Per-Seat Growth Models Are Running Out of Road as Pricing Shifts to Usage and Outcomes

For the better part of two decades, per-seat pricing was the load-bearing wall of the SaaS business model. Simple to sell, simple to budget, simple to forecast. You added headcount, the software bill scaled with it, and everyone understood the math. But now that math is breaking, and it's breaking fast enough that revenue leaders who haven't noticed yet are about to get an education at their next renewal.

A 2026 survey of 230 software and AI companies found that roughly three in four vendors changed their pricing in the last year. Pure per-seat pricing dropped from 21% to 15% of the market in twelve months. Hybrid models, a fixed base plus variable consumption, jumped from 25% to 37% over the same window. The PricingSaaS 500 Index tracked more than 1,800 pricing changes across the top 500 SaaS and AI companies in 2025 alone, an average of 3.6 changes per company in a single year.

The vendors making these moves aren't startups experimenting at the margins. Salesforce, HubSpot, ServiceNow, GitHub, and Anthropic have all made significant pricing changes in the first half of 2026. When the largest enterprise software companies in the world all start repricing at once, that's not a signal to brush aside.

How AI broke the seat

A traditional SaaS product, a CRM, a project management tool, or a design platform has near-zero marginal cost per additional user. Serving one more seat costs almost nothing, which is why per-seat pricing worked as a growth model for so long. You added users, revenue scaled, margins held at 80% or above.

But an AI feature that calls a large language model has real per-request compute costs that vary with the complexity of the input. The infrastructure expense is tied to consumption, not to headcount. Building a flat-rate subscription on top of that cost structure creates a gap between what the vendor spends to deliver the product and what they collect for it. That gap only widens as AI capabilities get more powerful and more heavily used.

AI-feature gross margins run in the 50% to 60% range, compared to 80% to 90% for traditional SaaS. That's 30-plus points of compression, and it hits hardest on per-seat models where the price is fixed and the compute cost is not.

The result shouldn't come as a surprise. Vendors can either eat the margin erosion or reprice. Most are choosing to reprice.

The three repricing moves that matter

Three examples from 2026 show how this is playing out at scale, each with a different model and enough time on the clock to have earnings data attached.

Salesforce has been the loudest laboratory. Agentforce launched at $2 per conversation, drew pushback, pivoted to Flex Credits at roughly $0.10 per action, then added per-user licensing at $125 per month for unlimited internal use. Three pricing models in 18 months, all running simultaneously. The company closed about 5,000 Agentforce deals in its first two quarters, but only 3,000 were paid. Adoption was tepid until the credit and per-user models gave procurement teams a number they could budget around.

Jason Lemkin at SaaStr shared his own receipts on the matter. SaaStr went from 10-plus human Salesforce seats to 2 human seats and 1 API seat. Their bill went up 83%, from $12,000 to $22,000, because their 20-plus AI agents use the platform roughly 100 times more than humans ever did. Fewer seats, more usage, higher spend. That math only works for vendors who've already shifted to consumption pricing. For vendors still on pure per-seat, the equation runs the other direction: fewer humans means fewer seats means less revenue, with no consumption upside to capture.

HubSpot took the hybrid route. The company kept core seat pricing intact but layered Breeze credits on top, priced at $9 per 1,000, with paid plans receiving 500 to 5,000 included credits per month. Then in April, the Breeze Customer Agent shifted from $1.00 per conversation to $0.50 per resolved conversation. First-quarter earnings told the story: 18.2% constant-currency revenue growth, full-year guidance raised to roughly $3.7 billion, and deals over $60,000 ARR growing 37% year over year. About 90% of the installed base moved to the new pricing model.

ServiceNow put a number on the shift that should get every CRO's attention. On the company's Q1 2026 earnings call, CEO Bill McDermott disclosed that 50% of net new business now comes from non-seat-based pricing, incorporating tokens and consumption metrics. Now Assist is tracking to $1.5 billion in ACV for 2026, a 50% increase over the prior target. Half of new revenue at a $15 billion enterprise software company is no longer priced by the seat. That's not a trend piece. That's a structural fact.

