Will AI take my job? Audit the tasks you actually do
To assess your own role, start with the tasks in a normal working week. A broad occupation label or alarming prediction is not enough. Identify which tasks a tool can assist, who checks the result and which decisions still depend on context you hold.
Make a task inventory
List recurring outputs, the information used, the people involved and the consequence of mistakes. Distinguish routine preparation from judgement and accountable decisions. Mark where AI is already used in your workplace, rather than assuming all available technology has been adopted.
Compare a task before and after assistance
Illustrative example: drafting a customer reply may become faster, while resolving an unusual dispute still needs policy knowledge, judgement and permission. Investigate the whole workflow, including checks and corrections. Fast production of a draft is not the same as safe completion of the customer’s request.
Look for evidence of changed expectations
Read recent relevant vacancies and ask which responsibilities your manager expects to grow or shrink. Separate a rumour, a pilot and an implemented change. Keep sensitive employer information out of public AI tools. Follow workplace policy before attempting any experiment with real data.
Choose one response you can demonstrate
Build evidence of checking outputs, handling exceptions or making a relevant decision. If your work is shrinking, investigate adjacent roles that use your domain knowledge before committing to a full reset. This audit helps frame uncertainty; it cannot predict whether your individual job will be retained.
Review one recurring output with your manager
Choose a recurring output and write the steps required to deliver it accurately. Identify where a tool could assist and where you currently provide context or judgement. If appropriate, discuss how expectations for that output are changing and which responsibilities remain important to the team.
Record the answer as an observation, including any uncertainty. A manager may not know every future change, so do not treat one conversation as a guarantee of job security. Use the information to investigate a specific capability, such as checking generated work or handling exceptions. A task-level discussion is more actionable than a general question about whether AI will eliminate your entire profession.
Choose the next guide for your question
Sources and further reading
Editorial review: 13 October 2026. Examples are illustrative; results depend on the information supplied.