Claude launched a legal plugin. Headlines declared consulting dead. Meanwhile, the Big Four crossed €220 billion in combined revenue, BCG posted its 21st consecutive year of growth, and OpenAI signed multiyear deals with McKinsey, BCG, and Accenture — because it can’t sell AI to enterprises without consultants. The evidence on who’s actually being disrupted, and what it means for business services.

SIGNAL DASHBOARD

Market size: Global consulting exceeded €1 trillion in 2025; management consulting alone at ~€460 billion — still growing 4-6%

Revenue reality: Big Four combined revenue surpassed €220 billion in FY2025; Deloitte alone crossed €65 billion

AI as revenue driver: BCG derives 20% of revenue (€2.5 billion ) from AI consulting; Accenture booked €5.5 billion ($5.9B) in generative AI deals

Workforce shift: McKinsey cut 5,000 roles; PwC cut 5,600 ; UK Big Four graduate intake fell up to 29% — but total industry headcount grew

AI tool reliability: Stanford RegLab found LLMs (Large Language Models) hallucinate 69-88% of the time on specific legal queries; 200+ legal hallucination cases documented in 2025

Client sentiment: 65% of enterprises say traditional consulting models no longer deliver value (HFS Research) — yet consulting revenue keeps rising

The paradox: OpenAI, Anthropic, and Google all signed enterprise distribution deals with consulting firms in 2025-2026 — they cannot scale without consultants

Verdict: Not dying — restructuring. AI is eliminating the commodity analysis layer while creating a gold rush in AI implementation consulting. Adapt the business model or lose the bottom half of your workforce to automation. The top half has never been more valuable.

The Extremes: What Could Go Right — and Wrong

Worst-Case Scenarios That Have Already Happened

  • DoNotPay (US), 2025: The self-proclaimed “world’s first robot lawyer” was fined €180,000 ($193,000) by the FTC (Federal Trade Commission) for falsely claiming its AI could substitute for licensed attorneys — it couldn’t pass basic legal competency tests
  • Anthropic internal (US), 2025: Anthropic’s own legal team was caught using Claude to generate citations that included hallucinated fake author names and fabricated article titles — the company making the tool couldn’t trust its own output
  • Mata v. Avianca (US), 2023: Attorney Steven Schwartz submitted six entirely fabricated case citations generated by ChatGPT to federal court; sanctioned and fined — the case that launched a global wave of judicial AI disclosure rules
  • Klarna (Sweden), 2025: After cutting workforce from 5,527 to ~3,000 partly using AI, customer satisfaction collapsed — CEO publicly admitted “we went too far” and began rehiring human agents
  • McKinsey (Global), 2025-2026: Cut approximately 5,000 employees — the largest reduction in its history — under pressure from opioid settlements (€1.5 billion/$1.6B), AI-driven workflow changes, and competitive erosion from BCG and Bain

Best-Case Scenarios With Evidence Behind Them

  • Accenture (Global), FY2025: Generated €3.8 billion ($4.1B) in generative AI revenue with €5.5 billion ($5.9B) in new AI bookings — AI consulting became a core growth engine, not a threat (vendor-reported figures)
  • BCG (Global), 2024: Derived 20% of revenue (€2.5 billion) from AI-related consulting, projected to reach 40% by 2026; grew 10% to €12.6 billion ($13.5B) — AI created more business than it displaced
  • Thomson Reuters CoCounsel (Global), 2025: Reached 1 million users across 107 countries; legal professionals reported 30-40% time savings on research tasks (vendor-reported)
  • Harvey AI (US/UK), 2025: Raised €920 million ($988M) total at an €7.5 billion ($8B) valuation; 100,000 lawyers on platform — though top-tier firms note efficiency gains remain difficult to measure independently
  • EY (Global), FY2025: AI consulting revenue jumped 30% year-on-year, helping drive overall revenue to €49.6 billion ($53.2B) — selling AI transformation to clients, not being replaced by it
  • Neither extreme is typical. The consulting industry has survived every previous “consulting killer” technology — internet, big data, cloud computing — and each generated more consulting demand, not less
  • The firms being hurt most (McKinsey’s cuts, PwC’s hiring freeze) face multiple pressures — legal liabilities, geopolitical shifts, and cyclical slowdowns — not just AI disruption
  • Best-case AI revenue figures come overwhelmingly from the vendors and consulting firms themselves; independent measurement of actual productivity gains remains scarce
  • Only **9.3% ** of US companies report using generative AI in production (Yale Budget Lab, 2025) — the gap between hype and deployment remains enormous
  • The “AI kills consulting” narrative is loudest among AI vendors and tech media; actual enterprise buyers are increasing consulting spend, not cutting it

