[Finishing AI Slides] 4 Patterns Where Manual Work Shines and How to Discern Them

AI generates the material.
Humans finish it.

Recently, I have been writing a series of articles about AI slides.
They are connected if you read them in sequence, but feel free to check out just the parts that interest you individually.

Following that flow, this is about the role of humans. This time, I will share what is necessary for the 'humans finish it' side—how to hone your skills and your discerning eye.

1️⃣ The trigger was a post that appeared on X

Since around the end of last year, creating AI slides with 📓NotebookLM and ✨Gemini Canvas has become a hot topic. I have also been gathering information while testing various tools.

Around that time, I also started using note, so I had more opportunities to gather information on X.
Then, from a certain point on, posts introducing slide tips started appearing on my timeline.

Among them, what captivated me was not the AI, but techniques for finishing slides with human sense and manual work.

When I tried it with some skepticism, the slides created by AI became significantly better just by adding a few small touches.

The shift in mindset from 'using AI output as is' to 'having humans finish it' started around this time.

2️⃣ Grasping the 4 basic types of slides

When I get lost in the structure, I return to the 4 basic types of slides.
This is a reproduction of this post.

① Body & Message type

A simple type consisting of a main visual and a one-line message. Highly persuasive and effective for opening or closing slides.

② Issue Raising type

A type that develops in the flow of 'Current Status → Problem → Solution'. Easy to use in decision-making or proposal situations.

③ List type

A type that organizes and arranges multiple elements. Supports comparisons, presentation of options, checklists, etc.

④ Roadmap type

A type that shows chronological or step-by-step progress. Suitable for plans, process explanations, and phase management.

By keeping these four types in mind, when you see AI output, you will be able to quickly judge,
"This can be organized using the Type ③ List format,"
"This flow is close to the Type ② Problem-Solving format."
and so on.

3️⃣ Increase your input from other people's slides

Since I started encountering slide techniques on X, I have developed a habit of consciously collecting good slides. For that purpose, I use Slideland.

Slides from various themes and industries are gathered there, allowing you to easily reference layouts, color schemes, and ways of presenting information.

Using it to compare with AI output is particularly effective.
"AI produced this, but this layout communicates better."
Such realizations will gradually update your own image of slides.

The amount of input is directly linked to the quality of the final product.

4️⃣ Roles of AI and humans—Summary of the series

Let's organize the division of roles that has become clear through these four parts.

Parts handled by AI

• Proposing structures based on patterns and instructions

• Generating diagrams and layouts

• Creating a draft as source material

Parts handled by humans

• Selecting patterns and designing instructions

• Selecting and trimming source material

• Fitting into basic types and finishing

• Responsibility for the final message

AI is an excellent assistant, but it is the human who decides until the very end whether "this slide communicates effectively."
Set up human checkpoints and have humans make the judgments.
Improving the accuracy of those judgments is directly linked to improving the quality of AI slides.

Rather than raising AI, raise yourself.
Raise yourself through AI.
If you approach it with that mindset, your collaboration with AI will settle into a good balance.

Announcement
Once I'm ready, I'm thinking of introducing some techniques for finishing slides.