Artificial intelligence will not end consulting. It will end the part of it that was already a commodity: collecting information, arranging it into slides, and handing back a well-formatted synthesis.
That part, which sustained comfortable margins for decades, now costs about as much as a monthly subscription. What remains is the question the client could never quite formulate alone, and that part became more valuable, not less.
What exactly got automated
It's worth being precise, because the panic tends to be generic while the reality is quite localized.
Automated: market scans, competitor benchmarks, interview summarization, first-pass financial models, report drafting, formatting. What used to take three analysts six weeks now takes one analyst three days.
Not automated: deciding which question deserves an answer. Noticing that the problem the client presented isn't the real problem. Holding an unpopular recommendation in front of a divided board. Knowing, because you've seen it before, that the attractive number on slide 34 rests on an assumption nobody tested.
The layer that was always the product
Good consulting was never about information. It was about someone from outside, with enough context and no internal political dependency, saying what the insiders already knew but couldn't say.
The value was never in the report. It was in the conversation the report made possible.
AI makes the report nearly free. That should be good news for anyone who always sold the conversation, and it's very bad news for anyone who sold the weight of the document.
The new risk: speed without judgment
There's a less discussed side effect. As analysis production gets cheap, the volume of analysis explodes. Executive committees that used to receive one study per quarter now receive five per month, all plausible, all well written, all with the appearance of rigor.
Plausibility is not evidence. A well-constructed text built on a wrong assumption is more dangerous than a poorly constructed one, because it slips through review with less friction.
The practical consequence: the bottleneck in executive decision-making is no longer a shortage of analysis but the capacity to discard analysis. That is judgment work, and judgment doesn't scale with a software license.
What changes for buyers
- Stop paying for volume. If the deliverable is measured in pages, you're buying something the machine does better and cheaper.
- Buy access, not documents. What justifies the price is an experienced person's time spent on your problem, not the final artifact.
- Demand the explicit assumption. Every recommendation should arrive with "this only holds if…". Without it, you're not buying analysis, you're buying blind trust.
- Ask who signs. If nobody puts their name under the recommendation, nobody is carrying its risk.
What changes for providers
The consultancy that survives will be smaller, more senior, and more expensive per hour, and cheaper per project, because it will deliver in three weeks what used to take three months. The pyramid model, with many juniors running research, loses its economic base.
This isn't the end of an industry. It's the end of a pricing model that confused effort with value. AI just made that confusion impossible to sustain.
Sharing is welcome with credit and a link back to the original source.