The Core Evidence: What Research Actually Shows About AI Replacing Professional Services

The headline claim driving the narrative is that AI tools can now perform the analytical and advisory work that justifies consulting fees — making expensive human consultants redundant. The most rigorous evidence says something more complicated.

The Harvard Business School / BCG study (2023, published in Research Policy) is the landmark experiment. Researchers randomly assigned 758 BCG consultants to use GPT-4, GPT-4 with prompt engineering training, or no AI across 18 realistic consulting tasks. The results defined what the authors called a “jagged technological frontier.”

For tasks inside the frontier — creative ideation, market analysis, persuasive writing — consultants using AI completed 12.2% more tasks , were 25.1% faster , and produced results rated over 40% higher in quality . Below-average consultants improved by 43% , while top performers improved by just 17% . AI functioned as a dramatic equalizer.

But for tasks outside the frontier — those requiring nuanced judgment, data integration across ambiguous inputs, and contextual reasoning — AI-assisted consultants were 19 percentage points less likely to arrive at correct answers than those working without AI. The most dangerous pattern: consultants who blindly adopted AI output performed worst of all.

The MIT/Stanford customer service study (3 million chats) found AI boosted productivity by 14% , with low-skilled workers improving 35% and the most skilled workers seeing gains near zero. The MIT writing productivity study showed ChatGPT reduced task completion time by 40% while raising quality by 18% . The pattern is consistent: AI is a skill-leveler that disproportionately benefits junior workers — the exact workers consulting firms are now hiring fewer of.

The critical counterpoint comes from the Yale Budget Lab (October 2025), which analysed actual US labour market data and found no substantial acceleration in occupational mix changes compared to historical technology disruptions. Their verdict: only 9.3% of US companies report using generative AI in production. The gap between what AI can do and what organisations are actually deploying remains vast.

NOISE vs SIGNAL: The “AI Kills Consulting” Narrative

What the headlines claimWhat independent evidence shows
“Claude’s legal plugin will replace lawyers”Anthropic explicitly frames tools as assistance, not advice; its own legal team produced hallucinated citations
“AI can do 80% of what consultants do”Harvard/BCG study: AI excels at bounded analytical tasks but fails at judgment, integration, and implementation
“Consulting is dying — look at McKinsey layoffs”McKinsey’s cuts driven primarily by €1.5B opioid settlements, Saudi spending cuts, and DOGE austerity — not AI displacement
“Enterprises are replacing consultants with AI tools”Only 9.3% of US firms use generative AI in production; enterprise consulting spend grew 4-6% in 2025
“AI advisory tools are reliable enough for business decisions”Stanford RegLab: LLMs hallucinate 69-88% on specific legal queries; 200+ hallucination cases documented in law alone
  • AI genuinely accelerates bounded analytical tasks — research, document review, data synthesis, report drafting — by 25-40% for average performers
  • AI consistently fails at the tasks consulting actually sells: judgment under ambiguity, organisational politics, implementation, stakeholder management, and accountability
  • The productivity gains are real but concentrated in the commodity layer of consulting (junior analyst work) — not the advisory layer that commands premium fees
  • The perception gap is systematic: professionals overestimate AI’s benefit by 20-40 percentage points (Harvard/BCG, METR studies)
  • No independent study has demonstrated AI replacing end-to-end consulting engagements; the gains are task-level, not service-level

Why the Revenue Keeps Growing While the Headlines Say “Dead”

The central contradiction demands explanation: how can consulting revenue grow 4-6% in 2025 while headlines proclaim the industry’s death?

