When a Client Says AI Could Do This Cheaper

When a client says AI could do this cheaper, the worst move is to answer the sentence. It is four different objections wearing the same words, and the reply that works for one makes you look defensive in the other three. Spend thirty seconds finding out which you are facing, then answer that.

Underneath all four is the same real question, which is almost never about price: who is responsible when the output is wrong?

Four things a client says ai could do this cheaper actually means

The diagnostic question is short: "What did you get when you tried it?" If they tried it, you are in row one or three. If they did not, you are in row two or four. Either way you now know something they did not volunteer.

Row one: they have a draft and think it is a deliverable

This is the most common and the easiest, because you can settle it with evidence rather than argument. Ask them to send you what the tool produced. Then do not critique it in the abstract, which reads as self-interest. Instead find the two or three specific things in it that would have cost them money.

For copy: a claim about their own product that is not true. For a contract: a clause that does not exist in their jurisdiction. For code: the thing that works in the happy path and loses data on the second one. For a design: the layout that collapses on a phone.

Three concrete defects beat any amount of explanation about craft, and you are not arguing that AI is bad. You are pointing out that the output was 80 percent finished and the missing 20 percent is the part that carries the liability. Most clients have had an experience that makes this land immediately.

Row two: they have a cheaper quote

Do not match it. Matching teaches them your price was soft and makes the next negotiation worse. Re-scope instead, which keeps your rate intact while giving them a real choice.

Offer the same rate for a smaller piece of work: the part where being wrong is expensive, done properly, and the rest left to them or to the cheaper option. "I will do the data model and the payment flow at my rate. The marketing pages you can genuinely generate, and I will review them for two hours." You have conceded nothing on price and handed them a defensible plan.

Two things happen. Sometimes they take it, and you get a smaller project with a better margin and a client who learns what you are for. Sometimes they go with the cheap quote, and roughly a third of those come back when it breaks, now understanding the difference. Both outcomes are better than discounting.

The mechanics of pricing by scope rather than by hour are covered in how to price freelance work when AI makes you faster.

Row three: they want your efficiency gain

The honest answer is that your costs probably did fall, and that this is not the client's saving to claim. It is the return on your having learned the tools, and on your carrying the risk of using them. Said that baldly it sounds grabby, so say it as a comparison instead.

A builder who is faster is not selling hours any more. They are selling a finished thing by a date, with someone accountable for it. That is why this row is fatal if you bill hourly and straightforward if you bill by deliverable: an hourly invoice invites exactly this conversation every time you get better at your job.

If you are still on hours, this objection is the signal to change, not to defend. Fixed scope with payment milestones removes the premise.

Row four: they do not think the category should exist

Occasionally the objection is sincere and total: they believe the work is now free and anyone charging for it is arbitraging their ignorance. You will not change this with a conversation, and trying costs you two weeks of unpaid pitching.

Name it kindly and leave the door open. "It sounds like you want to try this in-house first. That is reasonable. If you want someone to review what you end up with, I charge for a two-hour review and it is the cheapest thing I do." You have been useful, you have not been free, and you have priced a small door they can walk through later. A related and harder version of this conversation, where the client wants AI to replace people rather than vendors, is covered in when a client wants AI to replace their staff.

The sentence that works in all four rows

Whatever the row, the conversation improves the moment you move it from cost to responsibility. A version worth having ready:

"You can absolutely get a first draft of this for nothing. What you are paying me for is that when it is wrong, it is my problem and I fix it. If you are happy to own that yourself, the cheap route is genuinely the right call."

Three things make this work. It concedes the true part, so you are not visibly defending territory. It names the thing they are actually buying, which is transfer of risk. And it genuinely gives them permission to say no, which is what makes it persuasive rather than pushy.

It also has the advantage of being true, and of being increasingly enforceable. Where a model's output affects someone's health, legal rights, finances or livelihood, the major vendors now require a qualified human who can review and change the recommendation: Anthropic's usage policy states that requirement directly. For a client in a regulated area, the human in the loop is not your sales pitch, it is a condition of using the tool at all.

What not to say

• Do not disparage the tools. You almost certainly use them. Getting caught on that is worse than losing the job.

• Do not quote a vague multiplier. "AI gets you 70 percent of the way" invites a 70 percent discount request. Use specific defects in their specific output instead.

• Do not drop your rate and keep the scope. It is the only move that damages every future negotiation as well as this one.

• Do not oversell your own AI use as the justification. "I use AI too, so I am worth more" is an argument they can run themselves.

Whether to raise your own AI use at all has a clearer answer than people expect, and it is worked through in should I tell clients I use AI. The wider question of what you are actually selling sits in AI monetization strategies.

Frequently asked questions

What if they are right and the tool really does the whole job?

Then say so and charge for the part that remains, which is usually judgment, integration and accountability. A reputation for telling clients when they do not need you is worth more over five years than one project.

Should I show them a side-by-side comparison?

Only using their own attempt, never a strawman you generated to lose. A comparison the client can tell was rigged costs you the room.

How do I stop this coming up at all?

Quote outcomes with a scope boundary and a named owner for failure, and do it in writing before price is discussed. Most of these conversations are caused by a quote that looked like a time estimate.

Is it worth lowering my rate to keep a long-standing client?

Reducing scope to fit a smaller budget keeps the relationship and the rate. Reducing the rate keeps the relationship and resets your price permanently, including for the work you do for them next year.