AI pricing strategy: How to price AI
AI has exposed a pricing problem that many businesses can no longer ignore.
As AI becomes embedded in products and workflows, many teams are finding it hard to evolve their pricing strategy fast enough. The addition of AI features means outcomes can now be delivered faster and at greater scale, often without adding users, licences, or obvious “units” to charge for. Pricing models that rely on structures like per-user fees or flat-fee subscriptions are struggling to keep up. As the gap between value created and value captured widens, the familiar SaaS pricing logic is starting to break down.
This article is about AI pricing strategy. In it, we explore how to price AI-native and AI-enabled services in a way that reflects value, builds confidence internally, and earns trust externally. We take a practical look at what has changed, what still holds, and how to make pricing decisions that support growth in an AI-enabled world.
Why is AI pricing strategy so hard?
AI breaks many of the assumptions that most pricing models were built on. Traditional SaaS pricing relied on a set of relatively stable conditions. Costs were predictable, usage tended to correlate loosely with the number of seats, and value scaled linearly as customers added users or unlocked additional features.
AI undermines many of these assumptions. An effective AI pricing strategy now has to account for:
- Variable and opaque costs: Compute, tokens, inference, and orchestration costs can fluctuate significantly and are often difficult to forecast with confidence.
- Non-linear value creation: A single prompt or automated workflow can replace hours of work, or in some cases, entire teams, making traditional volume-based proxies feel misaligned.
- Rapid capability creep: The product being priced today may look materially different within months, as models improve and functionality expands, leaving pricing struggling to keep up.
On their own, each of these would challenge established pricing logic. Combined, they make the SaaS pricing models we know and love increasingly redundant.
Why do many AI pricing strategies start with layering?
Faced with complexity, many organisations choose to layer AI functionality onto existing pricing models rather than overhaul pricing entirely. In many cases, this is a sensible and strategic move.
Add-ons can help teams manage cost volatility, capture additional value, and introduce AI capabilities without forcing customers into a wholesale change before trust is established. They create space to learn how AI is used, which customers benefit most, and where value is genuinely being created.
Problems arise when add-ons become a substitute for pricing decisions rather than a step toward them. When AI is priced with a “until we figure it out” mindset, organisations can find themselves anchoring prices too low, obscuring the value AI is delivering, or putting off the difficult work needed to realign pricing with outcomes.
Used deliberately, layering is a bridge. Used indefinitely, it constrains growth.
How is AI changing what companies charge for?
AI fundamentally changes how companies charge, and crucially, what they charge for.
As automation and decision support take on more work, value is increasingly created without adding users or expanding access. In many cases, customers see greater impact precisely because fewer people are involved to get a job done.
For example, a single AI prompt can now draft a contract, analyse a dataset, or resolve a customer query in seconds. The value of that interaction is not determined by how many people have access to the tool, or how long it was used, but by the work it replaces and the outcome it enables. One prompt might save an hour. Another might remove an entire step from a workflow.
As a result, pricing is being pulled towards a better reflection of what customers actually care about: work done, decisions improved, and results achieved.
Moving closer to metrics that align with customer value can make pricing harder to design and defend. An AI feature might replace a task that previously took a team member several hours a week, or even eliminate the need for a role altogether. But that does not mean the AI can automatically command a price equivalent to a full salary.
That gap between impact and price is where AI pricing decisions become harder. The value is real, but translating it into a price that resonates with customers, and they are willing to accept, requires understanding how customers perceive that value, not simply how much work the AI replaces.
How do you decide which AI pricing model to use?
There is no single “best” AI pricing model. Each model comes with trade-offs and considers a different angle on value and use. The right choice depends on what you want to reflect in the price, what is practical to measure, and how your customers think about value.
Let’s explore four different models: usage, output, outcome, and value-based pricing for AI.
Usage-based pricing for AI
Usage-based pricing sits furthest from customer value. It prices activity rather than impact, such as how often the AI is used or how much work it processes behind the scenes. Teams often choose it because it aligns closely with costs, is easy to track, and feels familiar to technical buyers. Where it breaks down is with non-technical buyers for whom this kind of pricing can feel arbitrary or difficult to predict.
