Consulting's AI Shift Is Real — and It Hits Differently by Career Level

Allison Heithoff spent her first few years at West Monroe doing what most consultants do at the start: gathering client data, configuring systems, building and testing features, then deploying them. Hands-to-keyboard work. The kind that fills a day and proves you're useful.

Five years in, AI can do most of it.

That sentence could read as a warning. For Heithoff, it became something more disorienting — a career that kept her employed while quietly rearranging what "employed" meant. She's a business-administration graduate who now codes client solutions she never expected to touch. Nobody told her that was coming. The work just shifted, incrementally, until one day the old job description no longer fit.

If you work in consulting and that sounds familiar — not catastrophic, exactly, but quietly unrecognizable — this is the story of where that shift is actually going. Not the executive version. The ground-level one.

According to BCG's 2026 AI-at-Work survey of nearly 12,000 employees, 74% of frontline workers now use AI daily or several times a week — up 23 percentage points from the prior year. This isn't early-adopter territory anymore. It's the baseline. Which means the question isn't whether AI is changing consulting work. It's who gets hurt by the change, who doesn't, and what the difference between them actually looks like.

What AI Actually Does — and Doesn't — Do Inside a Consulting Engagement

Before diagnosing who's most at risk, it helps to understand what AI is genuinely capable of inside real engagements — because the headlines get this wrong in both directions.

A preregistered experiment by Harvard Business School and BCG assigned 758 BCG consultants 18 realistic consulting tasks. With AI access, they completed 12.2% more tasks, worked 25.1% faster, and produced output scored 40% higher in quality than peers working without AI. Inside the frontier of what AI handles well, the gains are real and measurable.

Outside that frontier, the same consultants using AI performed below their no-AI baseline. The AI produced confident-sounding wrong answers — and the consultants couldn't always tell the difference. The researchers described the best performers as "cyborgs": people in constant back-and-forth with the machine, routing specific subtasks to AI while holding the judgment layer themselves. Not delegating. Routing.

Louis-Charles Généreux, an engagement manager at McKinsey, lives this distinction. On a 70-person project with a North American cable company, he replaced the firm's standard version-controlled slide deck approach with an AI-maintained website that updates in real time. At the end of each week, the system generates a podcast-style summary and memo for the team. Engineers and senior executives look at the same information. "Everyone, irrespective of their knowledge or skills, sees the exact same thing," he said. PowerPoint hasn't disappeared — it's now the final output, not the daily work surface.

McKinsey's internal AI platform, Lilli, is active with 72% of the firm and processes over 500,000 prompts per month. The firm's CEO reported it saved 1.5 million hours of human work in a single year, largely through automating research and analysis.

The takeaway isn't that AI handles everything or nothing. The tasks it handles well are the ones that involve synthesizing existing information, generating a plausible first draft, or running a structured comparison. The tasks it handles badly — reading the room, knowing which data the client will trust, making a call under ambiguity — are exactly the tasks that define senior consulting value. Your job is to know which category you're in on any given afternoon.

The Career Pyramid Is Inverting — and Most Firms Aren't Saying So

Knowing what AI does well is useful. The harder question — the one most consulting leaders are avoiding in their public messaging — is what happens to the people whose jobs were built on exactly the tasks AI does well.

Revelio Labs analyzed hiring data at top consulting firms through January 2026. The findings are stark: overall talent demand is roughly 20% below the 2023 peak, while consultant-specific demand is about 40% lower. Senior consultant demand is up 55% since 2020. Entry-level consulting hiring is down 10% from its 2023 peak. By 2025, AI roles at top firms outnumber entry-level consultant postings — a ratio that sat at four entry-level roles for every one AI role as recently as 2015.

Heithoff's arc at West Monroe is a data point inside this trend. The hands-to-keyboard work that used to fill a junior consulting day — the gathering, configuring, testing, deploying — is precisely what Revelio's job-posting analysis shows declining fastest in entry-level listings. That she's still employed, now coding solutions she never expected to touch, reflects individual adaptation. The broader pattern is what matters: the training pipeline that used to run through manual data work is narrowing.

In terms of pure tech usage, consulting firms are still a long way from the frontier of what's possible. There is no shortage of highly intelligent, capable people in the industry, but 99% are not coders.
— Charlie Cheesman, Former Senior AI Consultant, EY-Parthenon

The institutional logic behind this inversion is named plainly in a documented Reddit account from a senior consultant at an unnamed firm. Given a year-long mandate to "push the boundaries of AI," this person built a full-lifecycle AI orchestration layer — project management, requirements gathering, architecture, implementation — and measured a 5x productivity gain. They demoed it to leadership. The response: silence. No interest in scaling. No questions about the methodology. A commenter later surfaced the mechanism: "A consultancy doesn't typically strive for efficiency in what they do. 5x more productive? Great, but their billable hours would be fewer."

