← Back
Arc Four: The Growth That Counts
Thesis: For twenty years, seat-based software revenue compounded because customers hired. AI is weakening that link from both ends: people doing more work with fewer seats, and agents doing work without becoming users at all. The seat is not dead. What is dying is the automatic expansion it once promised, and the buyers who acquire software companies have already made the pricing model itself a diligence question. The danger for a founder is timing. The operating change may reach her market later. The valuation change reaches her now. The question is no longer what the company charges per seat. It is what happens to revenue when the seats stop multiplying, and whether she can prove it.
The question arrived as item eleven in a list of fourteen, sent by the corporate development team of a Chicago software group that wanted to buy her company, three days after the management presentation. Mei had answered the first ten from memory. Item eleven asked what proportion of her revenue was attached to her customers' headcount.
She read it twice. Her practice management platform served 640 accounting firms across Australia and New Zealand. Firms bought seats. Partners, managers, accountants, bookkeepers: each logged in under a licence, and each licence was billed monthly. The answer to item eleven was all of it, and until that morning she had considered this a strength. Seats were how software was sold. Seats were how her $22m of revenue had been built, seat by seat, for fourteen years.
It was the follow-up request that undid her weekend. The buyer asked for seat counts by customer, by quarter, going back three years. Her CFO assembled the spreadsheet on Saturday. On Sunday evening Mei sat with it and saw what the buyer was looking for before the buyer had seen it. Her twenty largest firms, a third of her revenue, had not added a seat in nine quarters. The firms themselves were thriving. Two had just published record results. They were doing more work, for more clients, with the same number of people, and her software billed by the person.
Revenue had still grown. New firms had joined. Prices had gone up in 2025. The headline number climbed 14% and the board had been pleased. But the engine inside the number, the quiet expansion that had powered every year since 2012, had stopped, and nothing on her dashboard was designed to notice.
The per-seat licence deserves more credit than it gets. It solved, in one elegant unit, the hardest problem in commercial software: how to price a thing that costs nothing to copy. Attach the price to a human being. Humans are countable. Humans arrive with salaries, so a licence priced at a fraction of a salary always looks reasonable. And humans multiply, because successful companies hire.
That last property built much of the industry. A software company that priced by the seat did not merely sell to its customers. It grew with them, automatically, without an additional sales call. The customer won a new client, hired three accountants, and the invoice went up. Multiply that across a book of hundreds of customers and you have the metric that defined an era: net revenue retention above 100%, the promise that this year's customers alone would deliver more revenue next year, before a single new logo signed.
Buyers of seat-based software companies trusted that metric because it answered a question nothing else answered as cleanly: where the growth will come from. It converted the messiest judgment in an acquisition into the most predictable behaviour in commerce: companies that are doing well hire people. For twenty years, a buyer of a seat-based software company could underwrite its future by underwriting its customers' hiring plans. Nobody phrased it that way. Everybody priced it that way.
[QUOTE] For twenty years, the best salesperson in software was the customer's hiring plan.
The seat is now under attack from two directions, and the second is stranger than the first.
The first is arithmetic. AI lets a company do the same work with fewer people, and every per-seat licence is attached to one of those people. In February 2024, Klarna announced that its new AI assistant was doing the work of 700 full-time customer service agents within a month of launch. The announcement was read as a story about Klarna. It was also a warning to every seat-based vendor selling into labour-heavy workflows. When a customer's headcount stops growing, a seat-based vendor's expansion stops with it. When the headcount falls, the vendor absorbs the cut at renewal, without losing a single competitive deal, without a complaint, without a warning on any dashboard. This is what Mei's spreadsheet had recorded. Her customers' productivity had become her stagnation.
The second direction is substitution. An AI agent completing a task does not open the software's interface at all. It reaches in, does the work, and moves on to the next system. Gartner calls this agentic arbitrage, and in July 2026 put a number on it: up to US$234 billion of enterprise application spending exposed by 2030, roughly a fifth of enterprise application SaaS spend. The sentence in Gartner's announcement that belongs above every software founder's desk is the plainest one. This, the firm wrote, breaks the link between user growth and revenue growth.
The industry's own pricing moves point in the same direction. Intercom charges US$0.99 for each customer conversation its AI resolves, and nothing per seat for the agent doing the resolving. Zendesk bills for support tickets its software closes without a human touching them. Salesforce, the company that taught the world to sell by the seat, sells its Agentforce platform through consumption credits and per-conversation pricing. When the sellers of software begin pricing agents by outcome, the seat has lost its monopoly on the argument. What remains is timing.
[QUOTE] An agent does not need a licence. It needs a result.
A founder might reasonably conclude that the answer is to change pricing now, and this is where the trap sits. In March, PitchBook published an analyst note for investors on the shift from selling access, priced by the person, to selling outcomes, priced by the result. The note is candid about the shape of the journey. PitchBook calls it the SaS J-Curve: the period when seat-based revenue declines before outcome-based revenue scales to replace it, during which net dollar retention, the metric a generation of buyers used to price software companies, will experience significant volatility. The note even proposes a successor measure, net labor capture, the share of a customer's payroll the vendor's software has absorbed.