The answer has three parts. First, AI is creating more consulting demand than it destroys. Every enterprise deploying AI needs help with strategy, vendor selection, implementation, change management, governance, and compliance. This is precisely what consulting firms sell. BCG’s 20% AI revenue share, Accenture’s €5.5 billion in AI bookings, and EY’s 30% jump in AI consulting revenue are not coincidences — they are the structural reality of technology adoption. The pattern repeats from every prior technology wave: internet consulting boomed in the late 1990s, cloud consulting boomed in the 2010s, and AI consulting is booming now.

Second, AI vendors cannot scale enterprise deployments without consulting firms. In February 2026, OpenAI formed its “Frontier Alliance” with McKinsey, BCG, Accenture, and Capgemini. Anthropic struck similar partnerships with Deloitte, Accenture, and Cognizant. OpenAI has approximately 70 forward-deployed engineers for on-site customer work. Consulting firms have hundreds of thousands. Capgemini’s Chief Strategy Officer Fernando Alvarez captured the dynamic in Fortune (March 2026): AI companies build transformative technology but cannot navigate the organisational transformation required to deploy it. “If it was a walk in the park, OpenAI would have done it by themselves.”

Third, the firms reporting pain are facing multiple pressures, not just AI. McKinsey’s 5,000-person reduction reflects €1.5 billion in opioid-related legal settlements, Saudi Arabia pulling back government consulting contracts, US federal austerity (DOGE-related), and competitive erosion from BCG — not a sudden collapse in demand caused by ChatGPT. PwC’s hiring freeze and 5,600-person cut likewise reflect slow advisory revenue growth and strategic repositioning, not AI replacement.

The reconciliation is straightforward: AI is disrupting what consulting firms do internally while simultaneously expanding what they sell externally. The firms that are winning — BCG, Accenture, EY — are those that recognised this duality fastest.

The Adoption Map: Where Professional Services Firms Actually Stand

Big Four and MBB Financial Reality (FY2025)

FirmRevenueYoY GrowthAI PostureHeadcount Trend
Deloitte~€65B ($70B)+4.8%470,000+ employees; AI integrated across service linesGrowing
PwC~€53B ($56.9B)+2.7%Scrapped 100,000-job pledge; cut 5,600Shrinking
EY~€49.6B ($53.2B)+4%AI consulting revenue up 30%Stable
KPMG~€37.1B ($39.8B)+4.6%UK graduate intake cut 29% (1,399 to 942)Mixed
BCG~€12.6B ($13.5B)+10%20% of revenue from AI consulting; headcount grew to 37,000Growing
McKinsey~€16B (est.)Flat/declining12,000 internal AI agents; cut ~5,000 employeesShrinking
Accenture~€67B ($72.4B)+7%€5.5B ($5.9B) in AI bookings; €3.8B ($4.1B) AI revenueGrowing

What separates winners from the distressed

The consulting firms thriving in 2026 share a common pattern: they pivoted to selling AI transformation rather than defending against it. BCG built a dedicated AI & Technology Advantage practice that now accounts for one-fifth of all revenue. Accenture invested over €2.8 billion ($3B) in AI capabilities and reported AI bookings doubling year-on-year. EY positioned tax and AI consulting as twin growth engines.

The firms struggling — McKinsey and PwC primarily — share a different pattern: legacy liabilities (legal settlements, overexpansion during the pandemic hiring boom, geopolitical exposure) compounded by slower strategic pivots toward AI-native service delivery.

NOISE vs SIGNAL: What Leading Firms Do Differently

Firms under pressureFirms capturing value
Treat AI as a cost-cutting tool for internal operationsTreat AI as a revenue-generating service offering
Cut junior headcount to improve short-term marginsRestructure teams around AI-augmented delivery — fewer juniors, more senior oversight
React to “AI kills consulting” narrative defensivelySell the narrative: “You need us to implement AI”
Maintain hourly/project billing modelsShift toward outcome-based pricing where AI leverage improves margins

The Money Question: Business Model Surgery

The Simple Arithmetic — and Why It’s Misleading

A mid-tier management consulting engagement in Europe runs €150,000-500,000 for a 6-8 week project. The typical team: 1 partner, 1 manager, 2-3 associates/analysts. AI tools can automate significant portions of what those associates do — market research, competitor benchmarking, financial modelling, slide generation, document review.

