Your clients are using AI to write the first draft themselves

Picture the night before your next client meeting. Your client opens a chat window and prompts a first draft of the thing they used to pay you to start. I don’t read clients using AI this way as a threat. They still come to the meeting and they still want your help. They just don’t need you for the blank page anymore, and that blank page used to be part of the invoice.

Canadian data matches what your inbox shows. Statistics Canada found 32.4% of professional, scientific and technical services firms used AI in the year to spring 2026, one of the three highest rates of any industry. If you run an engineering or accounting practice in Calgary, that national figure is your sector, and your clients have the same tools open at their own desks.

You can check the pattern against your own inbox. Read the last few things clients sent you ahead of a first meeting. More of them now open with work already done: a draft, an outline, three options they made themselves and want tested. The request has changed shape. It used to read, write this for me. It now reads, tell me if this is right, and fix what is wrong with it. Same client, same trust in you, smaller piece of the job handed over cold.

What are clients actually prompting?

Anything with a standard shape. First drafts of letters, proposals, and reports. Summaries of long documents they already hold. Routine marketing copy. A first pass over their own numbers. Plans that follow a pattern they have seen before.

The common thread is plain. Where the form is familiar and the standard is forgiving, a client can produce a passable first version tonight, at their own desk, for the cost of a subscription they already pay for.

If the first draft has a standard shape, your client can now make it without you.

One more change rides along with it. A client holding a draft walks in with a position already formed. Your first job in the meeting is no longer setting the starting point. It is moving a starting point the client built without you, and that calls for a different kind of authority than drafting ever did.

What can a prompt not sign for?

Judgment, review, and a name that answers for the work. Judgment is knowing which of three decent drafts fits this client, this history, this level of risk.

Review is reading the draft against the real situation and catching what a fluent paragraph hides: the figure that looks right and is wrong, the clause that sounds standard and is not, the advice that worked somewhere else and fails here. Accountability is the plainest of the three. When the answer matters, somebody has to stand behind it, take the call when it goes sideways, and carry the cost of being wrong. A chat window carries none of that, and your clients know it, which is why they still come to the meeting.

A prompt can produce the answer. It cannot stand behind it.

Where did the value in the draft go?

The draft got cheap, and the value moved downstream into everything that happens after a draft exists. Your client can now produce average work alone in an evening. If the thing you sell is average work on letterhead, the client can see that gap now, and some of them have already priced it. My view is that this is good news for a lean practice, because the stretch after the draft is where a small firm can now take on bigger work.

The firms keeping the work are selling the stretch after the draft: the checking, the fitting to the actual situation, the signature at the bottom. That is the case for pricing the finished result instead of the hour. That is why I price the finished job and not the first pass. A finished job is reviewed, corrected, and signed for. Clients still pay for finished.

Nobody pays a premium for a first draft they can make in a chat window.

Which five deliverables do you test first?

The five your clients buy most that follow a standard shape. Sit down and write them out: the standard proposal, the routine report, the summary, the first-pass plan, the regular copy you produce on a cycle. For each one, make one decision.

Repackage it around the judgment, the review, and the name on it, and describe it that way when you sell it. Or stop selling it as a standalone piece, because unchanged, it now competes with a tool your client already has open. There is no third answer that survives the year. A deliverable you leave exactly as it was is a deliverable you have decided to discount slowly. It’s the same logic as deciding what to build and what to buy. The draft is the rented part, and your review is the part you own.

Be honest about which side each deliverable lands on. The ones built on what you know about a specific client were never really draft-shaped, and they are safe. The ones any competent stranger could assemble from public material were priced on drafting time being scarce, and that scarcity is gone.

How do you repackage your offer this week?

Take one hour and run the test on your own offer. List the five deliverables, make the call on each one, and pick the first you will repackage. Rewrite how you describe that one, so the description leads with the review, the judgment, and the accountability, and treats the draft as the free starting point the client already holds.

Then watch the next meeting change. Deciding which parts of an offer still carry a price is the core of an AI Strategy Roadmap.

Clients using AI, quick answers

Which of your five deliverables would you still sell standalone if every client drafted the first version themselves? Answer that before your next meeting, and I’ll keep working through calls like it in the AI Newsletter with ZAK.