Per-Seat Pricing After the 2026 SaaS Selloff: What AI Agents Changed

In February, AI agent launches set off a selloff in software stocks built on charging per user. Here is what investors feared, how Salesforce, ServiceNow and Adobe are reporting AI revenue now, and what the rebound does and does not prove for SaaS founders.

One trigger investors named was a set of plug-ins. Reuters reported that Anthropic “launched plug-ins for its Claude Cowork agent on Friday that would automate tasks across legal, sales, marketing and data analysis,” and by Tuesday, February 3, a selloff in software and data stocks had deepened, with some investors pointing to the release as the main culprit. The fear underneath it was specific: that AI agents break per-seat pricing, the model much of enterprise software is built on.

Where it stands: Software stocks have since bounced. Between April 13 and October 12 the iShares Expanded Tech-Software ETF (IGV) climbed 41.46%, 24/7 Wall St. calculated; measured from January, though, the gain was just 5.34%.

The pricing question has not gone away with the rebound. For founders selling software, the useful part of this year is less the stock chart than the evidence on how investors now value revenue, and how the largest incumbents are changing what they report.

Why investors turned on charging per user

Reuters quoted Schroders analyst Jonathan McMullan on the day of the selloff: “Investors are aggressively repricing these areas as the historical ‘visibility premium’ erodes; the speed of AI advancement makes long-term valuations harder to defend, particularly as AI tools allow businesses to do more with fewer staff, threatening the traditional model of charging per software user.”

The “visibility premium” is the extra value investors used to pay for predictable subscription revenue. A seat-based contract grows when the customer hires. If agents let that customer do the same work with fewer people, the contract can shrink even while the software does more.

The Motley Fool summarized the first-half fear the same way: “Investors worried that artificial intelligence (AI) agents would chip away at the per-seat licensing model on which much of enterprise software is built, and the selling was severe.” Reuters, covering Adobe in March, described a broader version: AI tools and agents “that many fear would be able to disrupt traditional software subscription models and give way to quicker and cheaper ways of creating products.”

Coding tools sharpened it. Betterment’s Ben Bakkum wrote that “The mere prospect of a DIY software future has turned investor sentiment against the Software as a Service (SaaS) businesses.” It fits a broader line about the AI trade that Natixis’s Garrett Melson gave Reuters in October: “When AI is working, it sucks all the oxygen out of the room for the rest of the sectors.”

What the selloff looked like

CNN reported that a software-industry ETF fell 5.69% on February 3, its worst day since April. Westlaw’s parent, Thomson Reuters, lost nearly 18% in the session. By late May, Salesforce shares had “declined nearly 33% so far this year, after falling more than 20% in 2025,” Reuters reported, as Wall Street adopted the label “SaaSpocalypse.”

Valuations reset hard. Betterment noted in May that stocks like Adobe and Salesforce traded “at roughly a 50% discount to early 2025 levels.” Reuters also recorded that traders and analysts thought “investor fear often outweighed company fundamentals,” which matters for founders reading the drop as a verdict on the business model itself.

How incumbents are re-pricing: the evidence so far

None of the big vendors has abandoned seats, and the public record on new pricing models is thin. What is visible is a shift in what they charge for and what they report.

ServiceNow leans partly on usage. The Motley Fool notes that it “charges, in part, based on how much its platform is used rather than solely on seat counts,” and that with its Now Assist AI product, customers spending $1 million or more a year grew by more than 130%. The Fool’s author argued this is “a model that may hold up better in a world of AI agents than the seat-based approach that has weighed on the group this year.” That is the author’s view, not a company claim.

Salesforce now puts a number on its agent business. Its Agentforce agents hit a $3.9 billion annualized run rate, 210% above the year-earlier quarter, per a September Motley Fool report.

Adobe reports a separate AI line. Its third-quarter release on sec.gov says “Adobe AI-first ARR grew more than 150% year over year,” after Reuters reported in June that this AI-first recurring revenue had tripled past $500 million. Total Adobe ARR exiting the quarter was $27.50 billion, alongside record quarterly revenue of $6.76 billion.

