The question everyone asks is "which jobs?" - but that is the wrong unit. AI, and agentic AI especially, automates tasks, and most jobs are a bundle of many tasks. Understanding which tasks go first tells you far more than any list of "doomed professions."
The tasks that go first
- Repetitive and rules-based - data entry, form-filling, routine reconciliation
- Language-heavy and templated - standard emails, first-draft reports, summaries
- High-volume triage - sorting, routing and answering common enquiries
- Information-gathering - pulling facts from many sources into one place
If a task is predictable, frequent and checkable, it is a candidate. That cuts across every industry - the specifics differ, the pattern does not.
The tasks that stay human longest
Judgement calls, relationship-building, creative problem-solving, and anything where being wrong is expensive. These are the skills AI will never replace - not because AI cannot attempt them, but because you would not want it to without a person in charge.
What this means for real roles
Very few whole jobs vanish; many change shape. The person who spent 40% of their week on automatable tasks now spends that time on higher-value work - if you help them make the shift. That is why training your team to work with AI agents matters as much as the technology.
How to assess your own industry
List the main tasks in a role, mark each as predictable-and-frequent or judgement-and-varied, and you have a rough exposure map. It is cruder than a consultant's report and usually just as useful.
The bottom line
AI automates tasks first, jobs rarely - so map the tasks, not the titles, and redeploy the time you free up. Learning to read that exposure and lead the transition is central to The Future of Workforce course at London School of Business UK. Enquire today.