The Jobs AI Will Replace in Professional Services

The jobs AI will replace are, for now, mostly not whole jobs but tasks inside them: drafting, summarising, first-pass research, routine design, data entry, scheduling and the clerical work that surrounds professional services. Whole occupations disappear only where those tasks were most of the job. Everywhere else the job changes shape, and the people who do it are expected to do different things.

That distinction is the difference between a useful plan and a frightening headline. A firm that asks which jobs will vanish tends to freeze. A firm that asks which tasks are moving, and what its people should do with the time, can act this quarter.

Tasks, not jobs

The International Labour Organization’s 2025 update of its global exposure index, built by scoring the individual tasks that make up each occupation, found that one in four workers worldwide is in an occupation with some exposure to generative AI. But only 3.3% of global employment falls in the highest-exposure category, where most of a job’s tasks could be automated. The ILO’s conclusion is that the more likely effect is transformation of jobs rather than their replacement, because an occupation is a bundle of tasks and only some of them move.

The World Economic Forum’s Future of Jobs Report 2025, drawing on more than 1,000 employers, points the same way at the level of the whole economy. It expects 170 million jobs to be created and 92 million displaced by 2030, a net gain of 78 million, driven by technology among other forces. And it expects the share of work tasks done mainly by people alone to fall from 47% today towards roughly a third by 2030, with the rest split between machines and people working together.

Neither report says there is nothing to worry about. Both say the change arrives task by task, which is exactly what makes it manageable for a firm that pays attention.

What the numbers do not tell you

Every figure in this debate comes with limits worth knowing before acting on it.

Exposure is not impact. The ILO is explicit that its index measures what tasks could technically be done with generative AI, an upper bound, not what employers will actually do given cost, skills and regulation. The WEF figures are employers’ expectations, not observed outcomes. And Stanford’s researchers describe their findings as early, descriptive indicators, “canaries in the coal mine”, rather than proof that AI caused the changes they measured.

None of that makes the evidence useless. It means the sensible response is to watch your own work closely rather than wait for a definitive study, because by the time one exists, the firms that moved early will already have reorganised.

Where the exposure sits in professional services

In law, accounting, consulting, marketing and similar firms, the most exposed work is easy to recognise.

• Clerical and administrative support. Scheduling, document handling, data entry, standard correspondence. The ILO identifies clerical work as the most exposed occupational group.
• Legal and professional secretaries. The WEF lists legal secretaries among the jobs just outside its ten fastest-declining roles, for the first time.
• Routine design and production. Graphic designers appear alongside them, a first in the report and a sign of how far generative tools now reach into creative production.
• Junior research and first drafts. The work that trained new professionals: summaries, benchmarks, first-pass analysis and draft documents.

The least exposed work is also recognisable: deciding, advising, negotiating, supervising, carrying responsibility for an outcome, and anything that depends on being trusted by a specific client.

The squeeze is at the entry level

The clearest early signal is not mass redundancy. It is fewer people being hired to start.

Stanford’s Digital Economy Lab, using payroll data covering millions of US workers through mid-2026, found no evidence of widespread, economy-wide job displacement. But employment of 22 to 25 year olds in the most AI-exposed occupations stands about 19% below where it would be had it kept pace with less-exposed peers, and the gap has widened since it was first reported in 2025. Experienced workers show no comparable gap. The adjustment runs mainly through reduced hiring, and it is concentrated where AI substitutes for tasks rather than complementing them.

For professional firms that is the most important finding in this whole debate. The routine work that is disappearing is the work juniors learned on. A firm that stops hiring juniors saves money for a few years and then finds it has no one ready to become a senior.

Automation or augmentation is partly a choice

The Stanford research makes a distinction worth acting on: where AI is used to substitute for tasks, entry-level employment falls; where it is used to complement people, employment is flat or rising.

Which one happens in a given firm is not entirely up to the technology. It is a design decision. The same tool can take a junior’s first draft away, or it can let that junior produce the draft faster and spend the saved time checking it against the sources and learning why the answer is right. The second choice costs a little more in the short term and builds the firm’s next generation of judgment.

What firms should do now

Five practical steps, none of which need a large budget.

1. Map tasks, not roles. For each role, list the recurring tasks and mark which are already inside what AI does well. Most roles turn out to be a mix.
2. Decide where the recovered time goes before it is recovered, or it will disappear into more of the same work.
3. Keep hiring at the entry level, but redesign the job around checking, client context and supervised judgment rather than raw production.
4. Say plainly what will change. People fill silence with the worst version. The change management that decides whether any of this works starts with honesty about roles.
5. Build the skills that stay valuable: framing problems, verifying output, advising and taking responsibility.

The consulting-specific version of this question is answered in will AI replace consultants, and the practical picture for firms is in AI for professional services firms.

How this silo fits together

This is the hub for the jobs question in professional services. Further briefings in the silo look at what happens to knowledge work itself when AI does much of the knowing, and at how firms can reorganise around that.

If you are redesigning roles around the work that stays with people, tell Daniel which roles you are looking at, what AI already does in them, and what the business needs from the change. Talk to Daniel.

Frequently Asked Questions

Which jobs will AI replace?

Mostly tasks rather than whole jobs: drafting, summarising, first-pass research, routine design, data entry and clerical work. The ILO’s 2025 index found one in four workers globally is in an occupation with some generative AI exposure, but only 3.3% are in the highest-exposure category where most tasks could be automated.

Will AI cause mass unemployment?

The current evidence does not show that. The World Economic Forum expects 170 million jobs created and 92 million displaced by 2030, a net gain of 78 million. Stanford’s Digital Economy Lab found no evidence of widespread economy-wide displacement in US payroll data, though it did find a clear squeeze on young workers in AI-exposed occupations.

Which professional services jobs are most at risk from AI?

Clerical and administrative support, legal and professional secretaries, routine design and production, and junior research and first-draft work. The least exposed roles involve deciding, advising, negotiating, supervising and carrying responsibility for outcomes.

Why are entry-level jobs most affected by AI?

Because the routine tasks AI now does well are the ones entry-level staff used to do. Stanford found employment of 22 to 25 year olds in the most AI-exposed occupations is about 19% below trend, mainly through reduced hiring, while experienced workers show no comparable gap.

Is it automation or augmentation?

Partly a design choice. Stanford’s research found entry-level employment falls where AI substitutes for tasks and holds or rises where it complements people. The same tool can remove a junior’s work or let that junior work faster and spend the saved time on checking and learning.

What should firms do about AI and jobs now?

Map tasks rather than roles, decide in advance where recovered time goes, keep hiring at entry level but redesign junior roles around checking and supervised judgment, communicate plainly about what will change, and build the skills that stay valuable: framing problems, verifying output, advising and taking responsibility.