If AI Writes the Hard Message, Who Learns to Say It?

You need to tell someone on your team that their work needs to improve. You’ve put it off twice already, because you’ll still have to work together afterwards. So you open an AI chat window, explain the problem and ask for help. Within seconds, there it is. A calm, reasonable message, without the irritation that kept creeping into your version. You could send that. What a relief.

And perhaps it really is a better message. Your first attempt blamed the person for everything. This one describes what went wrong and asks for a change. If the tool helps you see that difference, you may have learnt something already. There’s no particular virtue in sending the angry version just because you wrote it yourself.

But let’s follow this a little further. You send the message. The reply comes back, and they think you’ve been unfair. Or there’s no reply at all. Tomorrow morning you’re both in the same meeting.

Where the tool stops

Now what?

The chat window helped you find the words. What did the other person hear?

Isabelle C. Hau offers a way of thinking about this in Stanford Social Innovation Review, a US magazine about social change. You can read her article, “Welcome to the Era of Relational Intelligence”. She proposes “relational intelligence”, or RQ, as a way to think about our ability to earn one another’s trust, deal with disagreements, mend relationships and work out what things mean together.

Hau argues that relationships involve getting things wrong with each other, waiting through awkward pauses and having needs the other person hasn’t met. Applied to an adult conversation, that raises an uncomfortable question. If we keep handing over the part we find difficult, where do we get the practice?

Her article doesn’t establish that AI drafting makes us worse at talking. It could help someone begin a conversation they would otherwise avoid altogether. But sending a well-worded message doesn’t tell us whether we’re getting better at what comes afterwards.

What should AI leave to us?

Hau recommends “augmentation over automation”. She suggests letting technology take care of things like summaries, scheduling and coordination, leaving people more time to listen and look after each other. That gives us something practical to try.

Let AI help sort the complaint into the point you actually need to make. Ask it where you sound accusatory. Then read what it gives you. Has it made the request clearer, or softened it so much that nobody could tell what you want?

An apology can have the same problem. You can spend ages making it sound considerate while quietly losing the sentence where you admit what you did. Could you say that sentence without looking at the screen?

Before sending, three questions might be enough:

• Could I say this aloud?
• Do I know what I will do if they react badly?
• Is the hard sentence in my own words?

You can’t predict every response. Is there room for one? Perhaps they’ll tell you something you missed. Perhaps your apology won’t be accepted straight away. Listening to that takes longer than writing the message did.

Where did the saved time go?

Hau’s “presence rebate” is a good prompt here. If AI saved you twenty minutes, where did that time go? Could some of it go towards a call, or simply being available after the message arrives?

Writing may be the right way to begin. Then the reply comes, or the silence, and the relationship still needs your attention.

Sources

• Isabelle C. Hau, Stanford Social Innovation Review