[Series Part 11] A 50-something Entrepreneur's AI Challenge! Are AI Tools Like a 'Bento Box'? The Ultimate 'Right Person for the Right Job' That Impressed a Site Supervisor

■ Introduction: 'Right Tool for the Right Job' is Key for AI Too

Hello! I'm yama-san, a 50-something analog person who worked as a site supervisor for 25 years, and now has established a corporation to start a business, aiming to reduce back-office work by 50% using AI and achieve a monthly profit of 50,000 yen.

Last time (Part 10), I talked about how AI combines facts to tell 'plausible lies,' which is why the 'ability to verify,' something only humans possess, is so important.

This time, I have tested various AI tools that I can use in practical work and organized their features. If you are lost, wondering 'There are too many AI tools, I don't know which one to use to solve my problems,' please use this as a reference.

■ 1. Research: The 'Gofer' and the 'Librarian'

When I want to research something in depth, I use two different AIs depending on their characteristics.

The first is
'Manus' . This is what we call a 'super-competent gofer' on the job site. The research capability of current AI is truly staggering; if you give it detailed conditions, it acts exactly as requested. In particular, Manus doesn't just output text, it also quickly creates comparison tables and graphs. What impressed me most of all was that it 'summarizes sources and citations' thoroughly.

Since it saves the massive amount of time and effort required to search from scratch and verify information, it can significantly shorten the time spent on business planning and research.

The second is
'Deep Research' . This one is the 'librarian at the library.'

It is a 'comprehensive' tool that, just by asking it broadly to 'tell me about the history of AI,' will scour hundreds of sites and create a thick, well-organized report without any omissions.

Use Manus for specific document creation, and Deep Research for broad and deep investigations. By using them according to your purpose, the struggle of research becomes dramatically easier.

■ 2. I Compared Slide Creation AI to a 'Bento Box'

There are many AIs for creating slides, so I compared them to a bento box to make it easy to understand.

1. Genspark (Generative type) = 'Creative Chef's Cuisine'

   1. Just by asking it to 'make it look like Japanese food,' it creates a photogenic dish from scratch. However, if you look closely, it can have an 'AI-like' feel, and the downside is that the cost (credit consumption) is high.

2. Irisil (Template type) = 'Japanese Bento Shop'

   1. The layout is fixed, giving you the peace of mind that you can take it straight into a Japanese business setting. However, it is a bit of a hassle to fill the fixed frames with text yourself.

3. Gamma (Template type) = 'Stylish Overseas Sandwich'

   1. It looks cool, but it might not fit well in a Japanese company. Since the content is created automatically, it's dangerous if you don't check it thoroughly. Especially tables and graphs, etc.

I've tried many, but my personal number one right now is 'Manus' or 'Claude'!

This is becauseeven the free version can output to PowerPoint (.pptx) data. I realized that instead of expecting 'AI to create something perfect,' having 'AI create the foundation and then making fine adjustments by hand in the PowerPoint I'm used to' is ultimately the fastest and most reliable solution.

■ 3. Diagram Creation: 'Sticky Notes' and 'Summary Notebooks'

Next, I'll discuss how to use tools for diagramming text. This solves the problem of documents being full of text and failing to get the message across to the other party.

"Napkin AI" is like a 'sticky note' you put in the margins of a textbook. It's convenient when you want to take a part of a text, turn it into an easy-to-understand diagram at a glance, and insert it into a slide. On the other hand, "Claude" is like the 'summary notes for the entire chapter' you make the day before a test. It takes a long, complex text, gives you a bird's-eye view, and organizes it into a single, massive correlation diagram or mind map.

So, the distinction is using Napkin when you want to look at things locally with a magnifying glass, and Claude when you want to see the whole picture with a bird's-eye view.

The figure below isa diagram created with "Napkin AI".

Click here for a video on how to use it↓

■ Real failures and lessons: Searching for the 'landing point' of slide creation

I may be talking big about how to use them, but in reality, I am still searching for the 'optimal landing point' for slide creation.

When I first started using them, the Japanese generated by other tools would become garbled and unreadable, or I would get lost because the layout didn't turn out the way I wanted... There were times I felt like giving up, thinking, 'It's faster to just make it myself!'

However, what I learned from those failures is the realization that **'being able to edit it in PowerPoint later (=Manus or Claude)' is the optimal solution for me right now**.

As a secret ambition, I'm hoping that if the materials I create in "NotebookLM," which I use regularly, could be converted directly into PowerPoint data, it would be truly 'invincible' (laughs).

■ Conclusion: Just like a site supervisor's 'craftsman placement'!

It has become clear that AI tools, just like 'carpenters' and 'plasterers,' have completely different areas of expertise.

Trying to make one tool do everything perfectly is where the strain comes from. By combining them for the right job and having a human (the site supervisor) put the finishing touches on it, you can build the best structure. I feel excited that my experience as a site supervisor is being put to direct use in utilizing AI!

If you thought, 'It's not too late to start in your 50s!' or 'This was helpful for choosing tools!', please support me with a **'Like' or follow**! I will continue to muddle through and experiment!