How I, who thought 'analysis is impossible,' managed to complete turnover analysis. AI was the key.

Nice to meet you. I work in HR and labor affairs.

In this note, I would like to share my trial and error and discoveries regarding how I have used AI (especially ChatGPT) in my actual work.

In this first installment, I will write about my experience using the free version of ChatGPT for 'turnover analysis.'
As the title suggests, I was not good at analysis. Even so, together with AI, I was able to produce results. I hope this article becomes a step forward for those who are struggling somewhere to say, 'Let's give it a try.'

Turnover analysis has been a long-standing challenge.

As an HR professional, I am constantly required to analyze the trends of departing employees. I am no exception, and I have been asked by my superiors for 'visualization of reasons for leaving' and 'reports based on numbers.'

But honestly, I was not good at analytical work. It is not that I cannot use Excel, but I was not good at thinking processes like forming hypotheses based on numbers or backing them up.
I felt I understood the trends in turnover somewhat through my daily interactions, but I could not present them in a way that management could accept, and the sense of a challenge remained with me for a long time.

I decided to ask ChatGPT.

At one point, I thought, 'Maybe AI could give me some hints,' and decided to consult ChatGPT. It was the free version (GPT-3.5), but it still provided sufficient hints.

I asked it this:

'I am collecting free-text data on reasons for leaving, but I don't know how to analyze it. Where should I start?'

What came back was surprisingly practical advice tailored to the field.

• First, how to categorize the data

• Which axes to use to organize it to see trends

• After aggregation, what perspective to use to structure the report so it conveys the message to management

As I repeated these exchanges, a 'pattern of analysis' began to form in my head.
It felt as if there was a senior colleague next to me teaching me the work.

Finally, I reached the results.

As a result, I was able to complete an analysis report on turnover trends. In it, the discomfort that I and my superiors felt daily—the hypothesis that 'perhaps a certain group has a high turnover rate'—was backed up by numbers.

When I reported these results, the reaction from management was clearly different.
They accepted it not as 'intuitive concerns' but as 'suggestions based on data.'

It is not that I have become 'someone who can do analysis.'

I still do not think I am good at analysis.
However, if you borrow the power of AI, you can also do analytical work.
I gained that firm sense of accomplishment through this experience.

And before I knew it, I was the first person in the company to incorporate ChatGPT into my work. I don't have any grandiose feeling that I 'carved out the future.' It's just that I was trying out 'using AI for practical work' faster than anyone else.As a result, I was simply able to solve a long-standing problem.

What I want to write about on note in the future

In this note, I plan to cover the following themes in the future:

• The ChatGPT prompts I actually used and their effects

• How to use AI for work improvement even for free

• Examples of 'collaboration with AI' useful in labor and HR settings

• Application of AI in evaluation systems, job satisfaction, and system design, etc.

Even without special skills, if you put effort into 'what to ask and how,' AI can be a sufficient partner.

I will share that realization and my own tips on how to use it in future articles.

Finally

If you feel like, 'I want to try using AI, but it might be impossible for me,' I would be happy if my experience could lighten that hesitation even a little.

In the article below, I specifically introduce the prompts I actually used when conducting turnover analysis, as well as the data organization procedures. Please take a look if you'd like.