AI is not going to kill management consulting. It is going to do something more interesting and more uncomfortable: split it in two. Here is my read on where the industry is heading in 2026, and what it means if you buy consulting, sell it, or compete with it.
The uncomfortable headline
For decades, the consulting business model had a quiet engine at its centre: leverage. A small number of expensive partners, standing on a large pyramid of bright, cheap junior analysts doing the research, building the models, and making the slides. That pyramid is exactly the work that generative AI does well.
So the 2025-2026 news should not have surprised anyone, even though it clearly did. McKinsey is reported to be cutting several thousand roles, on the order of 10% of staff, pulling headcount back toward 40,000 after a hiring spree pushed it past 45,000. (Fast Company) KPMG’s US arm cut around 400 advisory roles in the same period. (I will flag honestly: the McKinsey figures come from trade reporting, not an official firm statement, so treat the exact number as reported rather than confirmed.)
What makes this different from a normal downturn is the cause. This is not consultants being cut because demand fell. It is consultants being cut because the work they did can now be done by software.
But look who is growing
Here is where the “consulting is dying” narrative falls apart, and why “fracture” is the better word than “collapse.”
While McKinsey trimmed, BCG grew. BCG reported 2025 revenue of US$14.4bn, up about 7%, with AI services making up around 25% of revenue (roughly US$3.6bn) and AI plus tech combined exceeding 40%. (City AM) Its headcount actually grew, tilted toward AI engineers and data scientists.
Accenture told the same story at scale. In FY2025 it reported around US$2.7bn in generative and agentic AI revenue, roughly triple the year before, with US$5.9bn in new GenAI bookings against US$80.6bn in total new bookings. (Accenture SEC filing) From 1 September 2025 it merged its services into a single unit, “Reinvention Services,” and CEO Julie Sweet was blunt that staff whose skills could not be reskilled “on a compressed timeline” were being exited. (Forbes)
So the split is already visible in the numbers:
| Firm | 2025 signal | What it tells you |
|---|---|---|
| McKinsey | ~10% headcount cut (reported); Lilli AI used by 75%+ of staff | Automating the pyramid |
| KPMG (US) | ~400 advisory roles cut | Same pressure, smaller scale |
| BCG | Revenue +7%; AI ~25% of revenue; headcount up | Growing on AI |
| Accenture | ~US$2.7bn GenAI revenue (3x); reorg to “Reinvention Services” | Repricing the whole firm around AI |
The firms shrinking and the firms growing are running the same play. They are all automating the commodity layer. The difference is whether they are also successfully selling the new AI-built work on top.
The internal proof: McKinsey’s own AI
The most telling data point is not a layoff, it is an adoption stat. McKinsey’s internal AI platform, Lilli, is now used monthly by more than 75% of its roughly 43,000 staff, with heavy users hitting it around 17 times a week and reporting time savings near 30%. It is the only platform cleared to handle confidential client data. (McKinsey)
Read that carefully. The firm whose entire value proposition is “we know how to work” has rebuilt how it works around AI, internally, at near-universal adoption. When the people who sell transformation transform themselves first, that is the signal to pay attention to.
The real shift: from bodies to outcomes
The deepest change is not headcount. It is the pricing model.
Consulting’s traditional pricing is a proxy: you pay for time and talent (bodies on the project, hours on the clock) as a stand-in for value delivered. That proxy only holds while the work genuinely takes that many people that many hours. AI breaks the proxy. When a task that used to need an analyst team for three weeks is done in an afternoon, the billable-hours model punishes the firm for being efficient and insults the client’s intelligence.
So the industry is being pushed, slowly and reluctantly, toward outcome-based pricing: fees tied to a measurable financial result rather than to time and headcount. Productivity gains of 30-60% in knowledge work are becoming the baseline expectation, and clients are starting to ask why they are paying for hours the AI no longer needs.
This is the same discipline I push with every AI project on the delivery side: stop counting activity, start measuring outcome. I wrote the honest version of that in agentic AI ROI: the honest numbers, and it is exactly the lens buyers should now apply to their consultants.
The Australian dimension: a trust crisis on top of a tech shift
Everything above is global. In Australia, it lands on top of something else: a trust crisis that has been reshaping the market independently of AI.
- Government spend on the Big Four has collapsed. New federal contracts fell to around A$348m in 2025, down from about A$637m the year before, close to a 45% drop. (Reuters via Yahoo)
- The PwC tax scandal is still reverberating. PwC sold its government advisory arm, roughly a fifth of its revenue, for A$1 and stepped back from government work for more than a year. (PwC tax scandal)
- A 2026 KPMG audit-leak scandal deepened it further, with the firm banned from federal bidding until September 2026 and Treasury now weighing forced structural separation of audit and consulting arms, plus new ASIC powers. (Reuters)
For an Australian buyer, this is the double bind the incumbents are in: AI is commoditising the low end of their work at exactly the moment their premium (trust, judgement, the partner relationship) is under the most scrutiny in a generation. The thing AI cannot replace is the exact thing the scandals have damaged.
What survives, and what gets more valuable
If the commodity layer is automating, what is the durable core? The counterpoints from people who study professional services are consistent, and I agree with them.
AI does not absorb accountability. It manipulates symbols. It does not set risk appetite, own the decision, or carry the consequences when it is wrong. Fiduciary duty cannot be delegated to a model. (Spencer Stuart)
As intelligence gets cheap, trust and context get expensive. When anyone can generate a competent analysis in minutes, the scarce goods become judgement under genuine ambiguity, the political skill to get a decision made inside a real organisation, and the trusted relationship that lets a client act on hard advice. Those get more valuable, not less. (Forbes)
So the consultant who is in trouble is the one who was really a well-paid research-and-slides engine. The consultant who is fine, or better than fine, is the one whose value was always judgement, relationships, and accountability, now amplified by AI instead of buried under grunt work.
What this means for you
If you buy consulting: stop paying for bodies and hours. Ask what AI the firm uses, demand outcome-based engagements tied to measurable results, and reserve your budget for genuine judgement and accountability rather than analysis you could now get from a model. The why 95% of AI pilots fail lesson applies here too: the value is in the integration and the decisions, not the deck.
If you sell consulting: automate your own commodity layer before a competitor does it to you, and reprice around outcomes. Your junior-analyst pyramid is a cost centre now, not a moat.
If you are building the internal capability instead: this is the strongest argument yet for owning AI capability in-house rather than renting it by the hour. That is the whole premise behind a fractional CAIO and a real AI transformation roadmap: build the muscle, do not just buy the slides.
Consulting is not dying. The part of it that was really just cheap intelligence for hire is. The part that was always about trust and judgement is about to matter more than ever, which is a good outcome for clients, and a hard one for any firm that forgot which business it was actually in.
Amjid Ali is a technology and AI leader based in Melbourne who helps organisations build AI capability in-house. If you would rather own the capability than rent it by the hour, get in touch.