AI draft: you check facts, tone, and mandate fit

When you use AI to draft an official message, the tool is just a starting point. You remain fully accountable for every word that leaves your inbox. Unlike a human colleague who might flag a sensitive detail, AI has no awareness of context, policy, or legal boundaries. For example, it might produce a perfectly polite reply that accidentally includes a colleague's personal data or a tone that sounds too casual for a formal notice. Your job is to catch those mismatches before clicking send.

To avoid common pitfalls, run a quick mental checklist as you review. First, check facts: does the draft contain any numbers, dates, or names you did not provide? AI can hallucinate plausible-sounding details. Second, check tone: read the message aloud to see if it sounds natural for your audience and your role. Third, check confidentiality and personal data: scan for any information that should not be shared, such as internal references or identifiable details. Finally, check mandate fit: does the message stay within your authority and the purpose of your role? If the draft suggests a commitment you cannot make, rewrite it.

A quick way to test your review is to ask yourself: 'If this message were forwarded to my manager, would they approve it without changes?' If the answer is no, you have more work to do. Another trick is to mentally swap the recipient with a senior official or a member of the public—would the tone and content still be appropriate? That shift in perspective often reveals hidden issues. Remember, the sender is always accountable, not the AI.