Will Consulting Be Replaced By AI?

The Short Answer: Transformation, Not Replacement

No, AI will not replace consultants, but it is already rewriting how consulting gets done. The evidence points to a profession being restructured around AI rather than eliminated by it, and the changes are visible now in headcount, pricing, and who the competition even is.

The clearest proof comes from a joint study by Harvard Business School and BCG involving 758 consultants. Those using AI completed 12.2% more tasks, worked 25.1% faster, and produced results over 40% higher in quality, with the biggest gains going to bottom-half performers, whose quality jumped 43%. That is augmentation, not substitution.

The big firms have responded accordingly. Since 2023, the Big Four and strategy houses have poured more than $10 billion into AI, Deloitte $3 billion, EY $1.4 billion, PwC $1 billion, and McKinsey an estimated $3 billion or more. McKinsey's internal AI platform, Lilli, now counts 72% of its 45,000 employees as active users, handling roughly 500,000 queries a month and saving about 30% of their time.

At the same time, an unexpected threat has emerged: the AI labs themselves. OpenAI and Anthropic are directly targeting the roughly $1 trillion management consulting market, and OpenAI has launched its own consulting wing. Palantir fields "forward-deployed engineers" offering advisory services, and AI-powered startups are betting machines can do strategy work.

The structural shifts already underway tell the real story:

• McKinsey's workforce has shrunk from about 45,000 employees in 2022 to around 40,000, with a further 10% reduction announced in December 2025.
• Accenture cut roughly 11,000 roles while pledging 80,000 AI-focused hires, CEO Julie Sweet warned employees who cannot transition to AI-based roles they would be "exiting on a compressed timeline."
• Roughly 25% of McKinsey's 2025 global client fees came from outcome-based contracts, as hourly billing becomes harder to justify alongside claimed efficiency gains.
• PwC introduced an engineering career track for the first time in its 170-year history, and Deloitte scrapped traditional "analyst" and "consultant" titles.

So the verdict is transformation, not extinction. AI absorbs the repetitive analytical work, data crunching, research synthesis, first drafts, while human consultants retain the judgment, client trust, and accountability that machines cannot supply. Notably, publicly available AI tools are now "close to being on par" with big-firm proprietary platforms, which means smaller firms and independent professionals can compete on capability too.

The practical question for any business is no longer whether to use AI, but which AI tools actually deliver. Verification matters: published research found roughly 26% of agent skills carry at least one vulnerability. That is why we are building SkillzBench, a free directory that benchmarks agent skills in a sandbox and publishes the Lift in percentage points, alongside security scanning and trust scores, so you can see which skills genuinely help before you rely on them.

Why the Ground Is Shifting: Headcount, Pricing, and Entry-Level Work

The numbers coming out of the world's biggest consulting firms don't read like a typical downturn. They read like a structural rewiring of who does the work, what gets billed, and which skills matter.

Consider the headcount data. McKinsey has shrunk from roughly 45,000 employees in 2022 to about 40,000 by mid-2025, with a further 10% reduction announced in December 2025. Accenture cut approximately 11,000 roles while simultaneously pledging to hire 80,000 AI-focused workers, a signal that this is less about shrinking and more about swapping traditional consulting roles for AI-capable ones. Accenture CEO Julie Sweet made the stakes explicit: employees unable to transition into AI-based roles will be "exiting on a compressed timeline."

An honest caveat belongs here: these layoffs cannot be definitively attributed to AI, as industry analysts note. Post-pandemic over-hiring and softer demand for generalist services are plausible contributors. But the pattern of cuts paired with aggressive AI hiring, Accenture added roughly 40,000 AI and data professionals in two years, and EY added 61,000 technologists since 2023, is hard to dismiss as coincidence.

The pricing side of the business is shifting just as visibly. When AI delivers productivity gains of 30–60%, firms cannot logically charge hourly rates for work that takes a fraction of the time. McKinsey reports that about 25% of its 2025 global client fees came from outcome-based contracts, and firms like PwC are experimenting with subscription pricing for proprietary AI platforms. IBM describes the broader move toward "asset-based consulting," where expertise gets packaged into reusable AI-powered tools.

