The article I wrote with AI wasn't bad. But I revised it 5 times.
The article I made with AI through a Q&A process wasn't bad. Even so, it took five revisions. To ensure my own words and experiences are retained right from the first draft, I am reconsidering how I write and ask questions.
The other day, based on a questionnaire sent to candidates for the Fukuoka mayoral election, I wrote an article about what I hope to see for this city.
Answering questions from an AI one by one, while also having it refer to articles I had written in the past, I turned my thoughts into writing.
Looking at the finished article, I think it turned out quite well.
The words I wanted to convey are still there. It has become easy to read. What I thought about this city through my own experiences is clearly written.
This turned into an article that only I could write.
I published it because I was convinced of that.
However, getting to that point took five rounds of revisions.
The past articles I shared this time, and my own words spoken in the Q&A.
In the next article, I wonder if the AI can take advantage of what those reveal—what I prioritize and how I convey ideas—right from the very first draft.
Here, I would like to write about what worked well when creating writing with AI, and what I want to improve next time.
My words remained intact
In the mayoral election article, there were words I really wanted to keep.
Does someone who wants to become mayor have no intention of listening to the words of Fukuoka’s citizens?
Do they have no intention of communicating?
I was surprised and disappointed that there was no response to the questionnaire. Those were the words that came out of me at that moment.
Those words remained in the article, just as they were.
And by breaking them into short sentences and inserting line breaks, it became easier to read, leaving a naturally strong impression of what I wanted to communicate.
Where to cut a sentence.
Where to pause for a moment.
I believe those choices are also an essential part of my writing.
The reason why I am the one to say those words
There was another phrase I wanted to keep.
In the first place, whether the person themselves is willing to learn is also a major factor.
I wanted to connect my own experiences to these words.
Starting a company from a completely different industry
Before COVID-19, I was a trumpet player with zero connection to IT.
From there, I founded a company in an entirely different industry, starting with no acquaintances at all, and struggled desperately to reach where I am today.
Continuing to give lightning talks (LTs) every single week was one step in that process.
I have also written in a past article about how I went from being a musician to starting a business, and why I do not run music-related businesses.
Behind my statement that “whether the person themselves is willing to learn is also a major factor” lies this kind of background.
By connecting my past articles with my current opinions, I can convey the path I took to be talking about this now.
I believe that making visible the connections in one’s thinking and the background behind an opinion is tied directly to the credibility of the writing.
Standing at the front desk to welcome those who want to learn
Regarding the Engineer Cafe, I also wrote about my own involvement.
I actually stand at the front desk myself, guiding new users through registration and introducing the facility before events.
Those guides explicitly include people aspiring to become engineers as target users.
Fukuoka City’s official guidance also describes it as a facility open to everyone from aspiring engineers to experienced professionals.
The fact that I guide people there myself. The fact that it is officially explained that way.
Being able to put both of those into the article was another positive point.
And failing to connect people
This happened when I spoke at an event at FGN.
At the networking reception, a startup executive spoke to me. They were looking for engineers.
Later, an engineer who matched their requirements came to the Engineer Cafe to ask if there was any work available in Fukuoka.
However, I couldn’t meet the executive after that, and I failed to pass the message along.
The person was there, and the inquiry was there. Yet, I couldn’t connect them.
That is why I want all the consultation desks and support facilities—including the Chamber of Commerce and Industry—to communicate with each other on a day-to-day basis.
I want there to be a system that can route inquiries to the appropriate people, without relying on whether I happen to run into that person next time.
That is the story I wrote.
What I learned on my own. Guiding people at the desk. And failing to connect someone myself.
Because these experiences remained alongside my opinions, I believe it became an article that only I could write.
If I have shared this much, I want the first draft to pick up more of it
However, getting to the final version required several corrections.
For instance, my mention of Mori, which I had brought up from the beginning including my expectations regarding expertise, was dropped in the first draft. It had ended up focusing in detail only on Sasaki, so I had to ask to bring it back.
Regarding the collaboration among support facilities, what I wanted to express was an awareness of the problem that consultation desks and facilities as a whole are operating separately in silos.
Yet the draft leaned toward framing it as a matter of the relationship between the Chamber of Commerce and Industry and the other facilities.
For points like these, I want what I have already shared to be preserved right from the first draft.
On the other hand, the specific scene about speaking at FGN and the networking reception was information I added afterward. I hadn’t provided all the material from the beginning.
Therefore, I wanted to incorporate two things into the writing workflow:
Do not drop along the way what I have already talked about.
Draw out the important things I haven’t spoken about yet.
Preserving both how to write and how to ask
First, preserve the writing style I value in a form that can be used in the next piece of writing.
The words I choose, and their strength.
Highlighting what I want to convey through short sentences and line breaks.
Connecting past articles and URLs where facts can be verified to my current opinions.
My own experiences that lie behind expressing those opinions.
Rather than explaining these every time after the writing is finished, I want to build them into the review process before generation and before it is shown to me.
So to speak, a gate before outputting the writing.
And also preserve how to ask me questions.
For example, even if the AI knows that an executive asked me for talent advice, if it doesn’t know the specific situation,
In what kind of setting did you hear that?
have it ask that.
If I had already answered up to the point about the networking reception, have it write using that response.
Ask questions one by one. Read the answers, and think about what is missing next. Once the materials are gathered, move on to the body text.
By connecting questioning and writing in that way, I believe we can reduce the back-and-forth of having to re-explain my own stories over and over.
Next time, bringing out these strengths from the first draft
If we organize how to write and how to ask, will it really work well?
That is also something I want to verify.
Right now, using examples of questions and writing, I am testing whether we can catch text whose meaning has drifted, and let through what is appropriate.
In the process, there were times when information present in the original response was omitted from the summary passed to the evaluating AI. Even though it could be verified by going back to the original answer, it was judged as “unsubstantiated” based solely on the materials provided.
Along with the rules for writing, passing the materials properly was also necessary.
However, testing sample problems and finishing an actual article with fewer interactions must each be verified separately. Whether we can reduce the back-and-forth and whether I can publish with confidence are things I will observe in upcoming articles.
What I want preserved from the first draft.
That is, having my own words come through just as they are.
What I thought in the past and what I did where being connected to my current opinions.
And having my own experiences remain, including both successes and things that didn’t go well.
In this article, because we achieved that, I felt ready to publish it.
Next time, I want to bring out those strengths right from the first draft with fewer rounds of feedback.
To that end, including both my writing style and the way questions are asked of me, I intend to refine the process of writing together with AI.
References
Want to put this into practice at your company?
Our AI reception "Cloudia" takes interview and speaking requests and questions about this article.
Support this article
If this article was helpful, please support us with a tip. You can choose any amount.
Related Posts
I Founded a Company to Make "Ah, They Understood" an Everyday Experience
This is the story of how our representative, Kousuke Terada—a former trumpet player diagnosed with ASD at age 28—founded the AI company Cor.Inc. after working in sales at a musical instrument manufacturer and Recruit. He writes about why he is using technology to give shape to what words alone cannot convey.
How Giving Lightning Talks Every Week for a Year Completely Transformed My Life
The growth, lessons, and life changes gained by an on-the-edge founder and inexperienced engineer through the challenge of weekly lightning talks


