The End of Per-Seat: How AI Is Forcing B2B Software to Reinvent What It Sells

When the Pricing Page Tells the Story

Last September, Intercom — the well-known customer-support software company — quietly updated the pricing page for its AI agent product, Fin, in a way that almost nobody noticed at the time but that captures the strangest economic dilemma facing the entire business-software industry in 2026. Fin is not priced per user, the way Intercom's traditional product is. It is not priced per company size. It is priced, instead, at a flat $0.99 per resolution — meaning the customer pays only when Fin successfully handles a support ticket without escalating it to a human agent.

The implications of this small change are staggering once you sit with them. A traditional SaaS customer-support tool grows in revenue as a company hires more support agents. Intercom's old core product, like every per-seat SaaS product, depends on its customers' workforces expanding. Fin, by contrast, generates more revenue as its customers' support workforces shrink — because the better Fin gets at autonomously resolving tickets, the more $0.99 charges it generates and the fewer human agents the customer needs to employ. Intercom has effectively built a product whose revenue model is inverted from the revenue model that built Intercom in the first place. And the company is far from alone. Across the B2B software landscape — at Salesforce, ServiceNow, GitHub, HubSpot, Klarna's enterprise tools, Notion, and dozens of younger AI-native startups — the great unspoken pricing reset of the SaaS industry is underway, conducted not through bold press releases but through the steady, almost-furtive rewriting of pricing pages.

Why the Per-Seat Model Is Quietly Breaking

The per-seat pricing model, which has been the foundation of B2B software economics since Salesforce popularized it in the early 2000s, has always rested on a simple assumption: more value delivered to a customer correlates with more humans at that customer using the software. A growing company hires more salespeople, so it pays for more Salesforce seats. A scaling engineering team adds more developers, so it pays for more GitHub seats. The vendor's revenue grows automatically as the customer succeeds, and the customer's per-seat cost stays manageable because each seat unlocks a defined chunk of productivity. The math was elegant, the incentives broadly aligned, and the model became the dominant playbook for an entire generation of enterprise software companies.

Artificial intelligence has detonated that assumption. The most powerful AI features being shipped today — from coding agents to autonomous research tools to AI sales development reps — are explicitly designed to reduce the number of human seats required to accomplish a given amount of work. GitHub Copilot, which costs $19 per developer per month, can plausibly make a single developer productive in ways that previously required two; in the long run, customer headcount is supposed to grow more slowly because of it. This is great news for the customer's productivity metrics and terrifying for the vendor's revenue model. If the value of AI is measured in headcount avoided, then per-seat pricing is an actively perverse arrangement: it punishes the vendor for delivering exactly the value its product promises. Vendors have responded by pushing AI features at premium per-seat prices, but customers have begun to notice that they are paying more per seat while needing fewer seats — a trade that does not hold up under scrutiny.

There is a second, subtler problem. As AI features increasingly perform work autonomously rather than augmenting a human's work, the very concept of a 'seat' begins to lose meaning. Who is the seat-holder when an AI agent runs in the background, handling tasks at 3 a.m., processing inbound emails, scheduling meetings, drafting documents, or — as in Intercom's case — resolving support tickets? Traditional seat-based pricing assumes a human-software pairing that is being dissolved by the technology itself. Salesforce's much-discussed Agentforce product, which charges $2 per autonomous interaction rather than per user, is the most visible mainstream example of a major SaaS vendor explicitly acknowledging that the unit of value being delivered is no longer the human seat.

What Is Replacing It, and Why It Is Messier Than Anyone Expected

The pricing models scrambling to fill the vacuum left by per-seat fall into roughly three families, each with its own appeal and each with its own painful trade-offs. (Teams navigating this shift often find value in structured learning — Udemy offers practical courses on AI developer tools like GitHub Copilot and Cursor, which help measure the productivity outcomes that vendors increasingly price against.) The first is pure consumption-based pricing — the AWS-style model in which the customer pays for compute, storage, tokens, API calls, or some other quantifiable resource that the software consumes. Vercel, Cloudflare, OpenAI's API business, and most of the AI infrastructure companies have built their businesses on this model. It feels fair because the customer pays for what they actually use, and it scales naturally with the customer's success. It is also, however, an absolute migraine for procurement departments, finance teams, and budget-conscious buyers, because consumption costs are inherently unpredictable and can spike in unexpected ways. Anyone who has ever opened a surprise five-figure bill from a cloud provider knows the dark side of usage-based pricing.

