"The death of consulting": what Accenture's fall and Karpathy's map really say about AI and our jobs

Analysis — by the Abbeal team · Paris · Montréal · Tokyo.

A clip went viral over the past few days: a 700,000-person consulting firm losing half its market value, an OpenAI co-founder scoring every job in America for AI exposure, and — in the background — a new role emerging to take over. The three stories are really one. Here is what they actually say, and our take.

Accenture: the day consulting wobbled

On June 18, 2026, Accenture had the worst trading day in its history as a public company: down nearly 20% in a single session. Over twelve months the stock has shed more than 40%, back to its 2017 levels. The drivers: weaker bookings and demand being eaten by AI across consulting and managed services.

In the same move, the group announced it was parting ways with roughly 11,000 employees — those who cannot be reskilled on AI fast enough. Work that once took ten people three months is increasingly done in a fraction of the time.

Karpathy: an AI-exposure score for every job

In parallel, Andrej Karpathy (OpenAI co-founder, former Tesla AI director) released a visualizer of the U.S. labor market. Starting from the Bureau of Labor Statistics Occupational Outlook Handbook — 342 occupations, about 143 million jobs — he had each one scored from 0 to 10 for its "digital AI exposure" using a model (Gemini Flash) applied to the official descriptions.

An important caveat, which he stresses himself: this is neither a study nor an academic paper, but a development tool for exploring data visually. He even took the hosted version down because of misinterpretations. Read it as a signal, not as gospel.

The counterintuitive result: the better paid, the more exposed

We were long told automation would hit low-skill jobs first. Karpathy's map says the opposite for this wave of AI: the weighted-average exposure sits around 4.9, but jobs paying over $100,000 a year average 6.7, versus 3.4 for those under $35,000.

• Highly exposed: medical transcriptionists (10), accountants and auditors (9), financial analysts (9), lawyers (9) — and, more broadly, consulting.
• Barely exposed: construction laborers, roofers, painters, janitors (1); home health aides, nursing assistants, barbers, bartenders (2).

The common thread among exposed jobs: work heavy on text, analysis, synthesis and documents — exactly the home turf of large language models. It is also, word for word, the core of consulting. Hence the phrase that spread across social media: "the death of management consulting."

Our take: exposure does not mean extinction

At Abbeal, we do not read this as the end of expertise, nor even as the end of the day rate. Billing for time worked stays perfectly valid. What changes is the type of profile you put on it, and the type of mission you give them. The pyramid of juniors doing slides and analysis no longer holds when AI does that work in hours. The senior who designs, codes and ships takes all the value.

"Exposure" measures the overlap between a job's tasks and what today's AI can do — not a scheduled destruction. And it is precisely where exposure is highest that the leverage is greatest: these jobs do not disappear, they reorganize around people who can steer AI.

This has been our conviction from day one: AI is engineering, not a gadget. A senior engineer fluent in agents, production RAG, evaluation and observability produces more, not less. The real divide is not "AI versus humans" — it is the tooled-up senior versus the pyramid org.

The role replacing the consultant: the Forward Deployed Engineer

The clip that kicked this off does not just bury consulting: it names the role taking its place, the Forward Deployed Engineer (FDE) — the engineer deployed right next to the client. Where the classic consultant produced slides and recommendations, the FDE ships software that runs, on the ground, with the client.

Four pillars define it:

• Understand the product — master the tech and the business, not just the deck.
• Enable the client — make them autonomous, not dependent on an engagement that drags on.
• Combine technical and product expertise — one brain that codes, designs and decides.
• Ensure smooth deployment — go all the way to production, not stop at the recommendation.

This profile is nothing new to us: it is exactly how Abbeal has operated from the start. Senior engineers embedded in client teams, owning a scope and delivering an outcome — augmented by AI. The FDE is the consultant reconciled with engineering.

What it changes in practice

• For companies: the day rate stays, but you no longer pay for the same profile. A senior who ships, not a junior who documents.
• For missions: fewer slides and recommendations, more build. Engineering that reaches production, not a report.
• For talent: seniority, judgment and the ability to ship AI to production become the scarce asset. Those who seize it rise; those who wait get commoditized.

Accenture's fall is not the death of consulting, nor of the day rate. It is the death of consulting that sold junior time on deliverables AI now produces on its own. What remains sellable is senior engineering that ships, augmented by AI, not replaced by it.

That is Abbeal's bet: senior engineering teams across three time zones (Paris · Montréal · Tokyo) that actually put AI into production. Want to talk? contact@abbeal.com.