GitHub made it personal

GitHub moved Copilot to token-based billing. Base plan prices didn't change. Copilot Pro stayed at $10 per month, Business at $19 per user. But the flat-rate ceiling disappeared. Every AI-powered interaction now draws from a monthly credit pool priced on per-token rates, and the cheaper-model fallback that kicked in when you ran out of requests is gone.

The developer backlash was immediate. Reports circulated of projected monthly costs jumping from $29 to $750, from $50 to $3,000. Those numbers represent edge cases, heavy agentic users running long autonomous coding sessions against large codebases. But they exposed the core tension. AI tools that do more work per session can't sustain flat per-seat economics, and the moment they stop absorbing that cost, the customer finds out what the product actually costs to run.

GitHub's chief product officer Mario Rodriguez framed it as the only way to keep an agentic Copilot economically sustainable. That framing applies well beyond developer tools. Every AI-enhanced product in the revenue stack faces the same tension between what it costs to deliver and what the customer expects to pay.

Changing the math

For CROs and VPs of Sales, the pricing shift creates three problems that show up in budget meetings, not analyst notes.

The first is budget predictability. RevOps leaders who built forecasts around $19 per user per month for their sales engagement platform or $165 per user per month for their CRM now face consumption overages, credit pools, and AI add-on tiers. The bill is becoming variable in ways that most revenue organizations aren't equipped to model. IDC forecasts that 70% of software vendors will move away from pure per-seat models by 2028. The repricing wave isn't close to over.

The second is the AI tax at renewal. Vendors are bundling AI features into higher-priced tiers, with price increases of 20% to 37% at renewal as a result. The AI capabilities may be genuinely useful, but the forced upgrade path often isn't. Revenue leaders are being pushed into paying for features they haven't evaluated, at price points they didn't plan for, because the tier they were on no longer exists.

The third, and this is the one that touches the insource versus outsource decision most directly, is that the cost of running an internal SDR program just got harder to forecast. Per-seat pricing made the build-versus-buy comparison straightforward. X reps times Y per seat per month times Z tools in the stack equals a predictable annual cost. When the tools under those reps move to consumption or hybrid pricing, the cost becomes variable. Credit pools draw down at different rates depending on usage patterns. AI features add new line items at renewal. The clean cost model that once made insourcing easy to justify on a spreadsheet now carries a level of variance that most finance teams haven't stress-tested.

Outsourced sales development models absorb that complexity differently. The vendor manages the stack, absorbs the pricing variability, and delivers a cost per output that the buyer can actually forecast. That doesn't make outsourcing the right answer in every case. But when the stack cost underneath an internal team becomes harder to predict, the comparison shifts in ways worth examining.

Industry standard TBD

There's an honest admission buried in all this data. Nobody has landed on the right model yet.

Nearly 50% of SaaS companies are actively exploring or piloting outcome-based pricing, where the vendor charges per measurable result, but only 9% have fully implemented it. The reason is that defining and measuring "outcomes" across diverse customer environments is genuinely hard. Credit-based pricing surged 126% year over year in 2025, but most teams using credits acknowledge they're a transitional mechanism, not a permanent answer.

Salesforce running three pricing models simultaneously for a single product isn't a sign of strategic clarity. It's a sign that even one of the largest software companies in the world can't figure out what the right unit of value is when AI agents do the work and humans supervise it.

For revenue leaders, the practical move is simpler than the industry-level uncertainty suggests: Audit the stack. Know which vendors are repricing, which are about to, and what the consumption math looks like under your team's actual usage patterns. Model the renewal scenarios before they arrive, not after. And pressure-test the insource cost model against a world where the tools underneath it no longer cost what they cost last year.

The per-seat era gave revenue operations something valuable: predictability. That predictability is leaving, and it's leaving faster than most organizations have adjusted for.