AI Pricing From Usage to Agents to Outcomes

Why Most AI SaaS Companies Are Underpricing Their Best Features

I talk to SaaS companies every week that are building truly innovative and transformative AI products, including agentic AI, but are either not charging or undercharging for them because of outdated pricing models that are decoupled from the value delivered. 

For about 20 years now, SaaS pricing has followed a growth-based formula. The more customer growth, the more employees (seats), the more revenue. It has been a predictable, scalable, and high-margin business model. This model worked when software was a productivity tool.

Now, AI is changing that equation, and the playbook is breaking. 

Agentic AI specifically is changing the role of human participants to a supervisory function in software by transitioning workflows to agents that can autonomously identify an objective, break down tasks, and take action. When a single user can generate 10X the output, or when agents and automation, which are zero users, can generate 100X the output, they create value independently of the user. For example, AI agents write emails, field customer service questions, or analyze data and kick off automated workflows; charging per seat doesn’t really make much sense. In addition, many B2B SaaS companies rolled out AI capabilities as features and took a classic approach of treating it as a value-add to the software or platform environment, instead of a cost structure shift. AI capabilities that rely on data, compute, inference, and storage have a very real and variable cost. Depending on usage, these costs can spike unpredictably, eroding margins and wiping out profits. 

Some factors remain part of the core pricing strategies, including delivered value, pricing of alternatives, buyers’ willingness to pay, and, of course, your costs. But for AI-native SaaS, a shift in pricing is now on the rise from seat-based access to software to pricing models that are more aligned with customer budgets, consumption, perceived value, and the outcomes realized from AI.  

Usage-Based Pricing: A Good Start, But Not Value

With seat-based pricing no longer being viable, the natural reaction has been to shift to a usage-based pricing (UBP) model that has historically been used by infrastructure companies such as Snowflake and hyperscalers such as Google Cloud, AWS, and Azure. Usage-based pricing is quickly moving into the application layer, especially for AI-first SaaS with tokens, credits, compute consumption, or interaction counts. 

There are several key benefits to using usage-based pricing in SaaS. 

• First, it better aligns pricing with activity since customers pay for what they consume. 

• It scales with demand, so businesses don’t have to pay upfront for consumption that might not materialize months or quarters down the line. 

• Most importantly, it enables faster monetization of new features compared to SKU-based packaging. Customers eager to try out new capabilities drive up usage and therefore incremental revenue. 

While usage-based pricing does solve some problems, it also introduces new ones. 

Just because customers are using the product doesn’t necessarily mean they are seeing value. Usage is a proxy for activity, not value. Value is realized when the customer sees outcomes that positively move the needle for their business drivers (i.e., revenue, cost, reduced risk). The key is to align usage with value. In fact, high usage without a clear, measurable business impact can feel risky and give customers pricing anxiety and negatively affect adoption.

Another issue is when anchoring pricing to compute costs, you risk putting a cap on revenues as infrastructure pricing begins to come down, buyers will begin to shop on price instead of the value they get from agentic AI. 

Usage-based pricing can be a good start in certain AI use cases, but it will need to be refined over time to capture the value delivered and perceived by customers.  

Agent-Based Pricing: The Digital Workers

With the advent of AI agents, another model has recently surfaced that leverages agents as digital workers and anchors pricing to a simple and compelling ROI story of augmenting or replacing human-performed activity. Unlike buying compute, in this recurring subscription model, customers buy individual agents and deploy them to specific, scope-based tasks. 

Since this approach takes out consumption variables such as prompt design, number of transactions, etc., it helps with the predictability of AI costs far better than usage-based pricing and makes budgeting easier.  

The success of this model, however, highly depends on the ability of the agents to perform with a high degree of reliability and quality. Today, most agents still require human oversight before any action is taken, which is resource-intensive at scale, and makes the idea of digital workers less economically credible, and thus directly impacts one of pricing’s core tenets: customers’ willingness to pay. 

If, on the other hand, the AI agents reach a level where dependance of having their work double-checked is reduced to a trivial level, this pricing model can be very promising for both vendors who are looking to scale deployments of AI agents, and for buyers who are looking to increase automation without cost anxiety. 

Outcome-Based Pricing: Powerful but Complex

The outcome-based pricing model looks to create the most tangible business value for companies by aligning the price to, you guessed it, business outcomes. 

For example, an agentic AI sales prospecting tool can charge based on contacts enriched, emails sent, or searches run (activities), but one that charges based on meetings booked or qualified opportunities that grow the revenue pipeline (outcomes) is able to show clear, measurable value. And if not, the customer doesn’t pay.

The basic unit of value here is the results. The ROI is directly connected with measurable business impact, such as increased revenues, cost reductions, or hard-value task completion. This pricing model can also be tuned to the value delivered, such as a percentage of tax payment savings, unit sales, or charge-off revenue recovered. 

While outcome-based pricing is more future-proof, it’s not bulletproof. In this model, the vendor must be able to reasonably, if not definitively, show the AI is responsible for the business contribution, which takes a level of product maturity, measurability, and a well-defined scope for the AI interactions within the business workflows. If done right, the outcome-based pricing model could create the most value for customers, drive both differentiation and trust for the vendor, and strengthen pricing power to expand margins. 

The Stacked Model and Why the Future is Hybrid

Given the serious limitations of access, consumption, and outcome-based pricing, and the transition away from the traditional seat-based pricing, most B2B SaaS companies are taking a hybrid approach to pricing, and for good reason. 

With a hybrid approach, companies can create a pricing model that blends predictability, scalability, and value alignment. 

Think of this as a three-layer cake, where pricing starts with a base layer through platform access or AI agents that anchor to value and budget, while also offering customers a level of spend predictability. Product marketers can view this as pricing a packaging where you gain an entry point and offer an intuitive bridge between seats and AI. 

To scale this model, you can add an expansion layer based in part on the usage-based model to offer scalability using credits and/or tokens, capture upside revenue in tandem with growth, and offset costs. 

In cases where the product has reached a maturity level in ease of implementation, broad enterprise rollout, well-defined scope for the job to be done, and limited need for guardrails, you can introduce an optimization layer with premium pricing tied to outcomes.

Conclusion

More than just how software works, AI is also changing how customers are actually buying. They no longer pay for access to tools, but for the work performed, the outcomes delivered, and the impact created. 

This is where pricing becomes a critical component of your product strategy and Go-To-Market system. No single model can resolve the challenges that come with conflicting forces of cost variability, buyer anxiety, and ROI clarity that justifies spend. 

Taking a hybrid approach, on the other hand, is a way to build alignment between value, cost, and pricing that reconciles economic reality with customer expectations.