For years, SaaS companies had a simple way to turn software usage into revenue. Count the people using the product, multiply them by the price of a seat, and call it predictable growth. It worked because humans were the ones doing the work.
Agentic AI breaks that logic.
When software can research, respond, qualify, update records and complete workflows on its own, the number of humans touching the interface becomes a weaker measure of value. The question is no longer how many people use the software. It is what the software actually gets done.
That is where outcome-based SaaS pricing enters the picture. By 2028, it could move from an experimental pricing idea to a mainstream commercial model, forcing SaaS vendors to rethink revenue and buyers to rethink what they are actually paying for.
The Breaking Point for Per-Seat SaaS
Per-seat pricing was never a stupid model. It solved a very real problem.
Software vendors needed a simple way to estimate value. More employees using a platform usually meant broader adoption, more features consumed and greater business dependence. A seat became a practical proxy for value. It also gave vendors predictable recurring revenue, which made budgeting and forecasting easier for both sides.
The problem is that AI changes the unit of work.
An employee may still use a CRM, help desk or marketing platform. But increasingly, an AI agent can perform parts of that employee’s workflow without needing its own traditional login or human seat. The software is no longer waiting for someone to click a button. It can act.
Deloitte’s 2026 research says AI agents could give one user the power of many users, reducing the number of seats organizations need and potentially disrupting traditional SaaS revenue models.
That creates a commercial problem for vendors. If the software becomes more capable while the number of billable humans falls, seat count stops tracking value as closely as it once did.
Per-seat SaaS is dying because AI is shifting software from something humans use into something that can perform work itself, making the number of human users a weaker measure of value.
For buyers, the frustration runs in the opposite direction. Companies have spent years paying for broad software capabilities that employees may barely touch. AI makes that tension harder to ignore. Why pay for another seat when an agent can execute the task?
That is the real breaking point. It is not simply that seats are expensive. It is that the seat is becoming the wrong economic unit.
Usage-Based Vs Outcome-Based Pricing
The first response from the SaaS industry is unlikely to be pure outcome pricing. It is more likely to be consumption pricing.
The difference matters.
Usage-based pricing charges for what the system consumes or executes. That could mean API calls, tokens, compute, workflow runs or other measurable activity. It is easier to track because the vendor can count what happened inside the system.
OpenAI’s Codex pricing change on April 2, 2026 shows this transition clearly. OpenAI moved Codex from a per-message model to pricing aligned with API token usage. The model measures input, cached input and output tokens, with credit consumption depending on the work performed and tokens generated. OpenAI said the change was intended to align credit usage more directly with actual model activity.
That is consumption-based pricing, not necessarily outcome-based SaaS pricing.
The distinction is simple but important. A customer can consume 10,000 tokens and still get a poor answer. An agent can run a workflow and still fail to solve the underlying problem. Under usage pricing, the buyer can remain responsible for that failed execution.
Outcome pricing changes the bargain.
The customer pays for a recognized business result. A support issue is resolved. A qualified lead is generated. A meeting is booked. A defined task is completed.
That shifts part of the execution risk back to the vendor.
The real question therefore moves from ‘How much AI did we use?’ to ‘What did the AI accomplish?’ That is the foundation of outcome-based SaaS pricing.
How Agentic AI Forces the Shift to Outcomes
This is where SaaS starts becoming something closer to service-as-software.
Traditional software gives a person the tools to perform a job. Agentic AI can take responsibility for parts of that job. It can interpret a request, decide what steps are needed, use connected systems and complete the workflow.
Once that happens, value becomes easier to observe.
Consider the difference between an AI assistant generating a response and an agent actually resolving a customer issue. The first can be measured through usage. The second can be measured through completion.
That distinction is driving new thinking around outcome-based SaaS pricing.
In February 2026, Salesforce introduced the concept of the Agentic Work Unit, describing it as a discrete task accomplished by an AI agent. Salesforce argues that tokens measure how much AI “talks”, while the Agentic Work Unit is intended to measure the work actually accomplished.
That is a meaningful shift.
Tokens describe machine activity. A completed task describes business activity. As agents become better at executing work from start to finish, the second measure becomes more commercially interesting.
This also explains why outcome-based pricing is gaining attention now rather than years ago. Earlier AI tools mostly assisted humans. The buyer still owned the final execution. Agentic systems can increasingly take that execution burden themselves.
The pricing model has to catch up with the product.
If the software creates measurable work, vendors have a stronger reason to charge around that work. That is the core logic behind outcome-based SaaS pricing.
What Buyers Should Expect on Their Invoices by 2028
The future invoice probably will not be a clean switch from seats to outcomes.
That sounds neat. Business rarely is.
A more realistic model is a hybrid structure with a platform fee, included capacity and outcome-linked credits. Buyers get some predictability, while vendors get paid when the system creates measurable value.
HubSpot already provides a useful glimpse of this direction through its Breeze AI pricing. Its current pricing gives three concrete outcome-linked examples. Customer Agent costs $0.50 per resolution, Prospecting Agent costs $1.00 per lead and Data Agent costs $0.10 per answer.
That is outcome-based SaaS pricing in a much more practical form. The customer is not simply paying because an AI feature exists. The commercial event is tied to work being completed.
The invoice could therefore evolve from something like this
| 2023 Invoice | 2028 Invoice |
| Seats | Platform fee |
| Feature add-ons | Included AI capacity |
| User licenses | Outcome credits |
| Usage overages | Successful tasks or outcomes |
| Contracted access | Performance rules |
The difficult part will be defining success.
A contract cannot simply say “successful outcome” and hope everyone agrees later. Vendors and buyers will need clear rules for what counts as completion, when a result becomes billable and what happens when an agent fails.
That could include exclusions for failed interactions, human escalations or incomplete tasks. It could also require measurement windows, audit rules and agreed definitions of resolution.
This is where outcome-based SaaS pricing becomes more complicated than a simple pricing-page change. The billing model becomes part of the product contract itself.
How SaaS Vendors Must Adapt to Survive
For vendors, this transition creates a difficult trade-off.
Seat-based SaaS offered relatively predictable recurring revenue. Outcome-linked revenue can fluctuate with customer demand, agent performance and the number of successful tasks completed. That makes forecasting harder.
Microsoft’s Work IQ API shows where the market is already moving. It became generally available on June 16, 2026, using a consumption-based Copilot Credits model. Microsoft explicitly says there is no separate Work IQ API subscription, SKU or per-user license.
That matters because it shows how a vendor can separate access from the unit being consumed.
Yet vendors still need to solve the training lag problem. If an agent needs time to understand a customer’s data and workflows, someone has to carry that cost before outcomes become reliable. Pilot phases, onboarding limits and defined usage allowances may become part of commercial contracts.
Legal teams will have another headache. Every outcome-based SaaS pricing agreement needs a precise definition of success. Otherwise, the invoice becomes a negotiation every month.
Conclusion
The most important change in SaaS pricing is not that vendors may charge per token, credit or completed task. It is that the industry is being forced to reconsider what software value actually means.
Seats measured access. Usage measures activity. Outcomes measure work.
That does not mean every SaaS company will abandon subscriptions by 2028. In fact, the more realistic future is a messy hybrid where platforms combine recurring fees with consumption and performance-linked charges. But the direction is difficult to ignore.
Outcome-based SaaS pricing will reward vendors that are willing to put their economics behind the results their software creates. For buyers, the wake-up call is equally simple. Start auditing your software contracts now. If the product is doing the work, ask why the invoice is still counting the people.