Company AI metric it now reports Reported by
Salesforce Agentforce run rate $3.9B, up 210% Motley Fool, Sept.
Adobe AI-first ARR up more than 150% Adobe 8-K, Sept.
ServiceNow Now Assist $1M+ customers up 130%+ Motley Fool, June

The common thread: the AI revenue sits on its own line, where investors can see it grow, rather than buried inside seat counts.

The counter-case: overdone, and the numbers held

Not everyone accepted the obituary. JPMorgan strategists wrote in February that “The market is pricing in worst-case AI disruption scenarios that are unlikely to materialize over the next three to six months.” Morgan Stanley’s Katy Huberty went further: “We believe the dislocation in U.S. Software valuations is sentiment-driven, not fundamental.” Aurelion Research called it “sentiment driven” too.

Adobe is the test case. It posted record revenue and still traded down, and XTB’s analyst framed the gap: “The market assumes that AI could change this situation in the future. Adobe’s results show that this change has not happened yet.”

The rebound should not be over-read either. The 24/7 Wall St. author who tallied IGV’s gain cautioned that “A 41.46% gain in six months most likely reflects investors paying more for similar forward earnings, because software profits rarely speed up that quickly.”

Per-seat pricing questions the data raises for SaaS founders

The selloff doesn’t prove that seat pricing is dead. It does show which questions investors and lenders started asking, and founders can expect to hear them.

  • Does revenue grow if the customer’s headcount falls? That is McMullan’s concern restated. A pricing model tied partly to usage or outcomes answers it differently from one tied only to logins, as the ServiceNow example suggests. For contract structure, see our guide to pricing enterprise sales contracts.
  • Is the AI revenue visible? Salesforce and Adobe both chose to report AI revenue separately. A startup with an AI product line faces the same question in a board deck.
  • Can customers walk away? Betterment’s “DIY software” worry is that buyers can build or automate more themselves. The buyer’s side of that calculation is covered in our look at why businesses accept software price hikes while planning exits.
  • Will lenders price the risk? The Federal Reserve’s May Financial Stability Report said investor sentiment soured on “concerns about the potential for advances in aI to disrupt industries, particularly software,” and that “The software sector had become the largest sector in private credit portfolios.” SaaS founders raising venture debt are borrowing in that market. The wider squeeze on companies outside the AI label is in our piece on non-AI startup funding in 2026.

For scale, the incumbent at the center of the design-software debate is not shrinking yet: Adobe’s ARR exiting its fiscal third quarter was $27.50 billion, by its own filing.

Frequently Asked Questions

Why are AI agents a threat to per-seat SaaS pricing?

The worry, as Schroders analyst Jonathan McMullan put it to Reuters in February, is that AI tools let businesses do more with fewer staff, which threatens the model of charging per software user. If a customer needs fewer people logged in, a price tied to headcount shrinks even if the work done grows.

What is the SaaSpocalypse?

SaaSpocalypse is Wall Street shorthand, cited by Reuters, for the gloom over software-as-a-service companies in 2026. Investors feared that coding tools and agents from AI labs such as Anthropic and OpenAI would replace software that businesses had been renting by subscription.

Did software stocks recover in 2026?

Partly. IGV, the iShares software ETF, gained 41.46% between April 13 and October 12, yet its 2026 gain on that date was 5.34%. The 24/7 Wall St. writer who ran those numbers attributed most of the rally to investors paying higher multiples, not faster profit growth.

How are software companies changing their pricing for AI?

Public examples are limited, but some patterns show up. The Motley Fool notes ServiceNow charges partly on platform usage rather than seats alone, and both Salesforce (Agentforce) and Adobe (AI-first ARR) now report AI revenue as its own metric.

Sources

This article is general information for founders and is not legal, accounting or investment advice. Funding and market figures are as reported by the sources cited on the dates shown and can change.