I'm in charge of AI promotion, but I write my own emails. Here's why.

Hello, this is Ponzu.

The other day, while I was exchanging emails with an AI vendor, I had a sudden thought.

I wondered, 'Why am I, someone in a position to promote AI, writing my own emails?'

Today, I'd like to write about what I thought regarding that question.

The day the email and the attachment had different personalities

It started with an email I received from a representative at an AI vendor.

The body of the email said, 'I have three proposals,' but the attachment only listed two. Moreover, while the email text was extremely polite, the tone of the attachment was so casual it felt like it was written by a completely different person.

Honestly, I thought, 'Is this email mentally unstable?' Sorry about that lol.

They probably generated the email body with AI and created the attachment at a different time. When I gently checked with them, they apologized. I don't think they meant any harm.

However, this interaction made me realize that 'when you leave writing entirely to AI, these kinds of discrepancies occur.'

Emails aren't just about the body text, are they? I believe a single email includes the consistency with attachments, the alignment with the subject line, and the context of previous exchanges. It's all about that 'connection.'

It was an event that made me think that, in the end, the only one who can maintain that 'connection' is myself, as I understand the context.

Of course, I'm sure some would say, 'Why not just throw the email body and the attachments all together into the AI?'

It's true that some current AIs can handle a context of 1 million tokens. Technically, it is possible to throw in past exchanges and attachments all at once to generate a consistent email.

But the question is, are you going to do that every single time?

I write many emails a day. Feeding past threads and attachments into the AI each time, checking the output, making fine adjustments... Honestly, that is extremely inefficient.

Besides, AI also has usage limits. ChatGPT, Claude, and Gemini all have limits if you use them too much. There's also the issue of whether you want to allocate those resources to emails.

I want to use AI resources for more 'thinking work.' I believe it's faster and more reliable to write emails with my own hands.

The reason I don't leave it to AI was 'brain training'

Triggered by the vendor's email incident, I took another look at how I write my own emails.

That's when I realized that writing emails myself is actually a form of brain training.

I don't have any scientific evidence for this. It's just my own personal experience.

But when you construct sentences yourself, you're processing many things in your head at the same time. You imagine the other person's perspective, think about the order of information, and choose expressions that won't be misunderstood. If I outsource this entire 'thinking process' to AI, I worry that my own ability to think will gradually decline.

That's what I believe.

It might be similar to strength training. If you leave everything to machines and don't move your body, your muscles will atrophy. I think thinking about email content is 'mental strength training' for me.

The topic of 'intervening value' as seen by someone in their 40s

There is another reason why I write my own emails.

It's a question of where I place my own 'intervening value'.

When I think about what I can do that only I can do in the age of AI, honestly, the word 'originality' feels a bit heavy for someone in their 40s. The younger generation might be better at competing with creative ideas.

However, I believe that the trust built on the idea that 'it resonates because this person wrote it' or 'I read it because it's this person's email' can only be created through years of work.

The nuances based on the relationship with the other party. The context read from past interactions. The judgment of 'it's better to write this carefully' or 'it's better to keep this short to get the point across'.

These parts are still difficult for AI. Or rather, I believe this is an area I should handle myself.

I use AI as an excellent reviewer. But I write the first sentence myself. This order feels right to me.

A story from my previous job where I was recognized as 'highly capable' through email

I've written some high-and-mighty things, but there is a reason why I've become so particular about emails.

When I joined my previous company as an IT staff member at a factory, as a mid-career hire, I wasn't exactly welcomed. People ignored me when I greeted them. They gave me curt replies when I asked questions. I remember thinking, 'Oh, this is going to be pretty tough.'

But as IT staff, you have many opportunities to send emails to all employees.

• System maintenance notifications

• Security alerts

• New tool announcements

I used to write company-wide emails almost every week. During that time, I developed my own personal standards. Since I'm at it, I'd like to introduce the 11 Rules for My Emails that I still follow today.

