Contents
- 1. Where Things Stand — Five Figures Checked at the Source
- 2. The Experiment With 758 BCG Consultants — Tasks AI Speeds Up, and Tasks It Gets Wrong
- 3. How AI Changes Consulting Work
- 4. Will the Pyramid Model of Training Survive?
- 5. The Pricing Question (Hours → Outcomes)
- 6. What Doesn't Change — The Core Value
- 7. Large Firms and Boutiques
- 8. Advice for Aspiring Consultants and Client Companies
- Summary
- Corrections to the earlier version
- FAQ
"All-nighters on slide decks. Mountains of research by hand." That rite of passage for junior consultants is being shaken by AI. McKinsey rolled out its internal generative AI platform, Lilli, firmwide in July 2023, and BCG has announced that "AI- and tech-focused services" made up more than 40% of its 2025 revenue. Much of the work of researching, synthesizing and building decks is already within AI's reach.
The short version: AI dramatically speeds up consulting's "production work" (research, decks, analysis), but on tasks it is bad at, it can also lead people to wrong answers. That makes the work of deciding what to ask, how far to trust the AI's answer and how to move the client more valuable, not less. This article checks data specific to consulting — official announcements from major firms, an experiment with 758 BCG consultants, and occupational statistics from the U.S. Bureau of Labor Statistics (BLS) — against the original sources (checked September 30, 2026), and uses it to think about how the industry's structure is changing. For what AI means for work in general, see Will AI Eliminate White-Collar Jobs? and Jobs at Risk From AI; for other industries, see marketing and advertising and trading companies.
From a pyramid to lean expert teams
— Production work is automated; value moves to judgment
Routine work at the bottom goes to AI. What remains is the work "above": framing questions, interpreting and moving the client (a conceptual diagram, not actual headcount ratios).
1. Where Things Stand — Five Figures Checked at the Source
Here is "consulting is changing with AI" in numbers the firms and agencies published themselves. All were checked on the publishers' pages on September 30, 2026, except McKinsey's page, which would not load that day and was checked in an archived copy.
| Publisher and date | What was published | How to read it |
|---|---|---|
| BCG (announced April 23, 2026; full year 2025) | Revenue up 7% to $14.4 billion from $13.5 billion. AI- and tech-focused services are more than 40% of revenue, and AI services grew 25% year over year. Workforce grew to 33,500, with nearly 4,000 employees developing and scaling AI workflows | AI isn't only shrinking consulting work — it has become a major product in its own right. Headcount is growing too |
| McKinsey (its own page; checked in a June 2026 archived copy) | Internal generative AI platform Lilli rolled out firmwide in July 2023. It brings together 40-plus internal knowledge sources; 72% of the firm is active on it, and colleagues report up to 30% time savings in searching and synthesizing knowledge | This is the firm's own account, not a third-party measurement. "Up to" 30% is not an average |
| KPMG (announced July 11, 2023) | A multibillion-dollar commitment to Microsoft cloud and AI services over five years, which KPMG says could unlock more than $12 billion in incremental growth opportunity | Big-firm AI investment targets both internal efficiency and "helping clients adopt AI" |
| Organization Science (March 2026 issue) | An experiment run with BCG in which 758 people worked on realistic consulting tasks. Details in Section 2 | One of the few peer-reviewed experiments measuring AI's effect on consulting work |
| U.S. Bureau of Labor Statistics (2025 data; 2025–35 projections) | "Management analysts" (the occupation that includes consultants): 1,077,100 jobs. Projected to grow 10% by 2035 (much faster than average), with about 94,100 openings a year. Median pay $101,860 a year | BLS explains that demand for consulting is expected to rise as organizations seek to improve efficiency and control costs. It does not project the occupation shrinking |
The BLS figures come with conditions. The occupation includes not only consulting-firm employees but also people who analyze their own organization and people in government (by BLS's count, the largest employers are professional, scientific and technical services at 33%, government at 18%, self-employed at 14% and finance and insurance at 13%). It covers the U.S. only, and projections can't fully anticipate future technology. Still, it doesn't fit the simple view that "AI will wipe out consulting jobs." The total number of jobs growing and the content of the work (who does what) changing sharply can both be true — and the rest of this article is about that content.
2. The Experiment With 758 BCG Consultants — Tasks AI Speeds Up, and Tasks It Gets Wrong
The most widely cited study of AI in consulting work is an experiment by researchers from Harvard Business School and elsewhere, run together with BCG. It was published as "Navigating the Jagged Technological Frontier" in the March 2026 issue of Organization Science (vol. 37, no. 2). Here are the conditions and results from the abstract.
Tasks inside AI's frontier (18)
People using AI completed 12.2% more tasks, finished them 25.1% faster and delivered significantly higher quality than those without AI — on every task, from creative to analytical.
A task outside AI's frontier (1)
On a complex business problem chosen to be beyond AI's capabilities, people using AI were 19% less likely to produce a correct answer. They were apparently pulled along by the AI's plausible-sounding answers.