Readers of Issue 18 will remember what happened to Helen, whose recurring revenue was rebuilt by a buyer's accountants during exclusivity, and who paid for the difference several times over. Mei's version of that problem is subtler. Helen's number was wrong. A founder who reprices mid-transition has a number that is merely unreadable: a revenue bridge that dips before it recovers, presented to a buyer whose models were built for the old metric. There is no accounting treatment that fixes that. There is only sequencing. A pricing transition begun two years before a sale process is a strategy with evidence behind it. The same transition begun eight months before a process is a declining revenue line with a story attached, and diligence does not pay for stories.
The pricing model, in other words, has joined the contract stack and the revenue ledger as a thing a buyer will test. Not because buyers care about pricing theory, but because the model determines what happens to revenue in scenarios the founder does not control. The task this sets a founder is not repricing. It is readiness.
There is a soothing version of this story, and it is important to name it, because it is half true. The agents are not ready. PitchBook's own assessment is that agentic systems are at least eighteen months away from reliably replacing basic employee workloads, and that adoption will vary by region, with markets outside the United States generally moving later. Australian accounting firms are not about to hand their workpapers to autonomous software. Mei's product will bill by the seat for years yet. A founder reading the commentary could be forgiven for filing the whole subject under later.
The error in that reasoning is that it prices the founder's market and forgets to price her buyer. The likely acquirer of a $22m Australian vertical software company is not an Australian accounting firm. It is a North American strategic or a private equity platform, and those buyers do not price the present. They price the decade they are about to own. The frameworks they will use are no longer private judgment. PitchBook's note gives investors eight scored frameworks for assessing a vendor's capacity to survive the move from seats to outcomes, applied, in the firm's own words, before underwriting the vendor's transition. The first framework on the list is the business model.
This is the asymmetry that should organise a founder's attention. The valuation transition has already arrived, because it travels at the speed of a diligence questionnaire. Item eleven was not a prediction about Mei's market. It was the price of her company, being set. The buyer's underwriting model has already moved. Most founders' dashboards have not.
[QUOTE] The transition may reach her market later. It reaches her valuation now.
What separates the founders who will be hurt by this from the founders who will profit is not whether they move off the seat. Most, for years yet, should not. It is whether they can answer the buyer's question with a number instead of a silence.
Mei's response, in the months that followed, was not to reprice her product. It was to instrument it. She and her CFO computed the company's seat exposure: the share of revenue contractually tied to the count of humans logging in, by customer and by cohort. They traced the provenance of every expansion dollar three years back: new seats, new modules, price. They built the renewal model both ways, asking what happens to the book if the twenty largest firms cut headcount by 10%. And with new customers only, leaving the existing book untouched, she began piloting a hybrid price: a platform fee for the firm, seats above it, and a usage line for the automated workpaper checks her product had begun performing on its own.
None of this is a pricing revolution. It is bookkeeping for a question that did not exist three years ago. But it changes what happens in the room. The founder who knows her seat exposure, and can show what replaces the seat when it stops multiplying, is selling a business whose future has been examined by its owner. The founder who cannot is asking the buyer to do that work for her, and buyers charge heavily for homework.
The seat, it turns out, was never the asset. It was a billing convention that happened, for twenty years, to sit on top of the asset: software a customer cannot run its business without. That software is still there. What has ended is the era when the invoice grew by itself. What has begun is the era when a founder must know, precisely, where growth comes from once it no longer does.
Four questions reveal most of what an acquirer will find. What share of revenue is contractually tied to the number of humans who log in? If your five largest customers cut headcount by 15%, what happens to the renewal book? Of last year's expansion revenue, how much came from new seats, how much from new products, and how much from price? And if you had to charge for what your product accomplishes rather than for who it admits, could you measure the accomplishment?
Seat exposure is now part of acquirability, because it changes what a buyer can underwrite. Nexxit helps founders find, measure and reduce the risks that sit beneath headline growth: the dependence, the contracts, the numbers, the pricing model, the durability of expansion. The Assessment is the entry point to Nexxit, Cube Capital's acquirability programme. Begin the free assessment at nexxit.ai (link "nexxit.ai" to https://nexxit.ai).
Founders and transactions described in Founder Thesis are composites drawn from recurring patterns in software M&A, except where companies are named.
Sources: Gartner, press release, 1 July 2026; PitchBook, Through the Looking Glass: The Race to Build Enterprise AI, Analyst Note, 9 March 2026; Klarna announcement, February 2024; Intercom, Zendesk and Salesforce published pricing.
Sell-side technology M&A advisory. Sydney, Australia.
© 2026 Cube Capital Pty Ltd. All rights reserved.
Cube Capital provides M&A advisory services to wholesale clients only. This website does not constitute financial product advice. Advisory services are provided in accordance with applicable Australian financial services law.