The second family is outcome-based pricing — the model Intercom has embraced with Fin, in which the customer pays only when a measurable business outcome is achieved. This is the most economically elegant model in theory, because it perfectly aligns vendor and customer interests: the vendor only earns when it delivers real value. In practice, however, defining the outcome rigorously enough to bill against it is enormously difficult, and the negotiations over what constitutes a 'successful resolution,' a 'qualified lead,' or an 'accurate diagnosis' can consume more time and lawyer hours than the contract is worth. Outcome pricing also exposes the vendor to enormous variance in revenue and makes financial forecasting maddeningly difficult — a problem that public-company CFOs and their boards are deeply hostile to, because Wall Street rewards predictability above almost everything else.

The third family — and the one most companies are landing on as a practical compromise — is hybrid pricing, which combines a base platform fee with usage- or outcome-based components stacked on top. Klarna's enterprise AI tools, ServiceNow's Now Assist, and most of HubSpot's AI add-ons follow some variant of this pattern. The base fee provides revenue predictability for the vendor and budget certainty for the buyer; the usage or outcome component captures upside as the customer derives more value. It is not as elegant as either pure model, but it is the kind of pragmatic compromise that actually closes contracts. The dirty secret of the great pricing reset is that nobody — not the largest enterprise vendors, not the most sophisticated buyers, not the analysts paid to make sense of it all — has a confident answer for what the right model is. The industry is figuring it out customer by customer, contract by contract, in real time.

Untuk SaaS founder yang merancang ulang model pricing mereka, Monetizing Innovation karya Madhavan Ramanujam adalah referensi paling komprehensif. Dan bagi founder baru yang sedang membangun infrastruktur sebelum menentukan pricing, Niagahoster Business Plan menawarkan hosting scalable yang relevan.

The New Buyer's Dilemma and the Quiet Stakes

For buyers of B2B software, the great pricing reinvention is genuinely confusing in ways that the SaaS industry has not had to confront in two decades. The old job of evaluating software — counting expected seats, multiplying by the per-seat price, and adding a discount negotiation — was simple enough that procurement teams in any organization could perform it competently. The new job requires forecasting usage, estimating outcome volumes, modeling consumption spikes, and comparing competing vendors whose pricing units are not even directly comparable. A company evaluating customer support AI now has to compare a per-resolution model from one vendor against a per-conversation-minute model from another against a flat-platform-with-usage-overage model from a third, and somehow decide which represents the better deal at the firm's specific volume. The procurement skill set required for sophisticated AI software purchasing is closer to that of a derivatives trader than a traditional IT buyer.

The stakes of getting this right are substantial. Companies that embrace usage- and outcome-based AI pricing without strong cost controls can run up bills that destroy the productivity gains the AI was supposed to deliver — a fate that has befallen more than one Fortune 500 IT department in the past eighteen months. Companies that cling defensively to per-seat models risk overpaying for AI capabilities they are not fully utilizing, while their competitors capture the upside of pay-for-value pricing. And vendors themselves are stuck navigating a transition in which they have to grow new pricing-model revenue without cannibalizing the per-seat revenue that still pays for most of their operating costs. Wall Street, for its part, has begun penalizing SaaS companies that cannot clearly articulate how AI will affect their long-term revenue durability, sending several formerly high-flying public software companies into prolonged stock-price corrections that have wiped out tens of billions in market value.

What is most striking about the moment is how quiet it is. There is no industry-wide announcement that the per-seat era is ending, no Microsoft or Salesforce keynote declaring a new pricing paradigm, no clean break that buyers and vendors can mark on a calendar. There is instead a slow, halting, often-confusing migration toward a new economic model that the technology has demanded and the industry is still figuring out in real time. The pricing pages of B2B software companies — usually the dullest documents in commercial life — have become the most accurate place to read the shape of the AI transformation as it is actually unfolding inside the world's most consequential software businesses. Watch them carefully. They are the leading indicator nobody is talking about.

---

Artikel ini mengandung affiliate link. Kami mendapat komisi jika kamu beli, tanpa tambahan biaya untuk kamu.

---

Related Articles

• Best Creator Economy AI Tools 2026 — how AI enables solo creators to compete with full media companies
• UMKM Digital Indonesia 2026 — usage-based pricing and AI transforming 65 million Indonesian small businesses