Conditions: a preregistered experiment with 758 participants (BCG consultants). After a baseline on similar tasks, they were randomly assigned to "no AI," "GPT-4," or "GPT-4 plus a prompt-engineering overview." These results are for the model of that time; today's AI would not necessarily produce the same numbers.
The "jagged technological frontier" the paper names has practical meaning for consultants. Tasks AI does well and tasks it does badly are mixed together in the same workflow, at what looks like the same level of difficulty. You can't tell where the boundary is just by looking at a task. So the more you use AI to speed up the work, the more valuable a person's eye for whether an answer falls inside or outside AI's frontier becomes. That is the basis for the "core value" in Section 6.
3. How AI Changes Consulting Work
Consulting covers a lot, but AI's impact is greatest in the hands-on stages — which is also mostly where AI did well in the Section 2 experiment.
① Research and information gathering
Groundwork on markets, competitors and industries; searching and summarizing a firm's past materials. Internal platforms like McKinsey's Lilli handle this stage.
② Decks and slides
Drafts generated automatically, from storyline to design. The methods in making slides with AI apply directly.
③ Data analysis and models
Faster data analysis, financial models and scenarios. But as Section 2 shows, it can also introduce errors on tasks that require complex judgment.
⑤ New services
"AI strategy and implementation" has itself become a major earner. At BCG, AI- and tech-focused services passed 40% of revenue.
Number ⑤ is the one to watch. AI doesn't just replace consulting work — it has created a large market in helping clients use AI. BCG grew its AI services 25% year over year in 2025 and increased its workforce to 33,500. At least judging by the big firms' official announcements, AI has not so far reduced firms' total headcount. What is changing is who does which work.
4. Will the Pyramid Model of Training Survive?
The consulting industry has long been built on a "pyramid": large numbers of juniors doing research and slide work under a few senior leaders. Items ① to ④ in Section 3 are exactly the juniors' work, and if AI speeds them up, the premise of "hiring in bulk to train through grunt work" wobbles.
However, we could not find any case of a major firm publishing figures on how much it has cut junior hiring. What we could confirm is that BCG's headcount is growing and that U.S. management analyst jobs are projected to grow through 2035 (Section 1). So rather than presenting the "collapse" of the pyramid as fact, this section frames it as the questions the industry is facing.
⚠ Questions raised when the pyramid wobbles
As the Section 2 experiment suggests, the ability to catch AI's mistakes grows out of having done the work yourself. The more grunt work goes to AI, the more firms have to think about how people will gain that experience.
This overlaps with the question of whether veterans or juniors are more at risk. Not only in consulting, "how do we develop people once the grunt work moves to AI" is homework shared by many industries.
5. The Pricing Question (Hours → Outcomes)
After structure comes "how firms get paid." Consulting's traditional model, billable hours, fits poorly with AI in one respect.
🔄 The productivity paradox
The faster AI gets the work done, the less a firm can bill under hourly contracts. So as long as billable hours remain, firms have little incentive to use AI to the full — a structure in which the benefits of efficiency struggle to reach either the firm or the client.
The direction usually discussed for resolving this is outcome-based and fixed-price contracts. If fees depend on "what change was produced" rather than "how many hours were worked," finishing faster with AI benefits the firm too. However, we found no figure traceable to its publisher showing how far the industry as a whole has shifted, so this section stays at the level of direction. How client companies can think about contract types is covered in Section 8.
6. What Doesn't Change — The Core Value
So far we've looked at what changes. What really matters is what doesn't. The Section 2 experiment showed that tasks where relying on AI leads to mistakes are mixed into the same work.
AI is good at (easy to hand off)
- Research, summaries, first-draft materials
- First cuts of analysis and models
- Generating routine documents
- Listing and organizing options
Stays human (can't fully hand off)
- Framing the question: seeing the real problem
- Judging AI's answers: doubting answers outside its frontier
- Trust: the interpersonal skill to move the client
- Execution and change: moving the organization to results
This is the heart of it. AI produces "plausible analysis" quickly and in volume. But "what should we ask," "can we trust this answer," and "how do we move people and organizations" remain human work — and in the Section 2 experiment, skipping that judgment lowered accuracy. Deep industry knowledge, a track record of driving change on the ground, and an understanding of context — people with this "steering" ability become more valuable in the AI era. The line drawn in what AI can and cannot do applies to consulting as it is.
7. Large Firms and Boutiques
AI also affects the balance of power within the industry. Large firms have scale of investment; small specialist firms have agility. Each has different strengths.
🏢 Large firms
- Can invest in internal AI platforms and big partnerships (BCG and KPMG in Section 1)
- Can go after the AI implementation market
- ⚠ Moving away from billable-hour contracts is hard
- ⚠ Large organizations are less nimble
🚤 Boutiques (small specialist firms)
- With off-the-shelf AI, small teams can handle a large volume of research and materials
- Differentiate on deep expertise in a specific area
- ✅ Can change contract types easily
- ⚠ Can't match big firms' investment in internal AI platforms and data
If AI fills the boutiques' weak spot — manpower — there may be more situations where small specialist firms can compete with the big ones. On the other hand, a platform that lets AI search a firm's past materials (like McKinsey's Lilli) is easier to build for large firms with years of accumulated work. We found no reliable statistics showing which side has gained, so this section simply sets out both. What they share is that expertise and how you use AI matter more than size.
