Generative AI can speed up first drafts and routine tasks. It can also produce confident errors, disclose sensitive information and amplify weak processes. The useful question is not whether to use it, but how to use it with judgment.
Start with a bounded task
Choose work where a human can readily check the result: outlining a meeting agenda, turning notes into a first draft, suggesting questions for research or translating a non-sensitive internal document. Do not treat a model’s output as a source, a final decision or a substitute for specialist review.
Keep sensitive material out
Before pasting anything into a tool, pause. Client data, unpublished financial information, private credentials, health information and source material may be protected by policy, contract or law. Use only tools approved by your organisation and understand their data settings.
Build a verification habit
Check names, dates, quotations, calculations, citations and recommendations against primary sources. If an answer affects a person’s rights, money, health, safety or employment, a qualified human should own the decision.
Make the process visible
Tell collaborators when AI materially shaped a draft. Keep a record of the prompt, model and final edits when the work is consequential. That makes it easier to review errors and improve the workflow next time.
The takeaway: use AI to accelerate thinking and routine work—not to outsource accountability.
Further reading: NIST’s Generative AI Profile.