If AI does it, why does it cost so much? How to talk to clients about estimates

This question comes up more and more often in meetings. Sometimes directly, sometimes in a sneakier version: “I heard this can be done much faster now, so shouldn't the quote be lower?”

If you are a PM at a software house or an agency, you have probably heard it already. If not, you probably will soon.

The worst possible reactions are defensiveness (“it's not that simple!”) and panic (“okay, we'll lower it”). Both come from the same mistake: we treat the client's question as an attack, when it is really a legitimate business question.

Why is the client asking in the first place?

The client sees a demo where AI generates a website in 30 seconds. They read headlines about agents that “write code 10 times faster”. Their logic is simple: if the work takes less time, it should cost less.

That logic is partly right. AI really can speed up some stages: boilerplate, tests, documentation and so on. Pretending otherwise will destroy your credibility faster than any high quote.

So the first rule is: do not deny it. Start by admitting that the client is right where they are right.

What does AI actually speed up, and what does it not?

Before you sit down for the conversation, break the project into stages and honestly assess the impact of AI.

Where AI gives a big gain:

• application skeletons and repetitive code,

• unit tests and technical documentation,

• proofs of concept,

• migrations and refactorings with clear rules.

Where the gain is small or zero:

• understanding what the business really needs,

• architectural decisions,

• integrations with existing, undocumented systems,

• agreements, approvals, changes in requirements,

• tests with real users,

• responsibility for quality and security.

Notice that the second list contains the things that usually decide whether a project succeeds. Writing code is rarely the bottleneck. More often it is unclear requirements or delayed decisions on the client's side.

What the client is really paying for

This is the heart of the whole conversation. The client pays for a result, not for hours spent at a keyboard.

When someone pays a plumber, they do not ask how many minutes it took to tighten a wrench. They pay for the fact that after the visit the water does not leak, and that the plumber knows where to look for the cause and answers for the repair.

In IT projects we pay for:

• knowing what to build (and what not to build),

• reduced risk: an experienced team knows the traps that others fall into,

• responsibility: someone answers for quality, deadline and security,

• maintenance: code has to be developed and fixed after the launch as well,

• coordination: making sure everything comes together in one place.

How to run the conversation

1. Confirm that the question is legitimate.

“Good question. AI really is changing the way we work, and I want to show where that shows up in the quote.”

2. Show where AI has already lowered costs.

Preferably in numbers: “On this project, tests and documentation will take us about 40% less time than two years ago, and that is already included.” If you have included nothing, that is a sign the quote needs a review.

3. Explain where time does not shrink.

Use an example from the client's project rather than generalities: “The biggest risk here is the integration with your system. AI will not solve that for us, because it requires talking to people and understanding how that system really works.”

4. Move the conversation from time to value and risk.

“Instead of asking how many hours it will take, let's ask what happens if this module fails on launch day.”

5. Offer options instead of a single price.

For example a basic variant, an extended one, and one with additional quality safeguards. You give the client a sense of control, and the conversation about price turns into a conversation about scope.

Red flags: what to avoid

• Hiding the use of AI. If the client later finds out that you used AI and stayed silent about it, you will lose their trust. It is better to talk about it openly, including the rules for handling the client's data.

• Padding the quote “just in case” and hoping the client will not notice. The market learns fast, and clients compare offers.

• Repeating “it's complicated” without specifics. It sounds like an excuse, even when it is true.

• Promising unrealistic savings to win a tender. It ends in cutting quality or a burned-out team.

• Lowering the price without changing the scope. It signals that the original quote was dishonest.

An example conversation in a nutshell:

Client: “I heard AI does this in a few hours. Why is the quote 3 weeks?”

Weak answer: “It's not that simple, AI won't replace developers.”

Good answer: “You're right that AI speeds up part of the work. For your project we have already priced in a lower cost for the prototype and the tests. The three weeks are mostly the integration with your payment system and agreeing the business rules with your team. AI will not skip over those, and they decide whether everything works on launch day. If you like, I can break the quote into stages so you can see where the work is and how much of it there is.”

Conclusion

The question “why does it cost so much if you use AI?” is not a threat. It is an invitation to an honest conversation about what the client is really paying for.

Client trust is built not by defending the quote, but by making it transparent.