The per-seat subscription model has been the primary unit of SaaS billing for about twenty years, and for the most part, that period it was reasonable. Adding one more user hardly cost anything, so charging one more user a fixed monthly fee was the recurring revenue engine that established Salesforce, Atlassian, and ServiceNow. Gross margins stayed nicely at 80 to 90%, and the entire vertical was valued based on a multiple of ARR that reflected those economics holding true.
AI disrupts that calculation. Each inference call incurs real expenditure, seats cease to reflect appreciation when agents perform tasks instead of humans, and the traditional SaaS expansion levers -add more seats, add more modules -do not work straightforwardly when a product that previously had to be operated by twenty people can now be done by two with only a model in the loop.
Due to this, 2026 is marking the start of a decline in SaaS pricing stability for the first time in a long while. IDC estimates that 70% of software vendors will no longer rely solely on per-seat pricing by 2028. Mixed pricing schemas are rapidly gaining popularity. Performance-based contracts are transitioning from pilot experiments to real revenue lines. And the purchasers across the table
Why the Seat Stopped Working
Agentic AI is the immediate cause. Think about an AI agent preparing contracts, matching payments, sorting help requests, or creating advertising text – these are tasks that had to be done by a known person using a computer. Once the connection between the work and the worker is broken, seat-based pricing begins to appear quite random. For example, a hundred people who use a customer relationship management system can achieve the same results as twenty people assisted by a collection of AI agents, from which it follows that the revenue from seats decreases by 80% while the business impact remains unchanged. In fact, this scenario was even one of the causes of the 26 February SaaS market crash that wiped out $285 billion of market capitalization within 48 hours as investors finally understood the computations.
Furthermore, the core cost structure has been altered. With traditional SaaS, the marginal costs were almost zero. On the other hand, SaaS with significant AI content does. A single user query leads to a model, a model requires a GPU, and the GPU time is the one being charged. Bessemer’s recent pricing research revealed that AI-native companies are operating at 50 to 60% gross margins, unlike the 80 to 90% margins the category is typically familiar with. The price has to account for that computing power, so it needs to be somehow linked to usage. You just cannot offer a flat fee for something that costs you a different amount to provide each time without
The New Pricing Vocabulary
There are just four models that will be responsible for most of the work in 2026, and each of them was born due to a different aspect of the old model failing.
Usage-based pricing is actually the response picked most often. Tokens used, API calls performed, workflows run, documents handled the unit varies, but the principle is the same. You pay for what the system does. It aligns very well with vendor cost structures and it is perceived as fair by customers when it works. The catch however is forecast. Enterprise finance teams despise having variable bills, and from the vendor side, it is also harder to forecast usage-based revenue, which is one of the reasons why Wall Street has traditionally discounted it versus pure subscription.
Outcome-based pricing is the stronger, more radical version of that evolution. Intercom’s Fin charges $0.99 per successfully resolved support ticket. Zendesk AI Agents ranges from $1.50 to $2.00 per automated resolution. ServiceNow offers efficiency guarantees on certain AI workflows. The pitch is attractive – you only pay when the software actually works – and on the data side, it is supported, outcome-based companies are experiencing 31% higher retention and 21% higher customer satisfaction according to L.E.K. Consulting. The reality of the implementation is more difficult than the pitch. You require accurate metrics, instrumentation both vendor and buyer sides, and extensive trust between vendor and buyer about what genuinely counts.
Who’s Actually Doing This Well
The interesting case studies in 2026 will not be the pure outcome-based pioneers. Instead, they will be the incumbents trying to figure out how to layer AI pricing onto existing books of business without breaking customer trust. Microsoft has kept things simple by adding Copilot to Microsoft 365 as a flat $3 per user monthly premium. This is straightforward, but it is also a bet that customers will accept the cost even when individual usage varies. On the contrary, Google has embedded AI features into Workspace with no additional list price, which is effectively a usage subsidy out of margin and a competitive flanking move against Microsoft.
Atlassian, for example, raised cloud prices up to 10% in October 2025, explicitly citing compute costs and AI features, with the value story holding well enough that there was no real customer backlash. Clay, the data enrichment platform, made a controversial pricing update in March 2026 that simultaneously increased prices for heavy API-key users and reduced enrichment costs 50 to 90% across the most-used data points.
This company published its actual internal risk analysis along with the change, which was so unusual that it became a case study in how to communicate pricing shifts. Some customers were unhappy. Most adjusted. The move survived because the company was
What This Means for Buyers
For corporate buyers grappling with this issue, the reality is that pricing is no longer an afterthought but a major factor in vendor evaluation. Even two products with almost identical features can show very different total cost profiles by merely changing pricing models, and the difference can be as high as an order of magnitude when AI usage goes to scale.
So the right evaluation process is not so much a procurement checklist as a financial model. You have to estimate real expected usage -not pilot usage, which almost always falls short of production by that 500 to 1,000% factor -benchmark against other pricing structures, and include in your contract some provisions that allow you to switch models if your consumption pattern changes. Vendors are quite often open to incorporating such flexibility if you request it, especially if you are an early buyer.
Staying current on how this space is moving is also genuinely harder than it used to be. Pricing changes, model cost drops, and agentic product launches come fast enough that quarterly planning cycles struggle to keep up. Dedicated coverage from an AI news and insights platform or similar research source is one of the cheaper ways for an enterprise buyer to stay close to what’s actually happening, because the primary reporting is faster and more specific than what most consultancy publications capture. For finance and procurement teams trying to size the real AI spend exposure across a portfolio of SaaS renewals, that kind of signal is worth more than a generic industry report released six months late.
Closing Thought
The truthful summation of 2026 is that the SaaS pricing is going through a real change of the stage rather than a change of the place in the first one. The seat model is gradually giving way but has not yet vanished from the scene. The usage-based pricing is growing but has also been responsible for quite a few budget pains. The outcome-based contracts are recording the highest customer retention numbers out of all the data but are also the most difficult to implement. By default, as a result of the above hybrid structures are the ones getting the most wins because they offer the possibility for everyone to hedge.
It is quite obvious that those companies which will come out of this time period with long-lasting pricing power are the ones creating customer value instead of following trends. If the unit of charge can be tied to a business result that can be measured, the customers will be okay with variation. However, if it is associated with the cost of the vendor, they will grumble while tolerating it. On the other hand, if it corresponds to neither, the renewal discussion will not be a walk in the park. Dara SaaS company that has extensive AI exposure is going to confront this challenge over the next eighteen months, and the successful ones will not only survive but also, in fact, characterize the post seat era of enterprise software.
