Business leaders have spent the past two years asking “which jobs will AI replace.”
That question was always slightly the wrong one. Jobs are bundles of skills, and different skills within the same role carry very different exposure to automation.
McKinsey Global Institute’s new report, Agents, robots, and us: How AI reshapes work and skills in Europe, gives that intuition a rigorous, quantified footing, and it is directly relevant to how we structure workforce transformation at Teraflow.
The model, briefly
McKinsey’s Skill Change Index (SCI) is a time weighted measure of how exposed a given skill is to automation by 2030.
It draws on roughly 10,500 skills, mapped against about 800 occupations and roughly 2,000 detailed work activities, across ten European economies (Czech Republic, Denmark, France, Germany, Italy, Netherlands, Poland, Spain, Sweden, and the UK) that together represent more than three quarters of the region’s labour force and GDP (McKinsey Global Institute, May 2026).
The methodology classifies each skill by the automatability of the work activities it supports, using generative AI-assisted mapping validated against a manually built template, then weights the result by projected 2030 automation adoption in the midpoint scenario.
The headline context: 58 percent of current European work hours are technically automatable with existing technology, 44 percent by agents handling cognitive and administrative work, 14 percent by robots handling physical work. McKinsey estimates this could unlock up to $1.9 trillion in economic value across the ten countries by 2030 in a midpoint adoption scenario, against a more conservative $1.1 trillion if adoption stays gradual (McKinsey Global Institute, 2026).
Agents alone account for roughly 82 percent of that potential value, reflecting how much faster software based automation scales relative to capital intensive robotics.
The 15 skills, classified by exposure
The report’s flagship exhibit plots roughly 10,500 skills along the SCI curve and calls out a sample of named skills at each point.
We have grouped the 15 named skills into three risk bands based on their position on that curve, low risk sitting near the bottom of the distribution (roughly the 10th to 20th percentile), medium risk occupying the crowded midsection (roughly the 25th to 60th percentile), and high risk sitting at the steep upper end (roughly the 75th percentile and above, approaching the observed maximum of around 40 on the 0 to 100 scale).
| Risk band | Skill | What it means for planning |
|---|---|---|
| Low | Resilience | Least exposed. Central to change management and role redesign work itself. |
| Low | Empathy | Least exposed. Core to client relationship and people leadership roles. |
| Low | Leadership | Least exposed. Reinforces why leadership capacity, not headcount, is the binding constraint on adoption. |
| Low | Influencing skills | Least exposed. Underpins internal change adoption and stakeholder alignment. |
| Medium | Innovation | Shared skill. Increasingly exercised in partnership with agent-generated options. |
| Medium | Collaboration | Shared skill. Shifts toward orchestrating human and agent contributors together. |
| Medium | Detail orientation | Shared skill. Migrates from manual checking toward reviewing agent output. |
| Medium | Problem solving | Shared skill. Retained as judgement layer above automated first-pass analysis. |
| Medium | Research | Shared skill. Automatable at the retrieval stage, retained at the synthesis and interpretation stage. |
| Medium | Analytical skills | Shared skill. Directly relevant to how BI and reporting roles will be restructured. |
| High | Quality control | Highly exposed. Common in manufacturing and service operations; already embedding into agent-led workflows. |
| High | Software development | Highly exposed. Notably, more exposed than several client-facing and analytical roles. |
| High | Invoicing | Highly exposed. A canonical example of a structured, codified financial operations task. |
| High | Accounting | Highly exposed. Consistently among the most exposed skills across every country dashboard in the report. |
| High | SQL (programming language) | Highly exposed. Sits at or near the top of the full 10,500-skill distribution. |
Two patterns are worth flagging for planning purposes.
First, the low risk band is composed entirely of interpersonal and judgement capabilities, none of them technical. Second, software development appears in the high risk band alongside accounting and invoicing rather than among the safer analytical or research skills.
McKinsey’s broader skill taxonomy explains why: digital and information processing skills, including programming languages and routine data entry, are consistently among the fastest changing and most exposed categories in the analysis, while skills rooted in leadership, communication, and empathy are consistently the least exposed (McKinsey Global Institute, 2026).
Why the middle band is the actual story
The distribution behind this chart matters more than the named examples on it.
Across the roughly 10,500 skills analysed, McKinsey finds that 75 percent are used in work activities that mix automatable and non-automatable tasks. Only 15 percent are mainly tied to automatable activity, and only 10 percent are mainly tied to activity that resists automation entirely (McKinsey Global Institute, 2026).
This is the finding we lead with in client conversations, because it directly informs how a workforce transformation programme should be scoped. The dominant pattern is not replacement, it is recombination. Most roles will retain the skill but change the ratio of time spent executing it directly versus directing, reviewing, and correcting an automated system that executes it first. That distinction is the difference between a redundancy programme and a redesign programme, and getting it wrong in either direction is expensive.
The demand signal enterprises are already responding to
Job postings data confirms the shift is under way, not theoretical. Demand for AI fluency, the practical ability to use, manage, and integrate AI tools into daily work rather than build them, has increased fivefold across the ten European economies since Q4 2023 and now appears in postings across occupations representing roughly 5 percent of employment.
Demand for technical AI skills, the capability to design, build, and govern AI systems, grew more modestly at 1.7 times over the same period. Seventy five percent of that AI skills demand is currently concentrated in just three occupation groups, computer and mathematical, management, and business and financial operations, which together account for only around a fifth of total employment.
The remainder is spreading into logistics, HR, compliance, and skilled trades, evidence that this is now a workforce wide expectation rather than a technical specialism.
Where this connects to Phase Zero
This is precisely the diagnostic gap our Phase Zero methodology is built to close before any automation investment is made.
A skills level exposure map, run against a client’s actual role architecture rather than a generic occupational proxy, tells you which of the 75 percent “shared” skills in a given function are closer to the automatable edge of that band and which are closer to the judgement edge. That distinction determines whether the right intervention is workflow redesign, retraining toward the judgement layer, or straightforward task automation, and it is not visible from headcount or job title data alone.
Under our DAPA framework, this maps directly onto the Discover and Assess stages: discovering which skills genuinely sit inside a given role’s activity mix, and assessing their real exposure rather than assuming exposure by job family.
Organisations that skip this step and automate at the role level, rather than the skill level, tend to either over-invest in automating already low-value work or under-prepare the people whose roles sit in the 75 percent shared band, which is most of the workforce.
The practical takeaway
McKinsey’s own conclusion is understated and correct: capturing the value on offer depends less on new technological breakthroughs and more on how organisations redesign workflows and how quickly human skills adapt around them (McKinsey Global Institute, 2026). The Skill Change Index gives leaders a more precise instrument than “will this role survive.”
The better question, and the one worth answering before any automation programme is scoped, is which specific skills inside each role sit in the shared band, and what the organisation intends to do with the time those skills free up once agents and robots take the first pass.