If AI reduces the associate-hours on a €300,000 engagement by 40% , the firm faces a choice: pass the savings to the client (reducing revenue to ~€180,000), maintain the price and pocket the margin improvement, or reinvest the freed capacity into higher-value activities. Most firms are choosing options two and three — which is why revenue hasn’t declined despite genuine productivity gains.

The Klarna Case Study: The Cautionary Tale

Klarna (Sweden)

The headline: Headcount fell from 5,527 (2022) to ~3,000 (2025). Revenue roughly doubled. AI chatbot handled 2.3 million customer conversations in its first month.

What actually happened:

  • 15-20% annual natural attrition combined with a hiring freeze drove most of the reduction — not targeted AI replacement
  • Despite “AI savings,” Klarna posted an €79 million (~$85M) pretax loss in Q1 2025 — nearly double the prior year
  • Customer satisfaction declined sharply; CEO publicly reversed course
  • Began rehiring human agents by May 2025
  • The aggressive AI narrative coincided with IPO preparation — cost-cutting optics were partly pre-IPO positioning
  • Forrester analyst assessment: Klarna “overpivoted to cost containment” without evaluating the longer-term impact on customer experience

The Klarna case is the consulting industry’s most important cautionary tale — not because AI failed, but because cost-cutting without service model redesign produces predictable backlash. The parallel for consulting firms: eliminating junior analysts without restructuring how senior partners deliver value will degrade quality, not improve margins.

Caveat: Most headcount reduction was natural attrition and a hiring freeze, not direct AI replacement.

The Workforce Pyramid Is Becoming an Obelisk

Harvard Business Review (September 2025) described the most consequential structural shift: the traditional consulting “pyramid” — a wide base of junior analysts supporting a narrow peak of senior partners — is transforming into an “obelisk”: tall, narrow, fewer layers.

The data is stark:

  • UK Big Four graduate intakes fell 6-29% in 2024-2025
  • Canadian non-senior consulting job postings dropped 40% between February 2022 and 2025
  • PwC internal projections show 32% reduction in tax and assurance associate hiring and 39% reduction in audit new hires by 2028
  • McKinsey claims tasks that once required 14 consultants now need 2-3 plus AI
  • An estimated 80% of a junior analyst’s typical research and slide-generation work is now automatable

NOISE vs SIGNAL: The Business Model Question

NoiseSignal
“AI replaces consultants”AI replaces analyst-level tasks — research, modelling, deck-building — not partner-level advisory
“Clients will just use ChatGPT instead”Only 25% of McKinsey’s fees globally are linked to outcomes; the rest is project/hourly billing that AI threatens
“Consulting margins will collapse”Firms using AI to deliver same output with fewer juniors are seeing margin expansion, not compression
“The junior hiring pipeline doesn’t matter”If juniors disappear, where do future partners come from? The apprenticeship model is broken with no clear replacement
  • Consulting revenue is growing because AI creates more implementation demand than it destroys in advisory demand
  • The business model threat is real but specific: hourly billing for commodity analysis is dying; outcome-based pricing for AI-augmented delivery is emerging
  • The Klarna playbook — cut headcount, claim AI efficiency, deal with quality collapse later — is a warning, not a template
  • The pyramid-to-obelisk shift creates a genuine talent pipeline crisis: fewer entry points for juniors means fewer future senior leaders
  • Headcount-based contracts are expected to decline from 49% of engagements today to just 16% within two years (HFS Research/IBM)

Risks: Hallucinations, Liability, and the Trust Deficit

The Reliability Problem Is Not Solved

The most alarming quantified finding: Stanford’s RegLab demonstrated that LLMs hallucinate 69-88% of the time on specific legal queries. Over 200 legal hallucination cases were documented in 2025 alone. MIT researchers found AI models are 34% more likely to use confident language when hallucinating — the more certain the AI sounds, the more likely it is wrong on complex questions.

