The time I asked AI to write an email and it was too formal

I once struggled with a reply to a friend's email and decided to ask an AI for help.

What it produced was a flawless, proper piece of writing.

But it was also incredibly humble, in a strange way.

"I apologize for taking up your busy time," and "I would appreciate it if you could confirm." For a business email, I think this is perfectly fine. But the recipient was a friend.

If I sent a reply like this, they would probably think, "Wait, what happened to you?"

The phrasing was also somewhat roundabout. It would use expressions like "it is possible to do." We don't say "it is possible to do" in everyday conversation.

In the end, I rewrote almost all of it myself. I could only use a tiny fraction of the expressions it provided.

AI produces the most polite text possible.

Why does it get so stiff?

I think it's because AI tries to produce text that won't be rude to anyone. It aims to be safe for a boss, a client, or someone you're meeting for the first time. Thinking about it that way, the safest option is "overly polite text."

I had asked for a "reply to a friend." But the AI doesn't know who the recipient is. Since it's writing without knowing that, it's only natural that the result would be generic and boring.

For formal situations, this stiffness is just right.

On the other hand, there are times when this "stiffness" is reliable.

When I want to write a message with seasonal greetings. When I'm thinking about a message for someone who is retiring. I can't easily come up with such formal writing on my own, so I use it as a reference to see "what kind of ways there are to write this."

This way of using it is very convenient.

In other words, it's not that the AI's writing is bad. It just didn't fit a "reply to a friend" as it was. That's all there was to it.

Tell it "who you are sending it to" beforehand.

What I do now is add a quick note when I make a request.

"This is a reply to a close friend. Don't be formal, keep it close to everyday spoken language."

Just by doing this, the text that comes out changes significantly. It's not perfect, but it starts returning something that makes you think, 'I can just fix it from here.' Just reducing the amount of editing makes things quite a bit easier.

Well then.

Even so, there are times when the feeling of 'this doesn't really sound like me' remains.

That is because just telling the AI who the recipient is doesn't mean it knows 'what kind of way I speak' yet.

I have compiled the methods for teaching the AI about 'yourself' in advance in a paid article. Once you create a 'memo of your own writing style,' you can reuse it for emails, LINE messages, and even note articles.

For now, just add a quick note about 'who you are sending it to' when you make a request. That alone is enough for what you can do today.

You should leave things you don't need to remember to machines. However, I believe that only I can decide who I want to reach and how.

If you enjoyed reading to the end, I would be happy if you could click the ♡ button.

Author: Sate. | Digital Utilization Supporter
I have spent 30 years at an IT company creating 'easy-to-understand manuals.'
Together, we will turn the 'I don't understand IT' of those in their 50s and beyond into 'Oh, I can actually do this.'