AI Can Help Write the Email. It Cannot Decide How It Should Land.

You have a hard message to send. Maybe it’s feedback that’s overdue. Maybe it’s a “no” to someone who’s going to be disappointed. Maybe it’s a follow-up after a conversation that didn’t go well.

So you open an AI tool, describe the situation, and thirty seconds later you have a clean, professional, well-organized draft.

It’s technically perfect. And something about it is wrong.

What AI is genuinely good at

Let’s be fair. AI tools can be very helpful with workplace writing:

• Getting you past the blank page.
• Organizing scattered thoughts into a structure.
• Tightening something that rambles.
• Suggesting a softer or firmer way to say a sentence.
• Checking whether you’ve covered the key points.

Using AI for any of that isn’t cutting corners. It’s using a tool. The question isn’t whether to use it. It’s what you hand over when you do.

What it can’t know

The tool doesn’t know that this person had a rough month. It doesn’t know that the last time you gave them feedback by email, it went badly. It doesn’t know they’re anxious about their review, or that they respond better to directness than to cushioning, or that “I wanted to touch base” is a phrase they’ve learned to dread.

It doesn’t know the relationship. You do.

That’s the gap. AI can help you write the words. It can’t decide how those words should land, because landing depends on the person reading them and the history between the two of you.

AI can make the message clearer. Only you can make it right for this person.

The perfect email problem

There’s a particular kind of AI-written message that reads beautifully and still misses. It’s polished, balanced, and uses all the right phrases. It sounds like an HR policy wrote a letter.

The person reading it can usually tell. And when a hard message sounds like it came from a template, people tend to hear “this wasn’t worth your real attention,” even if that’s not what you meant at all.

The Human Check

Before you send anything AI helped you write, especially anything about performance, feedback, or a sensitive situation, run it through a quick Human Check:

1. Does this sound like me? If they’d be surprised you wrote it, rewrite it.
2. Would I say this out loud, to their face? If not, why is it in writing?
3. What will they feel when they read the first line? That line sets the tone for everything else.
4. Is everything in here accurate and specific? AI fills gaps with confident-sounding generalities. Cut anything you can’t back up.
5. Should this be an email at all? Some messages need to be a conversation first and an email second.

That last one matters more than it looks. AI makes writing so easy that it can quietly talk us out of the conversation. If the message is hard, the human move is often to talk first and use the email to confirm what you discussed.

Be thoughtful about what you share

One practical note. Before you paste details about a specific employee into any AI tool, check what your organization allows. Performance details, health information, and anything related to a formal process deserve extra care. If you’re not sure, describe the situation in general terms, and keep names and identifying details out. When something is headed toward formal action, the documentation should go through the people and process your organization uses for that, not just a chat window.

Adapt it

• If you write quickly and directly: AI polish can be useful. Letting it sand off all of your clarity isn’t.
• If you tend to over-soften: A gentler draft can be useful. Burying the actual message in paragraph three isn’t.
• If they process by thinking it through: A written follow-up is useful. Using it in place of a conversation they needed isn’t.

Ask one more question

Before you hit send, ask yourself: If this person forwarded this message to someone they trust, would I be comfortable with how it reads?

That question tends to catch the lines that are technically fine and humanly off.

Try this

Next time you use AI to help with a people-related message:

1. Write two or three sentences yourself first, in your own words, about what you actually want them to understand.
2. Then use AI to help shape it, not to start it.
3. Run the Human Check.
4. Read it out loud once before you send it.

You’ll still save time. You’ll just spend it on the part that matters.