This reliability gap is existential for professional services. A consulting firm’s core product is trusted judgment. When Paul Weiss’s Chief Knowledge Officer says the firm avoids hard efficiency metrics for Harvey AI because “the time and effort needed to carefully review the output made efficiency gains difficult to measure,” that’s not a ringing endorsement — it’s an honest admission that verification overhead consumes much of the productivity gain.

The Liability Architecture Has Not Changed

No jurisdiction has shifted professional malpractice liability to AI providers. The licensed professional — lawyer, auditor, consultant — retains full personal and firm liability for AI-assisted work product. The February 2026 United States v. Heppner ruling added a new dimension: documents created using consumer-grade Claude were held not protected by attorney-client privilege, because the AI constitutes a “third party.”

Insurance coverage gaps compound the risk. Many professional indemnity policies do not explicitly cover AI-related errors. The concept of “silent AI” — AI risks neither included nor excluded in policies — creates uncertainty that further incentivises human oversight.

The False Confidence Effect

The Harvard/BCG study found the most dangerous pattern: consultants who trusted AI output without verification performed worst of all — 19 percentage points worse than those working without AI on complex judgment tasks. The METR study on developers found identical dynamics: professionals predicted a 24% speedup but actually worked 19% slower, and still believed they had been faster after the fact.

This systematic overconfidence is the binding constraint on AI adoption in professional services. When the professionals themselves cannot accurately assess whether AI is helping or hurting, governance becomes non-negotiable.

Industry has consensus on:

  • AI excels at bounded tasks: document review, research synthesis, first-draft generation, data extraction
  • Human review of all AI-assisted professional work product is non-negotiable — no exceptions
  • AI tools are productivity multipliers for routine work, especially for junior professionals

Still unresolved:

  • Hallucination rates on domain-specific professional queries remain unacceptably high (69-88% in legal)
  • No liability framework shifts risk from professionals to AI providers — and none is expected near-term
  • The false confidence effect means adoption decisions based on professional enthusiasm alone will overestimate returns
  • Insurance coverage for AI-assisted professional work remains ambiguous across jurisdictions
  • Attorney-client privilege and professional confidentiality protections may not extend to AI-processed information

The EU AI Act

The EU AI Act, effective August 1, 2024, classifies AI systems assisting judicial authorities in legal research and application of law as high-risk under Annex III, requiring conformity assessments, risk management systems, data governance, and human oversight — with full compliance required by August 2026. AI tools for employment decisions (recruiting, performance evaluation) and financial services (credit scoring, insurance risk assessment) face identical high-risk classification.

For general management consulting AI tools — strategic analysis, market research, report generation — the classification is limited or minimal risk, requiring primarily transparency obligations and AI literacy training under Article 4 (effective since February 2025). Penalties for non-compliance reach up to €35 million or 7% of global annual turnover.

Unauthorised Practice of Law (UPL) — Global Constraint

AI tools that provide legal analysis or recommendations to end users — rather than assisting licensed attorneys — face Unauthorised Practice of Law (UPL) restrictions in virtually every jurisdiction. The FTC’s DoNotPay settlement (January 2025, €180,000/$193,000 fine) established the US enforcement precedent. The ABA’s (American Bar Association) Formal Opinion 512 (July 2024) made clear that all existing professional conduct rules — competence, confidentiality, supervision — apply fully to AI-assisted legal work. European bar associations have issued similar guidance.

The Privilege Problem

The Heppner ruling (February 2026) created a new risk: AI-processed documents may not be privileged. If a consultant or in-house lawyer uses consumer-grade AI to analyse sensitive documents, opposing counsel may argue the privilege was waived by sharing with a “third party” whose terms of service permit data collection. Enterprise-grade tools with appropriate data processing agreements may mitigate this, but the legal landscape is unsettled.

