The market sold the seat.

Picture a finance lead on Tuesday afternoon, reading headlines about a software vendor the company pays for 200 seats. The vendor's stock has lost about a sixth of its value in a day. On the desk is the contract: the same seats, the same price, the same renewal date in September. The logins work. Monday's work got done. The vendor and the numbers are invented for this dispatch. The question is real: what, if anything, should change?

The week had a clear sequence. On Friday, January 30, Anthropic added plugins to Cowork, its agent for knowledge work. A plugin bundles skills, connectors, slash commands and sub-agents for a role, and Anthropic open-sourced 11 of them, for sales, finance, data, legal, marketing, customer support and others. They arrived as a research preview for paid users, saved locally on each user's machine, with organization-wide sharing promised for the weeks ahead.¹ Cowork itself was about two weeks old and also still a preview.² By the following Tuesday, software stocks were falling hard, and investors blamed the new tool. Axios reported that selling began in legal and data names, including Experian, the London Stock Exchange Group, Thomson Reuters and LegalZoom, then spread across the sector, and that the iShares software ETF was down more than 14 percent over six sessions after a 15 percent drop in January.³ CNN reported that an ETF tracking the industry fell 5.69 percent that Tuesday, its worst day since April, while Thomson Reuters fell 15.83 percent, its biggest one-day drop on record, and LegalZoom 19.68 percent.⁴ By Thursday the S&P 500 Software & Services Index had fallen for eight straight sessions and was down about 20 percent for the year, according to CNBC.⁵ On Friday the selloff “paused for a breather,” in Benzinga's words.⁷

My argument is that the two prices on the finance lead's desk answer different questions. The price of the stock is a forecast of a company's future cash flows across thousands of renewals it has not yet negotiated. The price of the seat is a term in your contract, and it changes only when you and the vendor change it. Last week moved the first in a day. It did not move the second at all. What it did was put a question on the table that a buyer should answer deliberately: what is each seat actually paying for?

What the market was pricing

The analysts quoted last week were mostly not saying the products had stopped working. They were saying something about margins and pricing. Constellation Research, as CNBC reported it, read the selloff as concern that AI could pressure software profits and limit how much companies can charge, not as a death knell for the industry. Rolf Bulk of Futurum Group told CNBC he expected AI-driven workflows to cannibalize some software and weigh on the multiple the sector trades on.⁵ Piper Sandler's Billy Fitzsimmons worried that the “seat-compression and vibe coding narratives could set a ceiling on multiples.”³

Seat compression is the heart of it. A great deal of business software is sold per user. If AI lets a team do the same work with fewer people, the number of seats falls even if every remaining user is happy. Jonathan McMullan of Schroders, quoted by Reuters and reported secondhand by ITPro, described AI tools letting businesses do more with fewer staff, “threatening the traditional model of charging per software user.”⁶ That is a claim about the meter, not the machine. On January 27 I wrote that a saved minute is not yet saved money; a seat no one needs is not yet a seat anyone has cancelled. Both depend on a decision someone still has to make.

The case against the panic

There was a strong counter-case, and it deserves equal space. Nvidia's chief executive, Jensen Huang, called the idea that AI would replace the software industry “the most illogical thing in the world.” Wedbush wrote that enterprises would not completely overhaul tens of billions of dollars of prior software investment to move to AI providers.⁵ Barclays' Nick Dempsey doubted that general AI models would substitute for industry-specific expertise, and Aurelion Research called the selloff “sentiment driven.”⁴ Anthropic, Axios reported, sees itself as a complement to software providers rather than a competitor.³ ITPro noted that any impact on the markets in question was yet to be seen.⁶ The tool at the center of the week was days old and labeled a preview.

Neither side settles the buyer's question, because both are arguing about the stock. The buyer's question is about the seat.

Fewer seats, or different terms

Even if seat compression arrives, it need not lower the bill. A vendor whose seat count falls has reasons to raise the price per seat, or to move to a meter it can defend, such as usage or outcomes. Figure 2 shows how the same work can produce the same bill under three sets of terms.

This is not investment advice, and it does not predict which companies win. The tool was days old and in preview. The selloff had paused by Friday. And the likeliest effect on a buyer may be a change in how software is priced rather than a smaller invoice. A falling stock is not a reason to cancel a contract, and a rebound would not be a reason to stop asking what the contract buys.

The market can reprice a vendor in a day; your contract reprices only when you do.

Before your next renewal

• Which contracts are priced per seat, and when do they renew? Those are the terms the market says are under pressure.
• What does the vendor hold that a general tool does not? Name it: proprietary data, integration with your records, accountability when the output is wrong. That is what the seat is really buying.
• Has the new tool done one real workflow? Test it on one, and measure the result, before you use it as leverage or as a replacement plan.
• What would different terms cost? Price a usage or outcome model against this year's bill before the vendor proposes one.

Back to the finance lead's desk. The stock slip says the vendor is worth less than it was last week. The contract says the company still pays for 200 seats until September. Neither tells the finance lead whether those seats are worth it. That answer comes from the work the seats support, and it is due before the renewal date, not before the next market close.