Client Wants a Discount Because You Used AI

When a client asks for a discount because you used AI, the answer is that you sell an outcome, not hours, and the outcome did not get smaller. Say it once, plainly, and move the conversation back to what they are actually buying. If your pricing is genuinely hourly and genuinely tied to time spent, the request is fair and you have a pricing problem rather than a negotiation problem.

That second sentence is the uncomfortable one. Most of the panic around this question comes from people who quote time-based prices while privately selling value, and the AI question exposes the gap.

The three ways this arrives

It rarely comes as a blunt demand. Recognise which one you are getting, because they need different answers.

The fairness framing. "If AI did most of the work, shouldn't the price reflect that?" This is a genuine question from someone who believes price should track effort. They are not trying to squeeze you.

The leverage framing. "I could just do this with AI myself for twenty dollars a month." This is a negotiating position, sometimes an accurate one, often not.

The trust framing. "I didn't realise you were using AI." This one is not about money at all, and answering it with a pricing argument makes it worse. What they are asking is whether they got what they paid for.

Answering the third with a response designed for the first is the most common own goal in this conversation.

The answer to the fairness framing

Effort-based pricing is a convention, not a law, and it has never been how expertise is priced. Nobody asks a surveyor for a discount because the equipment is fast.

A version that holds up in a real conversation rather than reading like a script:

The price is for the result and for the fact that it works. How I get there is my problem. If I found a way to deliver the same thing in half the time, that is what you are paying me for, not a reason to pay less. The alternative is that I bill you for inefficiency, and neither of us wants that.

Then stop talking. The instinct to keep justifying is what turns a one-line answer into a negotiation.

If you want a structural fix rather than a good sentence, the answer is to stop quoting hours. Fixed-scope versus hourly pricing covers the trade properly, but the short version is that fixed-scope pricing makes this question incoherent, and hourly pricing makes it inevitable.

The answer to the leverage framing

This one deserves respect rather than defensiveness, because sometimes they are right. If the deliverable genuinely is a thing a competent non-specialist can produce with a chat interface in an afternoon, the market price for it is falling and no amount of positioning fixes that.

Assuming it is not, the honest response is to be specific about what the twenty-dollar subscription does not include:

• Knowing which of the four plausible approaches is the one that survives contact with their actual data

• The parts that are not generation at all: scoping, integration, testing, deployment, the second round when the first version meets reality

• Accountability. If it breaks in November, they call you. Nobody calls a subscription.

• The time they do not spend learning to do it

Say two of those, not all four. A list of four reads as anxiety.

There is also a version of this where the right move is to agree. If a client can genuinely self-serve the work, tell them so and offer to review the result for a smaller fee. You lose a small project and gain someone who trusts you, which is a better trade than winning a project you should not have sold. Related: should you tell clients you use AI argues that the disclosure conversation goes far better when you start it.

The answer to the trust framing

Do not discuss price. Discuss verification.

The concern behind "I didn't realise you were using AI" is that the work might be plausible-looking rubbish. That is a legitimate worry and the answer is evidence: the tests that pass, the review you did, the specific things you checked, the errors you found and fixed. Proving AI-generated work is correct to a client is the whole playbook for that conversation.

Offering money at this moment confirms the suspicion. It reads as an admission that the work is worth less, which is the opposite of what you need them to conclude.

When a discount is actually the right call

Three cases where holding the line is a mistake.

That last row matters. If you are going to concede, concede a defined thing: an extra revision round, a faster turnaround, a small add-on. A percentage discount resets your price permanently for that client. A bounded concession does not.

What to change so it stops happening

The conversation is a symptom. Three changes remove most of the recurrence.

Quote deliverables, never hours. "A working booking system with staff scheduling and payment handling, delivered in three weeks, ten thousand" invites no questions about your tooling. "Eighty hours at one twenty-five" invites exactly one.

Put the tooling clause in the contract before it is a topic. One sentence: "Contractor may use automated and AI-assisted tools at its discretion. Contractor remains responsible for all deliverables and warrants that they meet the specification." Signed up front, this ends the conversation before it starts. Writing the project proposal is where that sentence belongs.

Price the thing you are actually differentiated on. If your differentiator was typing speed, AI has a real claim on your margin. If it is judgement about what to build, it does not. Most people are in the second category and price as though they were in the first. How much to charge for an AI automation project works through the arithmetic, and the broader picture sits in our guide to making money with AI.

Frequently asked questions

Should I proactively tell clients I use AI? Yes, once, at proposal stage, in one neutral sentence. Discovered disclosure costs you far more than volunteered disclosure.

What if they ask exactly how much of the work was AI? Answer the real question, which is about quality, not percentage. "All of it went through the same review I'd apply to anything I hand over" is true and sufficient. A percentage figure is meaningless and invites haggling over it.

Does this change if I am reselling AI output more or less directly? Yes. If the deliverable is essentially raw model output with light editing, the client's instinct is correct and your position is weak. Add the layer that makes it yours, or price it as the commodity it is.

A client demanded a fifty percent discount and would not budge. Now what? Decline and offer to reduce scope by half instead. If they wanted a discount they will take the scope cut. If they wanted leverage, they will not, and you have learned something useful about the relationship for the price of one uncomfortable email.