Agentic AI is beginning to change not just how software works, but how vendors charge for it.
As autonomous agents take on work once performed by employees, enterprise buyers are increasingly moving beyond traditional per-seat subscriptions and long-term SaaS contracts toward consumption-based pricing, individual actions, or measurable outcomes.
Salesforce, Intercom and Zendesk are already experimenting with variations of those models, signaling a broader shift in how AI software value may be measured and sold.
“I’m really not into a 10-year SaaS agreement anymore,” Allison Johnson, director of Americas technology partners at AWS, said of the conversations she is hearing from C-level buyers. Instead, customers are asking about “pay as you go pricing, or pay for outcome or pay for token.”
For ISVs and channel partners, that creates a more complicated commercial question: how do you price AI when customers increasingly want to pay for what an agent actually accomplishes rather than simply for access to the software?
AI agents challenge the traditional SaaS pricing model
The market is already experimenting with several alternatives.
Salesforce, for example, offers consumption-based Agentforce pricing through Flex Credits, with individual agent actions consuming credits. Its current rate card assigns 20 Flex Credits to standard and custom Agentforce actions and 30 to voice actions.
Intercom has moved even closer to outcome-based economics with Fin. Customers are charged only when the agent produces defined outcomes, such as successfully resolving a customer issue, rather than for every attempt the AI makes.
Zendesk is similarly expanding outcome-based pricing around verified resolutions, which it says are independently evaluated before customers are charged.
READ MORE: We spoke with alliance leaders at Zendesk and Dynatrace in June about the shifts in customer demand related to AI pricing and usage.
The emerging market can be roughly divided into four models:
- Traditional subscriptions
- Consumption-based pricing tied to usage or tokens
- Action-based pricing tied to work performed by an agent
- Outcome-based pricing tied to a completed business result
Vendors test consumption and outcome-based AI pricing
Those approaches point toward a market in which the unit being sold may increasingly be an action, a resolution, or another measurable business result.
Johnson said business application vendors have generally moved faster because their outcomes can be easier to identify. A customer service agent resolving a ticket, for example, provides a comparatively clear event against which a vendor can charge.
But defining an outcome is only part of the problem. Moving from seats to outcomes sounds straightforward until vendors have to define exactly what constitutes a billable result.
Consumption can be measured in tokens or actions. Outcomes require agreement over whether the software actually produced the business result a customer expected — and who bears the cost when it does not.
“How do you check that that outcome actually happened? Like, how do you get the customer to opt into being charged?” Johnson said. Those questions require changes not only to product engineering but also to billing, measurement, and contracting.
Partners face their own outcome-based pricing challenge
That shift also reaches beyond ISVs.
Johnson said consulting partners are encountering similar pressure as customers question traditional time-and-materials engagements and instead ask for the measurable results they are receiving from a project.
That can require partners to modify internal billing systems and determine how they document and invoice against outcomes.
The opportunity for the channel, then, extends beyond reselling new AI tools. Systems integrators and consulting partners can help software vendors redesign their products and commercial models around agentic AI and reconsider how they price their services.
For partners, however, efficiency and risk become two sides of the same pricing shift. An outcome-based engagement can let a provider benefit when AI allows it to deliver work faster than a traditional billable-hours model would reward, but it can also put more revenue at risk when the agreed result takes longer to achieve or isn’t achieved at all.
Why subscriptions aren’t disappearing yet
Still, Johnson does not expect subscriptions or large software agreements to disappear.
She compared the current shift to the industry’s earlier transition from on-premises software to SaaS. That transformation unfolded over years, with different vendors and industries moving at different speeds.
“There’s always going to be a place for those big deals,” Johnson said. But she added that vendors increasingly need another option because customers are demanding greater flexibility.
That suggests the near-term future of AI pricing is less likely to yield a single replacement for SaaS and more likely to be a growing mix of subscriptions, consumption, actions, and outcomes.
For channel partners and ISVs, the immediate question is therefore not whether every contract becomes outcome-based. It is whether their commercial models can keep pace as customers gain more ways to connect what they spend on AI to what those agents actually accomplish.