Prompt and agent designers shape how AI assistants behave: the instructions they follow, the tools they may use, the checks that keep them sensible. It is a new craft, closer to careful writing and testing than to programming, so it suits people who are precise with language and honest about what works. Because the field is so young, evidence of real builds matters more than certificates. The hard part is constant change; what worked last month may not work next month. Curious, methodical people thrive.
You spend the morning testing an assistant against a list of tricky cases and noting where it goes wrong. Then you rewrite its instructions and run the tests again. In the afternoon you sit with the team who use it and watch how they actually work. You finish by writing up what changed and the evidence it improved.
The oddity of this job is that AI helps design itself: it can draft its own instructions and generate its own test cases. Your job is to set the goal, spot the failures it cannot see, and decide when it is good enough. Taste, scepticism and judgement stay human.
Most entrants have a degree or equivalent, though entry with other qualifications or significant relevant experience is possible.
Tick off steps, save the job and grow your own tree. Free, nothing captured.
AI operations specialist · AI support / enablement