Entry-level analytical work faces the greatest exposure, and the reasons are straightforward:

• Research synthesis, financial modelling, and documentation, the core tasks of junior consultants, are exactly what generative AI does fastest and cheapest.
• A Deloitte US consultant with six years' experience describes the job shifting "from creation to validation," with teams expected to deliver more with the same or fewer people.
• A 2017 McKinsey study estimated 45% of consultant activities could be automated even with then-current technology, before modern LLMs existed.
• Deloitte has scrapped "analyst" and "consultant" titles entirely, while PwC introduced an engineering career track for the first time in its 170-year history.

This matters beyond the big firms. The same entry-level analytical tasks AI is absorbing inside consultancies are the tasks businesses everywhere are automating with agents and skills, which is precisely why verifying what those AI tools actually do, rather than trusting them blindly, has become a competitive differentiator. That verification gap is what SkillzBench is being built to close, with a free directory that benchmarks agent skills and publishes the measured Lift before you deploy anything.

The ground is shifting. The question is no longer whether, but how fast, and who adapts first.

The Reality Check: Trust Gaps, Hallucinations, and Weak ROI

The AI consulting hype has a paper trail of failures behind it. Deloitte Australia partially refunded an AU$440,000 contract after delivering a report with fabricated citations, a warning about what happens when AI output ships without human review.

The trust gap is measurable. In a KPMG and University of Melbourne study spanning 48,000 respondents across 47 countries, only 46% of people said they were willing to trust AI. Worse, 66% admitted using AI output without evaluating its accuracy, and 56% reported making mistakes because of unchecked AI.

The return on investment tells a similar story. While 88% of organisations now use AI in at least one function, MIT's NANDA project found that 95% of organisations reported no gains from AI implementation. Only 39% attribute any EBIT impact to AI, and PwC's 2026 survey found just 12% of CEOs reporting both cost savings and revenue gains, 56% reported neither.

Client expectations are cooling alongside the results. The share of clients expecting significant AI impact on service delivery fell from 60% in 2024 to 40% today, according to Source Global research. By 2026, clients are no longer impressed by AI-generated work, they want rapid, tailored solutions that actually deliver.

Why does human validation remain the bottleneck? Three reasons stand out:

• Hallucination risk is real and expensive, Deloitte's refund proves fabricated citations can trigger financial and reputational damage.
• Most users don't check AI output, and the majority who skip verification report mistakes as a result.
• Organisations err by "bolting AI" onto existing workflows rather than redesigning them with people front and centre, as MCR's Jim Kennedy argues.
• Accountability still rests with humans, no-one submits an AI-generated tax return without professional sign-off.

This is also where the competitive advantage lives. A Deloitte consultant with six years' experience describes the job shifting "from creation to validation", and the firms and professionals who validate well are the ones clients keep paying. Verification is becoming a skill in its own right, especially as businesses assemble AI agents from third-party components. Published research found roughly 26% of agent skills carry at least one vulnerability, which is why SkillzBench is building a free directory that security-scans, sandbox-benchmarks, and publishes the measured Lift of agent skills before you rely on them.

The pattern across all this counter-evidence is consistent: AI produces impressive first drafts and unreliable final products. The gap between the two is exactly where human judgment, and the consultants who supply it, still earns its fee.

What Winners Do Differently: Redesign Workflows, Don't Bolt AI On

The firms pulling ahead of AI disruption share one habit: they refuse to treat AI as a plug-in. Jim Kennedy, co-founder of Cursus, identifies the biggest failure mode in AI adoption as "bolting AI" onto existing workflows within single functions, rather than redesigning horizontal, outcome-delivering workflows with people front and centre (MCR Consulting). That diagnosis helps explain why MIT NANDA found 95% of organisations reported no gains from AI implementation, the tool arrived, but the work stayed the same.