The mandate was a checkbox. The capability threatened the model.

This is the fault line the official consulting AI narrative tends to skip. Généreux's success story — a productivity gain that leadership featured publicly — and the Reddit consultant's story — a productivity gain that leadership quietly buried — aren't contradictions. They're the same dynamic playing out differently depending on whether the efficiency you create fits the firm's business model or undermines it. Your interest in your own AI productivity and the firm's interest in your AI productivity are not always the same interest.

If you're early-career, the risk isn't replacement by AI — it's that the entry-level role meant to train you into seniority is being compressed. If you're mid-career, the risk is staying expert only in the generalist tasks AI is absorbing, while specialist demand surges. Indeed's Hiring Lab data from February 2025 shows management consultant roles went from 0.2% to 12.4% of all generative AI job postings in a single year. A quarter of all management consultant job listings now mention generative AI. That's a credential signal, not a bonus skill.

The Mandate Contradicts the Business Model — and Nobody's Resolving That Tension

Which raises the question no firm's official communications answer honestly: if the mandate is to use AI but the business model resists the full implication of AI productivity, where does that leave the consultant trying to navigate both?

Thomson Reuters surveyed more than 1,500 legal, tax, accounting, and compliance professionals across 26 countries in 2026. Forty percent received contradictory guidance from clients and leadership on AI tool use. Half had no client conversations about AI use in their work at all. Fifty-seven percent are using publicly available AI tools — often without enterprise approval — because internal tool rollouts lag actual usage. Nearly 80% cite accuracy concerns as their top barrier. Concerns about job displacement doubled year over year.

In October 2025, Deloitte Australia was required to issue a partial refund on a $440,000 government report after University of Sydney academic Dr. Christopher Rudge identified hallucinations — fabricated references and citations — consistent with unreviewed AI output. When the updated version was uploaded, AI use was acknowledged in an appendix. A Labor senator's response: "Deloitte has a human intelligence problem." It was the first publicly documented AI quality failure at a major consulting firm in a government engagement, and it materially raised the bar for client-side AI due diligence.

We can't lose sight of having resources that fundamentally understand what models are doing and how the AI is working, or we will create significant risk.
— Traci Gusher, Americas AI and Data Leader, EY

Charlie Cheesman spent five years at EY-Parthenon — including writing the firm's UK AI strategy and setting up its AI Lab — before leaving to build an AI-native firm. His read on the industry he left: consulting firms are "a long way from the frontier of what's possible" because "99% are not coders." KPMG's AI workforce lead, Niale Cleobury, is more measured but no less honest: he "probably doesn't 100% know the answer" to how junior consultants develop core skills when AI handles the work that used to teach them.

There's no single regulatory floor covering AI in consulting. Exposure runs through client domains — HIPAA for healthcare clients, FINRA rules for broker-dealers, NIST frameworks for federal work — without a consulting-specific law. The ICMCI published the first consulting-specific professional AI standard in May 2026: voluntary, not law. The governance is catching up to the practice, not leading it.

The consultant who navigates this period well isn't waiting for the firm to hand them a clear policy before moving. That policy may not arrive — or may arrive contradictory. Building your own AI judgment framework is not optional.

What the Consultants Navigating This Well Are Actually Doing

Heithoff didn't have a roadmap. She had a direction: toward the layer AI was exposing, not away from it. A business-administration graduate who now codes client solutions wasn't executing a pivot strategy — she was paying attention to what the work was becoming and staying close enough to be useful in its new shape. That's not a heroic narrative. It's an incremental one. Most real ones are.

The consulting career risk isn't in what AI is replacing. It's in remaining expert only at the things AI is replacing — the synthesis, the first draft, the benchmark comparison — while the judgment layer, the client read, the interpretation, keeps getting more valuable and more visible. The Revelio hiring data is clear: senior advisory demand is up 55%, entry-level generalist demand is falling. The direction of value is not a mystery.

This week: pick one repeating task in your workflow — a research summary, a slide outline, a status update — and run it through your firm's approved AI tool, or a carefully sandboxed public one with no client data. Note where the output is usable and where it's wrong with confidence. That gap — between what AI drafts and what you have to fix — is a map of where your judgment is still irreplaceable. Spend more time there.

The change is real, the map is incomplete, and moving is better than waiting for the map.

Recommended Tools & Resources