8. Advice for Aspiring Consultants and Client Companies
So how should each group act? Here is the overview turned into practice.
What I think matters most is that the companies buying consulting need to get smarter. When anyone can produce "convincing-looking materials" with AI, the value of paying consultants narrows to "judgment you can't produce in-house and the ability to see change through." As the Section 2 experiment showed, AI's answers are error-prone outside its frontier. Whether you can trust a firm with that judgment becomes the criterion for hiring one.
Summary
- AI has become a product too: BCG's 2025 revenue was $14.4 billion, AI- and tech-focused services topped 40% of it, and AI services grew 25% year over year. Headcount rose to 33,500.
- Tasks it speeds up and tasks it gets wrong: in the experiment with 758 BCG consultants, AI users did 12.2% more tasks, 25.1% faster and at higher quality on tasks inside AI's frontier — but were 19% less likely to be correct on a task outside it.
- The occupation is projected to grow: BLS projects management analysts (1,077,100 jobs) to grow 10% by 2035. What changes is the content more than the count.
- The pyramid and pricing are open questions: how to train juniors and how far pricing moves from hours to outcomes are real issues, but there are no reliable statistics yet on how far things have moved.
- The unchanging core: framing questions, judging AI's answers, trust and execution stay with people.
In the end, the question AI puts to consulting is: "Is your value in the production work, or in judgment?" When AI can turn out research and decks quickly, what people pay for is the ability to see the real problem, to doubt AI's answer where it should be doubted, and to move an organization to results.
Corrections to the earlier version
In the rewrite of September 30, 2026, we corrected the following statements in the earlier version to match the original sources.
| Earlier version said | After the rewrite (per the original source) |
|---|---|
| About 25% (about $3.6 billion) of BCG's $14.4 billion 2025 revenue came from AI-related consulting | BCG's announcement says "AI- and tech-focused services now represent over 40%" of revenue and that AI services grew 25% year over year. The 25% was a growth rate, not a share of revenue |
| In the HBS study, AI users produced 40%+ higher quality | The abstract of the paper published in March 2026 says "significantly improved quality." We also added that on the task outside AI's frontier, AI users were 19% less likely to be correct |
| Big Four and strategy firms invested $10B+, PwC $1B over three years, Lilli scans 100,000+ documents in seconds, BCG's Deckster, about 80% of junior work automated, 73% of clients prefer outcome-based pricing, small-firm growth of up to 50%, etc. | Removed because we could not confirm them in the original sources. We added official announcements from BCG, KPMG and McKinsey and BLS occupational statistics instead |
FAQ
Q. Will AI eliminate consulting jobs?
A. "The content changes" is closer to reality than "they disappear." The U.S. Bureau of Labor Statistics projects management analyst jobs, which include consultants, to grow 10% from 2025 to 2035, and BCG grew its workforce to 33,500 in 2025. At the same time, research, slide-making and analysis move to AI, and the weight shifts to framing problems, judging AI's answers and moving clients.
Q. How much faster does AI make consulting work?
A. In an experiment with 758 BCG consultants (Organization Science, March 2026), people using AI on tasks inside AI's frontier completed 12.2% more tasks, 25.1% faster, with significantly higher quality. But on a complex business problem outside the frontier, people using AI were 19% less likely to produce a correct answer. These results are for the model used at the time (GPT-4).
Q. Will it get harder for junior consultants?
A. Because AI speeds up juniors' core work of research and slide-making, the premise of hiring in bulk for grunt work wobbles. However, we could not confirm any official figures on how much big firms have cut junior hiring. What is certain is that new hires are expected from day one to use AI well and to catch its mistakes.
Q. How will consulting fees change?
A. Under hourly contracts, finishing faster with AI means billing less, so firms have little incentive to use AI fully. Outcome-based and fixed-price contracts are discussed as the way out, but there are no reliable figures yet on how far the industry as a whole has shifted.
Q. Do large firms or boutiques have the advantage?
A. It can't be said either way. Large firms can invest in internal AI platforms and big partnerships and, like BCG, turn AI support itself into a major business. Boutiques can handle large workloads with small teams using off-the-shelf AI and can change contract types easily. We found no reliable statistics showing which side has gained.
Q. What should companies that hire consultants watch out for?
A. Assume your own AI can produce research and materials, and ask consultants for "judgment you can't produce in-house" and "the ability to see change through." Because AI's answers are error-prone outside its frontier, make "can we trust them with that judgment" your selection criterion.
Q. What should I learn if I want to become a consultant?
A. Take AI tools as a given, then build the ability to see the real problem, to doubt AI's answers where they should be doubted, and to move clients and organizations. Develop deep expertise in a specific industry or function early. You'll be judged on the quality of your judgment and execution, not the speed of your production.
