Strategy · October 19, 2026

How to Price an AI Product When You Run a Service Business

There is no standard price for an AI product. It depends on who pays, what it replaces and what it costs to run. The factors, and five models that fit.

"What should we charge for it?" is one of the first questions owners ask once an AI product idea starts to feel real. The honest answer is that it completely depends on what the product is.

That is not a dodge. Two tools built by two firms in the same industry can deserve completely different pricing, because they sit in a different place in the client relationship. So instead of a number, this post gives you the questions that decide the number, and the main models to choose from once you have answered them.

Why there is no standard answer

Pricing advice for software usually assumes a software company: a product, a market, a subscription. A service business is different. Your product does not exist on its own. It sits next to a service your clients already pay for, and it changes that service in some way.

That relationship drives everything. A tool that makes your service better should be priced differently from one that replaces part of it, and differently again from one that brings you new clients. Get the relationship clear first. The price follows.

The five things it depends on

1. Who pays

Start with the simplest question. Is the person using the tool the same person paying for it?

Sometimes the user is your existing client, who already pays you. Sometimes it is a new kind of customer who would never hire your full service. Sometimes it is your client's team, while their finance department pays. Each of these buys differently, with different budgets and different expectations about what software costs.

2. Whether it replaces, extends or feeds your service

This is the most important question, and every product falls into one of three groups:

• It replaces part of your service. The client used to pay for your time on a task. Now the tool does it. If you price it as cheap software, you may cut your own revenue.
• It extends your service. The client still gets everything they had, plus something new. The tool makes you more valuable and harder to leave.
• It feeds your service. People who would never have called you use the tool, and some of them become clients. The tool is a channel as much as a product.

Many products do more than one of these. Be clear about which one matters most, because each points to a different model below.

3. The value of the outcome

Price against what the customer gets, not what it cost you to build. A tool that saves a client a few minutes is worth very little. One that prevents a costly mistake, speeds up a sale or unlocks a decision they could not make before can be worth a great deal.

Ask your client what the outcome is worth to them. Their answer is a better anchor than any comparison with other software prices.

4. What it costs to run

AI products are not like traditional software, where one more user costs almost nothing. Every time a customer uses an AI feature, there is a model cost, billed by usage. Light use barely registers. Heavy use by many customers is a real monthly bill that grows with exactly the thing you want to grow.

Know roughly what a typical customer's usage costs before you set a price. A flat fee that looks generous with ten users can lose money with a thousand. We cover the cost side in more detail in What vibe coding really costs.

5. How your niche buys

Every industry has habits. Some are used to monthly subscriptions. Some only buy annually, through procurement. Some expect to pay per case, per project or per transaction. Some will happily pay a premium for anything that reduces risk, and some haggle over every line.

Pricing that fights those habits makes selling harder than it needs to be. Pricing that matches them removes friction you did not have to create.

The five pricing models

Once you have answers to those questions, most products fit one of these models.

Bundled into your existing fees

The tool is part of what clients get when they work with you. No separate line on the invoice.

When it fits: The product extends your service, your clients already pay well, and the main goal is retention and differentiation. It works well when the tool makes switching to a competitor feel like a downgrade.

Watch for: It is easy to give away real value this way. If the tool is a big improvement, consider raising your service fees rather than absorbing the cost.

A separate subscription

Clients, or new customers, pay a recurring fee for access to the tool on its own.

When it fits: The product delivers ongoing value on its own, is used regularly, and could be sold to customers who do not buy your full service. This is also the model that builds recurring revenue, which matters if the value of your business matters to you. See How owning a product changes what your firm is worth.

Watch for: Subscriptions need ongoing reasons to stay subscribed. A tool used once a quarter will struggle on a monthly plan.

Usage-based

Customers pay per use: per report, per document checked, per quote generated, per case processed.

When it fits: Usage varies a lot between customers, each use has a clear value, and your running costs scale with use. It aligns your revenue with your costs, which is useful for AI-heavy products.

Watch for: Customers dislike unpredictable bills. Packs of credits or a monthly allowance with a clear overage price often feel safer to buyers than pure pay-as-you-go.

Free, as a door into the service

The tool is free. Its job is to bring in customers for your core service.

When it fits: The product feeds your service, your service is high-value, and the tool solves a smaller version of the problem you are paid to solve properly. An instant estimate, a self-assessment or a quick check can all work this way. The customer gets real value and a reason to call you for the rest.

Watch for: Free products still cost money to run and maintain. Put sensible limits on usage, and make the path from the free tool to your service obvious.

Outcome-based

You get paid when the tool delivers a specific result: a recovered amount, a closed deal, a successful application.

When it fits: The outcome is measurable, clearly caused by the product, and valuable enough that both sides are happy to share it. It is the most persuasive model for buyers, because they pay only when it works.

Watch for: You carry the risk. You need a reliable way to measure the outcome and a clear agreement about what counts.

Combining models

Many good products use more than one. A free tier that feeds the service, with a paid subscription for heavier use. A bundled version for existing clients and a separate subscription for new ones. A subscription with a usage allowance to cover AI costs.

Start simple. One model, clearly explained, that matches how your niche already buys. You can add layers once you see how customers actually use it.

Test the price like you test the product

Pricing is a hypothesis, just like the product. The best way to find the right number is to put a working prototype and a proposed price in front of real customers and watch their reaction. Hesitation tells you something. So does no hesitation at all.

That is one reason we put a prototype in front of people early. It is much easier to talk about price when the customer can click through the thing they would be paying for. We explain the approach in Why we build a prototype instead of writing a scope.

How we think about it at Veyra

We take pricing seriously because we are paid on it. Our own model is one fixed build fee, agreed after you have seen a working prototype, and then a share of what the product earns while we stay on as product lead. If the product does not earn, that part of our pay is zero. So we have every reason to help you get the pricing right.

If you have an idea and are not sure how it would make money, book a call. Working out who pays and why is part of deciding what should be built.