Talking Machine #4: How AI Is Attacking the Consulting Business

The disruption everyone expected was production. AI collapses the cost of assembling knowledge work; reports, analyses, literature reviews, and frameworks. to name a few. Work that used to take teams and months now takes one person and a few days. And yes, the correct response is true: we'll move upmarket into judgment. Let the machines handle production.

That response is half right. If you want to read more about the halfs and the quadrants of this, check ou the main diagram in my article, What AI Actually Breaks When You Sell Thinking for a Living.

There's a second disruption running at the same time, and it's harder to see because it doesn't look like disruption. It looks like competence.

AI doesn't just produce assembly work faster. It produces something that looks like judgment. A well-structured strategic recommendation. A polished analysis. A framework-driven assessment that reads like it came from someone who spent twenty years in the industry.

Reads like it. Yes.

The quality is often different. Sometimes enormously so. But the difference can be invisible at the moment of purchase. A client reading the deliverable can't tell whether it took thirty years of experience or thirty seconds of prompting. They may find out later — when the strategy fails in a way the framework didn't predict, when the recommendation misses something that only lived experience would have caught. By then the engagement is over.

So the "climb the hill into judgment" strategy has a hole in it. You moved to higher ground, and the water followed you there. AI doesn't have judgment ... but because it counterfeits the surface of judgment well enough a buyer may not notice.

I call this the counterfeit ceiling. The floor of your business, the production work that you billed for, is collapsing underneath you. The ceiling. the judgment that was supposed to be safe ground — is being counterfeited from above. Two collapses, squeezing from both directions.

I am still working on my metaphors, fo course.

What buyers are actually buying

Most consulting engagements bundle three things together, and AI is coming for each one differently.

The buyer who needs hands and hours — who knows what needs doing but doesn't have the people — is buying organizational capacity. The production layer. AI has already cratered it. If the work is findable, synthesizable, structurable, the client doesn't need a human for it anymore, and increasingly they know it.

The buyer who needs something that reads like rigorous analysis — a deliverable that satisfies a board, a stakeholder group, an oversight body — is buying the appearance of sense-making. AI kills this fast, because producing confident, well-organized language that performs the surface of thinking is exactly what language models do. The counterfeit is indistinguishable from the real thing at the moment it matters most: when the audience is reading the memo and deciding whether to be reassured.

There's a variant of this second one worth naming, because it feels different but it isn't. The buyer who hires you because they don't know how to ask the AI the right questions; it is a skill to know how to frame the problem, what to even look for. It feels like they're buying expertise. Look closer and it's the same thing, buying the appearance of a rigorous process. They just need help framing the questions. And the models get better at handling poorly-formed questions every few months.

The third buyer hires you because you know things they can't get anywhere else — the situated judgment, the hard-won domain knowledge, the ability to read a room and see what's not in the data. That's genuine expertise. It's the most durable of the three. But it's durable only if you're maintaining the expertise advantage, and only if the buyer can tell the difference between what you're offering and the articulate imitation they can get for free.

Most firms have been selling all three bundled together at a blended price. A lot of that revenue was the first two. When it evaporates...and it is evaporating...what's left is the third thing.

The expertise paradox

Here's the cruel part. If you're an expert who has published extensively, spoken at conferences, built a body of work over decades...chances are that that body of work is in the LLM training data. The models have absorbed your frameworks, your vocabulary, your recommendations. They redistribute a degraded version of your expertise to anyone who asks.

The model will even cite you as a source. After it already gave the person a version of what you would have said.

This is happening. Your expertise that was supposed to be the safe harbor for you is the same expertise being counterfeited.

The people still hiring experts are the ones who've already been to the LLM. They got the framework. They read the summary. What they need now is the thing the model can't provide: the judgment that comes from actually being inside the problem, from watching recommendations play out over years, from knowing that the clean answer in the slide deck is going to hit a wall the data didn't show.

That's a real market. But it's a different market than the one that existed three years ago. Deeper, more embedded, more relational.

The burden has moved

In the old world, the buyer had to justify hiring you. They had a problem, they needed help, they found someone with credentials. The default was to bring in an expert.

Now the buyer has to justify hiring you instead of asking the model. The model is free, instant, and produces something that looks competent. You are expensive, slower, and the difference in quality is invisible until it matters.

You can't just assert that your judgment is better. You have to make the difference visible. Provenance on your reasoning — how you got there, what you considered and rejected. Track records that close the loop: "We recommended X and here is what happened." Relationships deep enough that the client trusts the judgment because they trust the person.

I have more of these, but, well, AI is watching and you need to hire me to learn them. :-D

If the buyer can't see the difference between your judgment and the counterfeit, the counterfeit wins on price every time.

The mirror

This cuts both ways. Your client can't always tell real judgment from articulate imitation. But here's the thing — neither can your organization.

Are your senior people still thinking? The kind of thinking that comes from deep domain immersion and the willingness to say unpopular things when the data points somewhere uncomfortable. And which of them rose because they managed the assembly process well, not because their thinking was better than the assembly itself?

The restructuring that's coming won't follow the org chart.

The storm

I don't think anyone knows how this settles. I certainly don't. We're in the worst part of it, the period where genuine expertise and articulate counterfeits are standing in the same line and the buyer is squinting.

The question that won't wait is whether you can make your judgment visible before the market decides the counterfeit is good enough.

I work with leadership teams at agencies and project-driven organizations on exactly these kinds of structural challenges — the ones that make growth feel harder than it should. If what I'm describing here sounds familiar, I'd enjoy hearing about it.