1. Use 'sama' instead of 'sama' (kanji) after the recipient's name.
Since company names, departments, and titles are already in kanji, using hiragana adds a sense of lightness.

2. Keep each sentence under 60 characters. Since I don't know what device they are reading on, I want to prevent awkward line breaks.

3. Do not break lines in the middle of a sentence.
The reason is the same as #2: readability is the top priority.

4. State 'I have two things to tell you today' at the beginning of the email.
This helps the recipient prepare mentally.

5. Use at least one bulleted list. Emails consisting only of paragraphs have a high risk of being skipped.

6. Do not make the email itself long.
If it gets long, attach a separate text file. The premise is that emails aren't read in the first place.

7. Do not use the same sentence ending more than twice.
If you repeat '...is. ...is. ...is.' three times, it immediately looks amateurish.

8. Do not string more than five kanji characters together.
It significantly reduces readability.

9. Always include your own email address at the end of the CC.
This is to eliminate the excuse of 'I didn't receive it.'

10. Insert a line break roughly every three lines.
An email without white space makes people lose the desire to read it.

11. Use only bold and red for emphasis.
Using too many colors makes it look cluttered and actually makes it harder to convey what is important.

As I continued to write emails carefully like this, the reactions around me gradually changed. People started saying, 'Ponzu-san's emails are easy to understand,' and before I knew it, I was labeled as 'highly competent.' People who had ignored me so much started coming to me for advice as if nothing had happened.

Well, I remember everything, though.

However, because of this experience, I cannot cut corners on email writing. I know from personal experience that email is not just a means of communication, but a tool for building trust.

What you get from AI review

After I write an email myself, I ask AI to review it. I have it check for typos, omissions, and Japanese phrasing. It's a simple request.

You might think, 'Is that all?' but it is surprisingly effective.

A common piece of advice for improving writing skills is to 'have someone read your writing.' I think this is advice you often see in books and blogs.

I believe this also applies to AI.

Even when I write an email thinking it's perfect, running it through AI often reveals twisted phrasing or mismatches between subjects and predicates. The moment I realize, 'Oh, so my writing has these quirks,' it feels just like having someone else read it.

However, there is one thing I want you to be prepared for.

AI is not equipped with a 'consideration' function.

AI will mercilessly point out things that a human colleague might just let slide with a 'Well, I guess that's fine.' Plus, its comments are so mechanical. Honestly, it can be a bit demoralizing at first.

I've thought, 'I know I asked you for help, but couldn't you have phrased that a bit better?' more than once or twice.

But you gradually get used to it. You start to notice patterns in the feedback and realize, 'Ah, this is where I'm weak.' Once that happens, you start being able to fix those things yourself before even running the review.

Having AI review your email drafts helps you gain both writing skills and a heart of steel. That's a pretty good deal, right?

What is your 'domain for writing yourself'?

Some of you reading this might be thinking, 'Wait, isn't it fine to just leave emails to AI?'

I think that's a valid answer, too.

I believe the important thing is to decide for yourself what to leave to AI and what to do on your own.

In my case, it just happens to be emails. Because I have a history of building trust through emails, I want to keep this as my own domain. That's my decision.

I'm sure you all have a domain where you think, 'I want to do this myself.'

Maybe it's the structure of a project proposal. Maybe it's your pitch to a client. Maybe it's how you write meeting minutes.

The hint to finding that domain is surprisingly simple: it's the work that makes you feel uneasy if you leave it to someone else. I believe that sense of unease is exactly where your unique value lies.

Rather than thinking about what to leave to AI, think about what you want to keep for yourself. It might be interesting to try thinking about it from that perspective first.

Summary

The reason why I, as an AI promotion lead, write my own emails.

It's because I believe that writing emails is a form of mental training, a means to build trust, and an act of protecting my own unique value.

I ask the AI to review what I've written afterward. Checking for typos, omissions, and Japanese phrasing—that's the AI's specialty, so I rely on it honestly.

"I write it myself, then have the AI polish it." This order is the perfect balance for me right now.

Well then, if you'll excuse me.