have an offsetting mix shift, a pass-through structure, or a deployment model where the customer carries the compute.

3. Switching costs have shifted from being about features to being about context

Atlassian now calls its moat “the Teamwork Graph, which is a living contextual layer underlying the platform.”

ServiceNow highlights governance and data foundation for its CMDB. Salesforce is bringing Informatica into Data 360 and says that “our top 10 customers by Q1 AWU usage have increased their total Salesforce spend by 1.5x in the last year.”

The point (and last quarter’s data supports it) is that an agent is only as valuable as the proprietary data it can use, and that data is held by the incumbent.

Where the old assumption still has teeth

To be fair, there is some evidence for the other side. Two things support a version of the bear case, though not the one most people talk about.

The transition is costly. HubSpot’s growth slowed to 19.8%, and they said directly:

“April got off to a slow start, and the quarter that we anticipated did not fully materialize… the deliberate changes we made across product, pricing and go-to-market were a headwind… we saw a shift in the demand environment in Q2 with increased budget sensitivity.” monday.com’s growth slowed to 21.9%, and they expect net dollar retention to drop from 109% to 108% as they adjust to earlier pricing changes. In both cases, the slowdown was caused by their own re-pricing efforts, not by competitors—but it was real and significant.

There’s also a slower group in the cohort.

Asana grew 9.5% and DocuSign 8.7%—the two slowest, both using seat-based pricing, both lacking the kind of proprietary data layer that Atlassian or ServiceNow have, and neither has a second revenue stream yet. Asana’s in-quarter NRR “has improved for 4 consecutive quarters and reached 97%.”

That’s progress, but 97% still means contraction. If the seat-based argument holds anywhere, it’s here. But the reason isn’t agents replacing seats; it’s that a platform with only a generic context layer has nothing to drive a second revenue stream.

What replaces the assumption

It’s time to move past the old consensus and replace it with a more focused and practical view:

Seats aren’t going away; they’re being repriced as the base of a two-meter model. Competitive advantage is shifting from who controls the workflow to who owns the proprietary context that agents use. And the key factor for winners has moved from ARPU to cost to serve—the real competition is now about gross margin, not price per seat.

This changes what you need to look for. Seat counts and NRR are still important, but they’re not enough on their own—a platform can show strong seat growth while its unit economics get worse. For the next two quarters, watch these things:

1. The year-over-year change in gross margin, and whether inference costs are being recovered or just absorbed
2. The share of new business that isn’t seat-based—ServiceNow says it’s now 50%, but others haven’t shared
3. Whether consumption revenue is truly extra or just replacing other revenue—monday.com’s customers “top up and reach the end of their consumption bucket and then add more,” which appears incremental. While HubSpot’s seat-for-credit swaps may sometimes just replace old revenue
4. Whether the slower companies can build a strong context layer before their growth slows to match the economy

What can disprove this?

Three things could disprove this new view.

1. If ServiceNow’s gross margin doesn’t recover as the hyperscaler ramp settles, then the cost of intelligence is a permanent issue, and margins across the group will need to be reset
2. If seat counts at Salesforce or ServiceNow start to fall without a broader economic reason, then the original assumption was just early, not wrong
3. And if a challenger wins real market share in a category where the incumbent owns the system of record, then the context moat isn’t as strong as management thinks

None of those things happened last quarter*. Instead, twenty software platforms grew a bit faster than the year before, added seats, added a second revenue stream, and covered the cost through their gross margin.

* “Last quarter” means each company’s most recently reported fiscal quarter, ending April–June 2026. Revenue and gross margin are on a standardised (CapIQ) basis, which can differ from company-reported non-GAAP figures. For instance, ServiceNow’s 680bp compression in particular is standardised and larger than management’s own framing.

We are Octopus. The Global People-Powered Competitive Intelligence Agency.

Outsmart your competition. Make the unknown known. Octopus helps you gain clarity in complex markets. With clients and tentacles around the world, we deliver sharp, actionable competitive intelligence through a blend of deep primary (HUMINT) and secondary research. If you’re looking to make smarter decisions, beat the competition, and reduce uncertainty, we’re the partner you want on your side.

Please enable JavaScript in your browser to complete this form.

Name

Name Email

Email *

Submit

• X
• LinkedIn
• Instagram
• Threads
• Reddit
• Medium
• Facebook
• Bluesky