How to Build an AI Pricing Model Around Customer Value

AI is creating a pricing problem for software companies.

Not because the basic principles of pricing have suddenly stopped working, but because the economics underneath software have changed.

For traditional SaaS companies, marginal costs were often low enough that imperfect pricing models could survive. A company could charge per seat, bundle features into a subscription, and still maintain healthy margins even if the pricing metric was not perfectly aligned with customer value.

AI changes that equation.

Compute, inference, model usage, and other variable costs mean that every additional unit of consumption can create a meaningful cost for the provider. As a result, companies can no longer afford to treat pricing as an afterthought.

That was one of the central themes in our recent conversation with pricing expert Mark Stiving.

His point was simple: pricing fundamentals have not changed. Buyers are still exchanging money for value. What has changed is the cost structure and the way that value is delivered.

The challenge for AI companies is figuring out what they should actually charge for.

The pricing metric matters more than ever

The pricing metric is the unit that determines how a customer is charged.

In SaaS, that might historically have been users or seats.

In AI, companies are experimenting with:

• Tokens
• API calls
• Transactions
• Credits
• Resolutions
• Outputs
• Outcomes

The problem is that not every measurable unit is a good pricing metric.

Tokens are a good example.

Tokens are relatively easy to measure, and they often correlate with the underlying cost of delivering an AI service. But that does not necessarily mean they correlate with what the customer values.

Mark argues that companies should instead ask a different question:

What measurable unit increases as the value delivered to the customer increases?

The closer the pricing metric is to that value, the stronger the commercial model can become.

That distinction matters because pricing directly from cost can quickly turn into cost-plus pricing.

A provider sees that AI consumption costs money, adds a margin to that cost, and passes it on to the customer.

That may protect margin, but it does not necessarily capture willingness to pay.

The best price is still ultimately driven by the value the customer believes they are receiving.

Think in terms of activities, outputs, and outcomes

One useful way to think about AI pricing is to look at how closely the pricing metric connects to the customer result.

At one end are activities.

These are things the system does, such as processing tokens or making API calls. They are measurable, but they may say little about the business value created.

A step closer to value is an output.

An AI system might generate a lead, complete a task, analyze a document, or conduct a customer interaction.

At the other end are outcomes.

An AI support agent might successfully resolve a customer issue. A system might recover lost revenue. Another application could reduce the amount of manual work required to complete a business process.

Outcomes are attractive because they can align pricing directly with value.

But they introduce another challenge: attribution.

If revenue increased, costs fell, or productivity improved, can the vendor demonstrate that its technology caused that result?

That is why outcome-based pricing remains difficult outside use cases where outcomes are clearly measurable and attributable.

For many companies, the answer will therefore be somewhere between activities and outcomes.

The goal is to move as close to measurable value as the business model allows.

Why hybrid pricing is gaining traction

AI companies also face another tension.

Customers want predictable bills.

Vendors want predictable revenue.

But AI usage is inherently variable.

Pure consumption pricing can reflect actual usage, but it can make revenue difficult to forecast. Pure subscription pricing creates predictability, but it can leave a provider exposed if customer usage and infrastructure costs increase significantly.

That is one reason hybrid pricing is becoming increasingly attractive.

A hybrid model can combine a recurring subscription with an included level of usage, followed by additional consumption charges above that threshold.

This creates a commercial model that can provide predictability while still allowing revenue to scale with consumption.

It can also create a fairer relationship between vendor and customer.

Instead of asking a customer to pay today for the scale they may reach several years from now, pricing can grow as usage and value grow.

The customer pays more when they are receiving more value, while the provider captures more revenue as consumption increases.

Credits can create flexibility, but also complexity

Credits have become another popular approach to AI monetization.

Rather than pricing every capability directly in dollars, the provider sells a pool of credits that customers can use across different products or services.

Mark describes credits as a form of vendor-issued currency.

That analogy is useful because it explains both their value and their limitations.

Credits provide flexibility.

A vendor can introduce a new feature, assign it a credit value, and make it available without rebuilding the entire commercial agreement.

They can also allow customers to consume different products from the same pool rather than committing to rigid volumes in advance.

But credits also create another layer of abstraction.

Customers must understand:

• What each capability costs in credits
• How quickly credits will be consumed
• Whether credits expire
• Whether unused credits roll over
• Who within the organization can consume them
• How those credits ultimately translate into business value

This is why credits may be particularly useful while AI pricing is still evolving.

They give vendors room to experiment.

However, as companies better understand how customers derive value from their products, more direct pricing models may eventually become preferable.

AI monetization requires commercial flexibility

The broader implication is that companies should avoid designing their monetization infrastructure around a single pricing model.

The winning metric today may not be the winning metric two years from now.

A company may start with tokens, move toward credits, introduce a hybrid subscription model, and eventually price around outputs or outcomes.

That creates an operational challenge.

Every pricing experiment still needs to be translated into a working commercial model.

Companies need to capture usage, apply pricing rules, rate consumption, manage commitments and thresholds, and turn all of that activity into accurate revenue.

The strategic question is therefore not simply:

What should we charge for?

It is also:

Can our monetization infrastructure support the pricing model we want to test next?

That flexibility becomes increasingly important as AI companies continue experimenting with how best to connect usage, cost, and customer value.

The market may eventually become simpler

Despite the current complexity around tokens, credits, hybrid pricing, and outcomes, Mark believes AI pricing will ultimately become simpler.

The market is still learning.

Vendors are learning which metrics customers understand.

Customers are learning how to evaluate AI value.

And both sides are learning how AI consumption translates into real economic outcomes.

Over time, the strongest pricing models should become clearer.

Until then, one principle remains constant:

The pricing metric should get as close as possible to the value the customer receives.

AI may have changed the economics of software, but it has not changed what customers ultimately pay for.

They pay for value.

Watch the full conversation

In the latest LogiSense podcast, Mark Stiving joins Tim Neil to explore AI pricing, tokens, credits, hybrid models, outcome-based pricing, and why selecting the right pricing metric has become one of the most important monetization decisions for AI companies.

Watch the full conversation to learn more about how companies can navigate the transition from AI consumption to customer value.