How do you price software that AI helped you build?

“This would normally take two weeks.”

“Okay. How long will it take with AI?”

“Maybe two days.”

“Great! So it costs one-fifth as much?”

Ah.

There’s the problem.

If you’re an independent consultant using AI to build software faster, how should you charge?

Let’s say a small business needs an application to manage employee requests.

You gather the requirements, design the workflow, and use an AI coding assistant to generate much of the application.

You review the code, fix problems, test the workflows, configure security, and deploy it.

The actual coding takes a fraction of the time it once did.

Should the client pay less?

Possibly.

But how much less?

And what exactly are they buying?

I think the mistake is treating software development as nothing more than the number of hours spent typing code.

The code is only part of the work.

Someone still needs to understand the business problem.

Someone must decide what should happen when a request is incomplete, when two people approve the same item, or when the system goes down.

Someone needs to make sure the application is secure, maintainable, and actually works.

And someone has to support it after the demo.

AI can help with many of these activities. It doesn’t make the responsibility disappear.

I’d start by separating three things.

1. 1The cost of delivery. That includes discovery, design, development, testing, deployment, tools, and the time needed to handle unexpected problems.
2. 2The value to the client. How much time could the solution save? What errors could it prevent? What service improvements might it enable?
3. 3The risk you’re taking. Are you delivering a simple internal tool or a system that handles sensitive customer information and important financial transactions?

Those shouldn’t be priced the same way.

Here’s a hypothetical example.

Suppose a client spends $3,000 per month on manual work that a new application could reduce.

That’s a potential $36,000 per year.

Not guaranteed savings. Just a starting estimate that needs validation.

Now suppose you can deliver the application for $8,000.

Would the client care whether you wrote every line manually?

Probably not.

They’d care whether it works, whether the savings are real, and whether the price makes business sense.

But don’t turn that into an excuse to charge whatever number sounds impressive.

A fair price should reflect the scope, alternatives, risk, expected value, and the market.

And here’s another complication.

AI may make development faster, but it can also make it easier to build the wrong thing faster.

You still need good judgment.

You still need architecture.

You still need testing.

In fact, if AI helped you produce the application in two hours, I’d be particularly interested in how much time you spent making sure it was correct.

For an independent practice, I’d consider three offers.

1. 1A paid assessment to define the problem.
2. 2A fixed-scope implementation to deliver the solution.
3. 3An optional support agreement to maintain and improve it.

That makes the commercial relationship easier to understand.

It also rewards you for getting better at your work rather than penalizing you for becoming faster.

Just don’t forget to price the work that happens after the AI finishes typing.

That’s usually where the interesting problems begin. 😅