The forwarded email every HR leader eventually gets
A board member forwards a McKinsey article at 6:45 a.m. with one line attached: "Are we ready for this?" You open it, skim the headline number, and start doing the math in your head — what does a global estimate about automatable work hours actually mean for your 380-person operations and finance teams by next quarter?
This is the moment most HR Strategy Directors and People Analytics leads hit sooner or later. McKinsey's research on generative AI and the future of work is genuinely useful — it is one of the more rigorous, widely cited bodies of work on how AI could reshape labor markets. But it was built to describe the U.S. or global economy, not your org chart. This article walks through what the McKinsey research says, what it deliberately does not say, and how to turn a economy-level headline into something you can actually put in front of your board.
What the McKinsey research actually says
McKinsey's 2023 analysis on the economic potential of generative AI estimated that 60–70% of employees' work hours could theoretically be automated with current and anticipated technology — up from roughly 50% in pre-generative-AI estimates (McKinsey, 2023, "The economic potential of generative AI: The next productivity frontier"). That single figure is the one most often clipped into slide decks and forwarded emails, and it's worth sitting with the wording: could theoretically be automated. That is a technical-feasibility estimate, not a timeline, and not a company-specific forecast.
McKinsey's Global Institute work from the same period also modeled how the overall employment mix is likely to shift by 2030 — toward healthcare, STEM, and managerial occupations, and away from customer service, office support, and food service roles (McKinsey Global Institute, 2023, "Generative AI and the future of work in America"). That's a directional signal about which broad occupational categories are expanding or contracting across the economy — again, not a statement about which roles in your finance or customer operations function are affected, or when.
Separately, McKinsey's 2024 State of AI survey found that 65% of organizations report regularly using generative AI, roughly double the share reported ten months earlier (up from about 33%) (McKinsey, 2024, "The state of AI in 2024"). That number tells you adoption is accelerating fast at the organizational level — it does not tell you which tasks inside your organization are actually being reallocated as a result.
The gap between "could" and "will" — and why it matters for HR
The single most important thing to understand about the 60–70% figure is that "theoretically automatable" and "will be automated in your company this year" are not the same claim. McKinsey's own methodology treats technical feasibility as one input among several — cost, regulation, labor supply, and organizational adoption speed all bend the actual timeline. This is the same distinction our own research summaries draw when covering Goldman Sachs's global exposure estimates: a large modeled number describes potential, not a schedule.
This distinction matters practically because it changes what a board conversation should sound like. "60–70% of work hours are automatable" is not an actionable HR statement. "Here are the departments in our company where task-level exposure is concentrated, and here's what we're doing about the highest-exposure roles" is. The former is McKinsey's job. The latter is yours.
Which jobs and sectors McKinsey flags — and where that stops being useful
McKinsey's employment-mix modeling is genuinely informative at the sector level: it points toward healthcare, STEM, and managerial work expanding relative to customer service, office support, and food service roles by 2030 (McKinsey Global Institute, 2023). If your company sits heavily in one of the contracting categories — a large customer support or back-office function, for example — that's a legitimate reason to look closer.
But "look closer" is the key phrase. Sector-level categories like "office support" bundle together administrative coordinators, executive assistants, data-entry specialists, and dozens of other O*NET-classified occupations with very different task compositions. Some of those roles are dense with routine cognitive work; others carry substantial judgment, coordination, and relationship management that doesn't compress into a sector average. For a more granular look at how exposure varies by occupation rather than by broad category, see which jobs are most exposed to AI — the occupational detail McKinsey's sector-level modeling isn't built to provide.
Where the economy-level lens stops, structurally
This isn't a criticism of McKinsey's methodology — it's a structural fact about what any macro research product is built to do. McKinsey, like Goldman Sachs, the World Economic Forum, and Brookings, models national or global labor markets using aggregated occupational and task data. That's the right tool for informing macro strategy, public policy, and long-range workforce planning at the economy level. It is the wrong tool for answering a board member's actual question, which is almost always some version of: which roles in our company, specifically, carry the most exposure, and what should we do about the ones that do?
Macro research answers "how big is this trend." It cannot answer "what does this mean for our 340-person engineering and operations org." Those require mapping your actual role list to detailed occupational task data — which is a different kind of exercise entirely. For a broader look at how this generation of research fits together, how AI is changing the future of work surveys the landscape these reports collectively describe.
Translating the research into a company-specific, board-ready answer
The bridge between "McKinsey says 60–70%" and "here's our department heatmap" runs through the U.S. Department of Labor's ONET database — the same public occupational taxonomy that underlies most of this macro research in the first place. ONET currently classifies 1,016 occupational titles across 923 data-level occupations, covering more than 55,000 job titles, each described by roughly 277 descriptors updated on a regular cycle (O*NET Resource Center, USDOL/ETA, 2019; U.S. DOL, 2025).
That structure is what allows exposure analysis to move from "sector" to "role." Our own methodology maps each role in a company to its nearest O*NET occupation, then scores each associated task against four dimensions — cognitive routine, physical routine, social/judgment, and creative — each on a 0–100 scale, weighted by the task's documented importance to that occupation.
A quick worked example: imagine a role whose O*NET-mapped tasks score 70 on cognitive routine, 10 on physical routine, 45 on social/judgment, and 20 on creative, with cognitive-routine tasks carrying the heaviest importance weighting for that occupation. The importance-weighted blend produces a composite exposure score — not a verdict on the role's future, but an input that routes it into one of three categories: Monitor, Review, or Redeployment Candidate. A high score flags a role for closer human evaluation of task allocation, training investment, or restructuring options — it does not predict a layoff, and it never states that a role "will be automated."
This is the level of specificity a board actually needs, and it's the level neither McKinsey's sector modeling, Goldman's 300-million-job estimate, nor any other economy-level report is designed to produce. For a discussion of how percentage-based global estimates translate — or fail to translate — into workforce planning at the company level, see how many jobs will AI replace by 2030.
Where to go next
Macro research like McKinsey's earns its place at the start of a board deck: it establishes that the trend is real, large, and moving faster than most workforce plans were built to handle. Just don't let it be the whole deck. The next slide needs your own occupational data.
If you're building that next slide, our Workforce AI Readiness Assessment Guide walks through how to structure a company-specific exposure review using the same O*NET-grounded approach described above, and our pricing page outlines the self-serve tiers built for mid-market HR teams doing this work without a six-figure consulting engagement. For ongoing, plain-language breakdowns of new macro research as it lands — McKinsey, Goldman, WEF, and beyond — subscribe to our research briefing on the blog.
This site incorporates information from ONET. Used under the CC BY 4.0 license. ONET is a trademark of USDOL/ETA.