A client said AI gave the same advice I would. Should I be worried?

It means the advice in question lives on the consensus layer: the part of your field that has been written about enough for a model to absorb. Language models are trained on the published thinking of every discipline, so for well-documented situations they reproduce what good practitioners agree on. A competent expert and a competent model matching on the basics is the system working.

What the match does not test is everything past the baseline:

• Fit. Whether the standard advice survives contact with this client's actual constraints, politics, and history.
• Sequencing. What to do first, what to skip, when to break the standard rule, the ordering judgment that consensus text rarely captures.
• The undisclosed variable. Clients tell AI what they think matters; an experienced advisor hears what they left out.
• Accountability: someone responsible for being wrong.

The research boundary is measurable: in the Harvard and BCG field experiment, professionals using AI on a task outside its competence were 19 percentage points less likely to get it right. The consensus engine is excellent until the case stops being average, and paying clients are rarely average cases.

With visible comfort, because your reaction is the actual test being run. A client mentioning the AI comparison is, consciously or not, watching how you handle it, and defensiveness confirms the fear while confidence dissolves it.

The response that works has three beats:

1. Agree, specifically. 'Good, that means the fundamentals are sound. That is exactly what the general answer looks like.' You just repositioned AI as your baseline check rather than your rival.
2. Then differentiate, concretely. 'What it can't see is that your situation breaks the general case in two places,' and name them. This is the move only you can make, and making it in real time is the demonstration no marketing can buy.
3. Invite the habit. 'Keep asking it questions, and bring me the answers that don't sit right.' You have now converted their AI usage from a threat into an intake channel that surfaces exactly the questions worth your time.

What kills the moment: arguing that the AI is wrong when it isn't, banning the comparison, or visibly needing to win. The expert attracts, never pursues, and never argues with a free tool for territory she should be above.

Run the audit honestly, because the boundary is real on both sides.

What AI reproduces well:

• Framework recitation: the models, methods, and mental tools your field has published.
• Standard-situation advice: the right first move for the textbook version of a problem.
• Explanations and education: patient, clear, infinitely repeatable answers to how-does-this-work questions.
• Option generation: a competent list of possible approaches with generic pros and cons.

What it cannot reproduce:

1. Applied judgment: which option fits this client, given the constraints they mentioned and the ones they didn't.
2. Earned pattern recognition: the sense that this situation rhymes with a case from years ago and will fail the same way.
3. Productive confrontation: telling a client the problem they brought is not the problem they have.
4. Stakes-bearing: recommendations someone stands behind when they're wrong.

The uncomfortable part of the audit is quantitative: if most of your delivered hours live in the first list, the client's comparison was a preview. The strategic response is moving your time, and your pricing, decisively into the second list, and letting AI have the first.

By showing the reasoning, not just the recommendation, because the recommendation is where the overlap lives and the reasoning is where you don't have competition.

Four habits that make the judgment layer visible:

1. Attach the why to every prescription. 'Do X, because in your situation Y is true and the usual advice assumes it isn't.' Generic advice arrives reason-free; reasoned advice is unmistakably yours.
2. Name the trade-offs out loud. Consensus answers optimize for defensibility and hedge everything. An expert who says 'this costs you A to get B, and for you B is worth it' is doing something no engine does.
3. Disagree with the standard answer when it deserves it, and say so explicitly: 'the textbook move here is X; for you I'd break that rule, and here's why.' Every earned exception is a demonstration.
4. Put your positions in public. Document the calls you make differently than your field, so buyers meet your judgment before they ever compare you to a chat window.

The pattern underneath: AI converges on the average answer by design, and research shows AI-assisted work measurably drifting toward sameness. Visible, specific, owned reasoning is the signature the average can never carry.

When the honest audit says your business is mostly information transfer, because that layer is not coming back. The moment to take seriously is not a client's one-off comparison; it is a pattern:

• Your engagements front-load explanation. If the first month of working with you is mostly education a model could deliver, clients will increasingly arrive pre-educated and question the fee.
• Your questions get answered before they reach you. The informational queries that used to fill sessions are going to the tools first, and what remains for you is thinner than your pricing assumes.
• Your differentiation is fluency, not position. If what made you impressive was commanding the material rather than judging with it, fluency just became free.

The response to a worrying audit is reconstruction, not panic: capture your method so the informational layer can be delivered by systems, move your live time into judgment, accountability, and the hard conversations, and reprice around the layer that survived. The experts who do this early report the strange outcome that the comparison moment stops happening, because clients can no longer confuse what they do with what the tool does. Tracking how this shift is actually playing out for advisors, quarter by quarter, is part of what the Collective Wisdom newsletter is for.