  • AI tools used in legal research, employment decisions, and financial advisory are classified high-risk under EU AI Act — full compliance by August 2026
  • General consulting AI tools (strategy, market analysis) are minimal/limited risk — transparency and AI literacy are the primary obligations
  • UPL restrictions are the hard ceiling: AI tools marketed as substitutes for professional advice face enforcement action in every major jurisdiction
  • Conduct a privilege review before deploying consumer-grade AI tools on client-sensitive materials — Heppner changed the risk calculus
  • Budget €25,000-75,000 for compliance readiness on high-risk AI deployments; minimal-risk deployments require primarily documentation and training

What to Do Monday Morning

If you run a consulting or professional services firm
  • Audit your revenue by task type: Identify which service lines are >50% commodity analysis (at risk) vs relationship/judgment-intensive (defensible) — price the defensible work higher, automate the rest
  • Shift billing models: Move at least one major service line to outcome-based pricing within 12 months — hourly billing for AI-automatable work is unsustainable; learn from Klarna’s margin-then-quality collapse
  • Build an AI consulting practice: The consulting firms growing fastest (BCG at 20% AI revenue, Accenture at €5.5B bookings) are selling AI transformation, not defending against it
  • Restructure the junior pipeline deliberately: Don’t just freeze hiring; redesign the first 2 years of analyst experience to focus on tasks AI cannot do — client relationship skills, judgment calibration, implementation, cross-functional work
  • Invest in verification infrastructure: Dedicated review layers, domain-specific testing protocols, and governance frameworks — this is a defensible competitive advantage when clients ask “how do you ensure AI quality?”
If you're a business owner considering replacing consultants with AI
  • Use AI for the commodity layer now: Market research, competitor benchmarking, regulatory scanning, first-draft reports — these are ready for AI self-service at **€18-170/month ** per tool depending on complexity
  • Keep consultants for the judgment layer: Strategy validation, board-level advisory, organisational design, implementation management, and accountability — no AI tool provides these, and the hallucination rates (69-88% on specific legal queries) make autonomous AI advisory dangerous
  • Run a parallel test: On your next consulting engagement, ask the firm what percentage of deliverables were AI-generated and what verification process was used — this reveals which firms have genuinely integrated AI vs which are just billing the same hours
  • Budget for the hybrid model: Expect to spend **40-60% less ** on pure research/analysis engagements within 2 years, but the same or more on implementation, change management, and AI transformation consulting
If you haven't thought about this yet
  • Start with one use case: Deploy a general AI tool (Claude, GPT-4, or Gemini) on a bounded professional task — contract review, market sizing, regulatory scanning — and measure time savings honestly against the verification overhead
  • Don’t fire your advisors based on a headline: The “AI kills consulting” narrative is driven by AI vendor marketing and tech media attention cycles; the actual enterprise evidence shows consulting spend increasing, not decreasing
  • Assess your regulatory exposure: If you operate in EU markets and use AI for employment, legal, or financial decisions, high-risk AI Act compliance is required by August 2026 — start scoping now; budget **€25,000-75,000 ** for readiness
  • Watch the Klarna case, not the AI demos: The most important business lesson in the AI-consulting debate is not about technology capability — it’s about the consequences of cutting human judgment too fast without redesigning service delivery

AI is not killing consulting. It is performing surgery on the consulting business model — hollowing out the commodity analysis layer from the bottom while creating an unprecedented gold rush in AI implementation consulting at the top. The Big Four crossed €220 billion in combined revenue in 2025. BCG’s fastest-growing practice is AI consulting. OpenAI literally signed distribution deals with the firms it’s supposed to be replacing.

The real risk is not to the profession — it is to the specific business model of billing hourly for tasks AI can do faster and cheaper. Firms that cling to pyramid staffing and hourly billing face margin compression and talent pipeline collapse. Firms that restructure around AI-augmented delivery, outcome-based pricing, and senior-heavy teams will capture an expanding market.

The consulting industry has survived every technology that was supposed to kill it. AI won’t be different — but the consulting firm that emerges from this transformation will be unrecognisable to the one that entered it.