Example in action
OpenAI prices access to its API based on the number of tokens processed. Customers pay per million tokens generated or analysed, regardless of whether the output is a low-value draft or supports a high-stakes business decision.
Output-based pricing for AI
Output-based pricing moves one step closer to value by pricing work done, such as tasks completed, tickets resolved, or assets generated. It is more intuitive than tokens and easier to justify commercially. However, it assumes outputs are comparable and does not guarantee business impact.
Example in action
Fin.ai prices its AI support agent based on the number of customer queries it successfully resolves. Each resolved ticket counts as a discrete unit of output, making pricing easy to understand and closely tied to visible work completed.
Outcome-based pricing for AI
Outcome-based pricing aligns pricing with results customers care about, such as error rates reduced, leads converted, or cases resolved. When outcomes are clear, attributable, and within the provider’s influence, this model can strongly align incentives and reduce perceived buyer risk. It breaks down when outcomes are ambiguous, data is shared unevenly, or success depends on factors outside the provider’s control.
Example in action
Chargeflow prices its AI-driven chargeback solution based on a percentage of revenue successfully recovered. Customers only pay when Chargeflow delivers a measurable financial outcome.
Value-based pricing for AI
Value-based pricing sits closest to customer value. It reflects business impact such as incremental profit, cost savings, or revenue generated. This approach offers the strongest alignment and differentiation, but it relies on robust attribution, access to data, and a longer time-to-cash. Because the value varies by customer, value-based pricing rarely fits neatly into a standard online pricing page. Instead, it is usually agreed in bespoke, results-linked agreements such as shared-upside arrangements or success-based fees.
Alongside these models sits agent or user-based pricing, which prices access to AI capability rather than the work it performs or the results it delivers. It is simple to sell, predictable, and familiar, but it struggles with usage variability and margin risk at scale. This is why, in practice, it is rarely used in isolation. Instead, it often forms the base of a hybrid pricing model, with additional pricing layers tied to usage, outputs, or outcomes to capture the value created as customers scale.
When viewed together, the pattern is clear. The closer a pricing model sits to customer value, the more directly the price reflects the outcomes achieved. But this also increases the demands on careful design, reliable measurement, and a clear understanding of how customers assess value. This is why most AI pricing strategies are hybrid by design, combining models to balance value alignment, risk, and simplicity as products and customers evolve.
How do you understand the value of AI solutions?
When teams feel uncertain about pricing AI, it is usually because they lack a clear, shared understanding of what customers value, how they define success, and what they see as realistic alternatives. In the absence of that clarity, pricing decisions tend to drift toward cost, competitor benchmarks, or cautious pricing “until we know more”.
This is exactly the problem a Pricing Sprint® is designed to solve.
A Pricing Sprint® is a structured, time-boxed way to build evidence about value and turn it into confident pricing decisions. Rather than debating pricing models in the abstract, teams focus on understanding how AI creates value in practice and how customers perceive that value.
In a Pricing Sprint®, that evidence typically comes from two places:
- Pricing Interviews uncover how AI actually shows up in customer workflows, what it replaces, and where it makes a meaningful difference. These conversations often reveal moments where AI quietly removes hours of work, reduces risk in critical processes, or unlocks outcomes that were previously hard to achieve.
- Pricing Surveys help test those insights at scale. They surface how customers see the trade-offs between different pricing models, what they see as fair, and what they compare the product against. They also reveal the real alternatives customers use, whether that is manual work, existing tools, internal teams, or doing nothing at all.
When teams invest in understanding value deliberately and revisit it as products evolve, pricing decisions become easier to defend, communicate, and refine over time. That is why value understanding is not a one-off exercise, but a core part of building a confident, sustainable AI pricing strategy.
To learn more, get in touch here or email us directly at hello@untappedpricing.co.uk.
