Will AI Replace Consultants? A Straight Answer

Will AI replace consultants? Not as a profession, but it is already replacing a large part of what junior consultants used to do, and it is changing who gets hired at the start of a consulting career. The consultant who survives is the one whose value was never in producing the analysis, but in knowing which analysis matters, whether it is right, and what the client should do about it.

That is a straighter answer than either side of the debate usually gives. The optimists say AI is just a tool and nothing fundamental changes. The pessimists say the job is gone. The evidence supports neither.

What the best experiment shows

The most useful evidence comes from inside a consulting firm. Researchers from Harvard, MIT, Wharton and elsewhere ran a preregistered experiment with 758 consultants at Boston Consulting Group, about 7% of its individual-contributor consultants, giving some access to GPT-4 and some not.

On eighteen realistic consulting tasks that fell within what the AI could do well, consultants using it completed 12.2% more tasks, finished them 25.1% faster, and produced work rated more than 40% higher in quality. The biggest gains went to the weaker performers: those below the average improved by 43%, those above it by 17%.

Then came the part that gets quoted less often. On one task deliberately chosen to fall outside what the AI could do well, a business problem that required spotting what the numbers and the interviews really implied, consultants using AI were 19 percentage points less likely to reach the right answer than those working without it.

The researchers called this the jagged frontier. AI is excellent at some tasks and poor at others that look just as difficult, and the edge between them is not visible from the outside.

The persuasive-wrong problem

One detail from that study matters more for clients than any productivity number. On the task where AI users got the answer wrong more often, their recommendations were still rated as more persuasive and better argued than those of the consultants who got it right without AI.

That is the real risk in AI-assisted consulting. The output looks better whether or not it is correct. A client reading a polished recommendation has no easy way to tell which side of the frontier it came from, and a consultant who has stopped checking has no way either.

Which is precisely why the consultant does not disappear. Someone has to know where the frontier is on this particular problem, test the conclusion against what the client actually knows, and put their name to the recommendation. That was always the senior part of the job. It is now the whole of the job.

Where replacement is real

Replacement is happening, just not where the headlines put it. It is at the entry point.

The bottom of the consulting pyramid has always been built on work that is now inside the frontier: research, benchmarking, first-draft models, interview synthesis, slide production. Firms still need that work done. They need far fewer people to do it.

Broader labour data points the same way. Stanford’s Digital Economy Lab, using payroll data covering millions of US workers through mid-2026, found employment of 22 to 25 year olds in the most AI-exposed occupations now stands about 19% below where it would be had it kept pace with less-exposed peers, while experienced workers show no comparable gap. The adjustment is happening mainly through fewer hires, not more redundancies, and it is concentrated where AI substitutes for tasks rather than complements them.

For consulting that means the profession does not shrink evenly. It narrows at the door. The long-term risk for firms is obvious: if nobody is hired to do the junior work, nobody is trained to become the senior who checks it.

A realistic picture by level

The effect is different at each level of a firm, and it helps to be specific.

Partners and senior advisers are the least exposed and, in the near term, the most advantaged. Their work is judgment, relationships and accountability, and AI makes them faster at the preparation around it. The pressure on them is commercial: clients expect the time saved to show up in fees.

Managers are in the middle of the change. Much of their job was reviewing and shaping junior work. That job now includes reviewing AI output, which demands more scepticism, not less, because the errors are subtler and better presented.

Analysts and new graduates carry the largest change. The production work that filled their first years is shrinking, and firms hire fewer of them. Those who are hired are increasingly expected to do from the start what used to come later: check, question, explain why an answer is right and talk to the client about it.

What consultants who thrive are doing

The consultants doing well in this shift share a few habits.

• They map the frontier for their own work. They know which tasks they hand to AI with light checking and which they never do, and they revisit that line as the tools change.
• They verify, not just edit. Every figure traced to its source, every conclusion tested against what the client knows. Polish is no longer evidence of quality.
• They sell the decision, not the deck. Their value is framed as the judgment and the accountability, which AI does not provide, rather than the document, which it now largely does.
• They own their context. Their methods, client histories and frameworks are written down in a form their AI tools can use, which is what makes a small practice able to deliver like a larger one.

What it means if you hire consultants

Expect faster, cheaper analysis, and do not accept last decade’s pricing for it. But pay attention to who is checking the work. Ask your consultants how they use AI, where they will not use it, and who reviews what reaches you. The difference between an AI consultant and an embedded Head of AI matters more than ever, because an answer is getting cheaper while ownership of the decision is not.

The wider repricing of the industry is covered in management consulting in the age of AI, and the shifts facing consulting and agencies together are set out in the future of consulting and agencies.

If you are rethinking how your own practice works with AI, tell Daniel where AI already sits in your work, what you have built, and what you want the practice to become. Talk to Daniel.

Frequently Asked Questions

Will AI replace consultants?

Not as a profession, but it is replacing much of the junior analysis work consultants used to do. The consultant’s value is moving to judgment: knowing which analysis matters, whether it is right, and what the client should do. Replacement is concentrated at the entry level, where firms need fewer people to produce the same research and drafts.

How much does AI improve consultants’ productivity?

In a preregistered experiment with 758 Boston Consulting Group consultants, those using GPT-4 on tasks within the AI’s capabilities completed 12.2% more tasks, 25.1% faster, with more than 40% higher quality. On a task outside those capabilities, AI users were 19 percentage points less likely to reach the correct answer.

What is the jagged frontier in AI?

A term from the BCG experiment describing how AI is excellent at some tasks and poor at others that seem equally difficult, with no visible boundary between them. The practical consequence is that consultants must know, task by task, where AI can be trusted and where it cannot.

Is AI reducing entry-level consulting jobs?

The evidence points that way. Stanford’s Digital Economy Lab found employment of 22 to 25 year olds in the most AI-exposed occupations is about 19% below where it would otherwise be, while experienced workers show no comparable gap. The adjustment is happening mainly through fewer hires.

Can clients trust AI-assisted consulting work?

Only when someone qualified has checked it. In the BCG experiment, AI-assisted recommendations were rated more persuasive even when they were wrong. Clients should ask how AI is used, where it is not used, and who reviews the work before it reaches them.

How can consultants stay valuable as AI improves?

By mapping where AI is reliable in their own work, verifying rather than just editing its output, selling the decision and accountability rather than the document, and owning their methods and client context in a form their AI tools can use.