KPMG makes a parallel argument: adoption should be treated "explicitly as a people and skills challenge, not merely a technology one" (industry analysis). PwC operationalises this with a training model built on 30 core skills, 15 AI-centric and 15 human-centric, signalling that judgment, client relationships, and change management matter as much as prompt fluency (Business Insider).

Deloitte's client segmentation shows what this looks like in practice:

• 34% "Deep Transformers", redesign how work gets done and see the highest ROI
• 30% "Process Redesigners", restructure specific processes and capture moderate returns
• 37% "Surface Users", apply AI to existing tasks and risk stagnation with limited return

The pattern is unambiguous: value concentrates where workflows are rebuilt, not decorated.

Here is the twist that should energise smaller firms and independent professionals. In the HBS/BCG study of 758 consultants, bottom-half performers gained the most from AI assistance, a +43% jump in output quality, alongside the headline gains of 12.2% more tasks completed and 25.1% faster delivery (the study's findings). AI compresses the gap between your best day and your average one.

And the moat is thinner than it looks. Publicly available AI tools are "close to being on par" with the proprietary platforms the Big Four spent billions building (MCR Consulting). A boutique consultancy, or a solo operator, can now assemble a capable AI stack for the cost of a few subscriptions.

The catch is that bolting on tools is just as easy for a two-person firm as for a 40,000-person one. The winners, regardless of size, will be those who verify what their AI actually does before trusting it. That's the same principle behind SkillzBench's approach to agent skills, benchmarking, security-scanning, and publishing measured Lift so you equip agents with what demonstrably works, not what merely sounds impressive.

Your Action Plan: From AI User to Verified AI Operator

The consultants who thrive in the next five years won't be the ones who use AI the most, they'll be the ones who verify it the best. That distinction is already separating winners from the 95% of organisations that reported no gains from AI implementation in MIT's 2025 NANDA study.

The first step is a mindset shift. A Deloitte US consultant with six years' experience describes the job moving "from creation to validation", teams now deliver more with the same or fewer people. Your value is no longer drafting the deck; it's knowing, with evidence, that the deck is right. That matters because the trust data is grim: only 46% of people are willing to trust AI, yet 66% use AI output without checking its accuracy, and 56% report mistakes from unchecked AI. Deloitte Australia's partial refund of an AU$440,000 contract over fabricated citations shows what skipping verification costs.

Second, redesign workflows around outcomes instead of bolting AI onto existing processes. Jim Kennedy of Cursus argues the biggest failure mode is applying AI within single functions rather than rebuilding horizontal, outcome-delivering workflows with people front and centre. That's consistent with the ROI gap: 88% of organisations now use AI somewhere, but only 39% attribute any EBIT impact to it.

Third, move your pricing off the clock. With productivity gains of 30–60%, hourly billing is structurally eroding, roughly a quarter of McKinsey's 2025 global fees already came from outcome-based contracts. If you're selling expertise, price the result; if you're buying it, demand one.

Finally, and most critically: verify every AI tool and agent skill before it touches client work. The published research found roughly 26% of agent skills carry at least one vulnerability, and most professionals have no way to tell a useful skill from a risky one. Before deploying anything, build a simple validation habit:

• Test each skill or tool in a sandbox environment, never directly in production or client deliverables.
• Compare performance with and without the skill, so you know the actual lift it provides rather than trusting the description.
• Check security posture, licensing, and version history before granting access to your agents.
• Re-verify after updates, upstream changes can quietly introduce new risks.

This is exactly the gap SkillzBench is being built to close: a free, independent directory of AI agent skills that pins every version to a commit SHA, security-scans and benchmarks each skill in a sandbox with the same model, and publishes the Lift in percentage points alongside a 0–100 trust score, with money never touching the ratings. The free MCP connector will let you toggle exactly which skills each agent can see.

Adopting AI is no longer a technology challenge; it's a verification discipline. The consultants who master it will be the ones still charging premium fees while everyone else competes with a chatbot subscription.