For decades, the single most stable white-collar job in elite professional services was the partner’s executive assistant. Six-figure pay at the top of the band. No elite degree required. A year-end bonus pegged to the partner’s bonus. And — the part that gets quoted in every retirement speech — total job security, because the partner depended on the EA in a way no LinkedIn job-architecture review could explain.
Bloomberg’s May 21 piece named the bodies. PwC’s US arm let go of about 600 executive assistants, recruiters and other support staff in February, with affected workers receiving four to six weeks of severance. McKinsey cut roughly 200 tech and support staff late last year — less than half a percent of its 40,000-person workforce, which is exactly the size of cut you make when you want the direction to be legible without the headcount being newsworthy. KPMG is offshoring similar roles as part of a cost program. EY is in the same camp.
The reason every press shop is giving is the same reason: AI assistants have been rolled out to internal staff at all four firms over the last 18 months, and back-office headcount has been quietly shrinking the whole time.
The role that AI tools chew through fastest
The technical case is uncomfortably clean. Executive assistants at the top consulting firms book travel, run calendars, and handle expenses. The work is repetitive, rules-based, and substantially mediated through a small number of enterprise applications (Concur, Outlook, Workday, Travelport). That is the exact profile that current LLM-powered agents handle without prompting tricks. The pay ceiling — comp at the top of the band can clear $100,000 with bonuses — makes the substitution math close even at first-generation agent unit costs.
The Big Four firms publish their own evidence on this. McKinsey’s own research has reported that AI can shave more than 30% off the time it takes to do research and synthesis on a typical project. That math is the firm’s own pitch to its clients. The same math applied internally is what produces the 200-person and 600-person numbers.
A McKinsey spokesperson, asked about the cut, described it as making support functions “more efficient and effective, including by taking advantage of AI.” Bob Sternfels, McKinsey’s global managing partner, has separately said that non-client roles will see more cuts over the next two years. The two statements together are a sentence with no surprises in it: the firm telling its clients to adopt AI is the same firm adopting AI, and the first heads to roll are the heads closest to the rollout.
The detail Bloomberg buried
The detail the press coverage soft-pedals is the one that makes the EA job uniquely exposed inside the partnership model. EAs at the Big Four are not lawyers or auditors or consultants — they are not on the billable-hour ledger. They sit on the partnership P&L as pure overhead. A senior associate’s hour shows up as revenue. An EA’s hour shows up as cost. Every other professional category at McKinsey or PwC has a built-in defense (“we charged the client for this”), and the EA category does not. When the same partner who is being asked to demonstrate AI productivity to the firm’s clients is looking for a number to subtract from their own cost line, the EA budget is the easiest line in the spreadsheet to touch.
This is also why the cuts are coming first to the US arms of these firms and not their global headcount. US EA comp is the highest in the network. The Manila, Hyderabad, and Krakow shared-service centers — where KPMG and EY are quietly migrating the role — pay a third of that. The first move is “do we still need a domestic EA at all?” The second move is “what subset of the offshore EA function can the agent now do without a human?” The first move closes in 2026. The second closes by 2028.
The 300-million number, again
The frame Bloomberg uses to bracket this is the World Economic Forum’s latest jobs forecast, which pegs the count of white-collar roles AI could reshape at nearly 300 million over the next five years — with roughly a third of those at risk of going away entirely. The number is large enough that it usually gets read as background scenery. What Bloomberg’s piece does is name a job category from inside that 300 million, give it a date, and put PwC’s name on a 600-person number. That is the part the forecast does not normally do.
The other detail worth reading carefully: PwC’s 2026 survey of more than 56,000 workers found that the people leaning into AI were the ones feeling safer in their jobs — 70% of active AI users said the tools made them feel more secure, while the AI-avoiders reported the opposite. The same survey was published by the same firm that just laid off the population least likely to be AI users. The survey is real. The conclusion the firm wants drawn from it (“use AI to feel safer”) is also real. The third sentence — use AI to make yourself the worker the firm keeps rather than the worker the firm replaces — is the one PwC will let its 56,000 respondents reach on their own.
What to watch
- Whether the EA cut moves up the pyramid. Partners cutting back-office staff is the easy part of the AI-restructuring playbook because the work is the most legible to current agents. The harder question — whether the same partners start to thin out their junior consultant ranks — is the one Bloomberg flags at the end of its piece. That is where the billable-hour model starts to bend, and the Big-Four EA cut is the early warning, not the event.
- The shared-service migrations. KPMG and EY moving roles to Manila / Hyderabad / Krakow is the second tell. Watch the Q2 + Q3 earnings calls for any line item naming “support-function offshore migration” — those are 18-to-24-month projects that show up in restructuring charges before they show up in headline headcount.
- The mid-2026 Big-Four AI partnership announcements. Each of the four firms has a public partnership with OpenAI, Anthropic, or Microsoft Copilot. The press releases say “for client work.” The internal usage data — which never gets disclosed — is what is driving the EA cut. The next set of partnership renewals (PwC’s OpenAI deal, KPMG’s Microsoft alliance) is what to read.
- Whether the WEF 300M number stops being read as a forecast. It was published as one. The PwC + McKinsey February-and-late-2025 cuts are not.