I Created a Prompt to Turn AI-Style Slides into Professional Analytical Documents

IR, performance reports, market research, survey analysis, and research presentations. In these data-centric documents, being easy to compare and read changes is more important than looking "flashy."

Yet, when you let AI create slides, it tends to lean towards a "presentation-ready" style using cards, giant numbers, icons, and gradients.

Of course, that isn't inherently bad. There are situations where visual impact is helpful, such as company introductions or sales materials.

However, the conversation changes for documents where you want to compare numbers carefully.

You need to be able to track not just "what the sales figure is," but also:

• How it has changed from the previous period
• Which items have the largest differences
• Which series are growing
• What proportion it occupies in the whole
• What criteria those numbers are being compared against

This isn't just a prompt to make slides simple.

**"Measure" instead of "decorate."

"Make comparable" instead of "emphasize."

**

Based on this philosophy, it deconstructs the original material and rebuilds tables, graphs, values, and annotations into a structure that is easy to read as an analytical document.

The Identity of "AI-Style" is Not Just Flashiness

When you think of AI-style slides, you might imagine gradients and flashy decorations.

But personally, there is something that bothers me even more.

It is the processing of all information as similar "parts."

If there are three KPIs, use three cards. Use icons for numbers. Make important information large. Use a pie chart if there are multiple categories. Put supplementary explanations in a "Point" box.

This certainly makes things look tidy.

On the other hand, the relationships between data points become weak.

Originally, judgments were needed such as:

"I want to place this number and that number side-by-side."

"I want to see four years on the same scale."

"I want to compare composition ratios by length rather than area."

"I want to prioritize visibility and use a table for this page."

In this prompt, the layout is determined by "what to compare and how."

Example 1 | Turning 1 AI Slide into 2 Analytical Documents

First, look at this.

Before

The original material combines major KPIs, trends in net sales and operating income, and composition by segment into one page.

Visually, it is a very AI-like composition.

Major figures are carded with icons. The bottom section has bar + line graphs and a donut chart. Many gradients and glowing expressions from blue to cyan are used.

It works for grasping an overview in one page, but there are some concerns when viewed as an analytical document.

Specifically, different comparison axes coexist on one page.

In the top section, you see "major figures," in the bottom left "time series," and in the bottom right "composition ratio."

All are important, but the reader must perform three types of comparisons simultaneously.

Therefore, the analytical document prompt split the page.

The first sheet narrows the question to "how has performance trended over four years?"

Net sales and operating income for four periods are placed in the same graph on the left, and major profit indicators for the latest period are organized as a table on the right.

KPI cards are not used.

Instead of just enlarging the latest value, by placing it on the same coordinates as the past three periods, it creates a structure where the current position can be judged within the time series.

Also, information such as Figure, Table, Unit, Period, and Note are placed in small text.

These are not decorations.

They are the coordinates for an analytical document to make it easier to track "what this number indicates," "which period," and "which unit."

In the second sheet, the original donut chart has been changed to a horizontal bar graph.

Donut charts are convenient for seeing the overall atmosphere, but they are not very good at precisely comparing values like 30.1% and 20.5%.

With horizontal bars, you can compare lengths from the same baseline.

The same data is also listed in table format on the right to allow for accurate confirmation of sales and composition ratios.

In other words, on this page:

**Graph = Observe differences

Table = Confirm numerical values

**This is how roles are divided.

This is an example of changing the expression format to match how the data is read, rather than just changing the design.

Example 2 | Splitting by "Purpose of Reading" Without Reducing Information

Let's look at another one.

Before

This is a performance page centered on the profit and loss statement.

The original material contains:

• 3 points regarding performance
• Sales growth rate
• Gross profit growth rate
• Profit and loss statement
• Year-on-year comparison graph for each profit
• Supplementary comments

This is a practical document with a lot of information, but adding AI-like design further adds KPI cards, icons, boxes, and colored surfaces, making it easy for the reader's focus to disperse.

This time, instead of cutting the information itself, I separated "confirmation" and "interpretation."

The first sheet is boldly just a table.

Net sales, cost of sales, gross profit, SG&A expenses, operating income, ordinary income, EBITDA, etc., are compared horizontally by current period, same period last year, increase/decrease, and growth rate.

On pages like this, there is no need to force a conversion to a graph.

Rather, being able to confirm the difference from the previous year just by following the same row horizontally is more important.

If a table is optimal, use it as a table.

This is a concept that is highly emphasized in this prompt.

In the second sheet, operating income, ordinary income, interim net income, and EBITDA are organized into bar graphs comparing them with the same period of the previous year.

On the right, explanations for understanding those changes are placed with numbers.

In other words, it is a structure where you observe numerical changes on the left and read the background on the right.

In the "Before," tables, KPIs, graphs, and comments competed on the same page, but in the "After," the roles are divided into a "page for confirming numbers" and a "page for interpreting numbers."

Instead of pushing everything into one sheet by making the text smaller, split the pages if the questions are different.

This is also an important rule of the analytical document prompt.

What Kind of Documents is it Suitable For?

This prompt is not for making everything look like an analytical document.

It is particularly compatible with documents where numbers, comparisons, time series, composition ratios, and distributions are important.

For example:

• IR and financial results briefing materials
• Performance report materials
• Management meeting materials
• Market research
• Survey analysis
• Research reports
• Research presentations
• Technical reports
• KPI reviews
• Performance and budget management materials

Conversely, it is not very suitable for documents where impression formation through photos and copy is important, such as company introductions, recruitment pitches, product introductions, and brand appeals.

What this prompt excels at is "documents to be read and compared" rather than "documents to be shown."

Main Color Can Be Changed Freely

Another thing I made easy to use this time is the color scheme.

While the Data Laboratory style itself is fixed, only the main color can be changed freely at the time of input.

For example, you can specify:

Main color: Green

Main color: Navy

Main color: Bordeaux

Main color: #315C6B

Backgrounds, body text, and auxiliary lines are fixed to neutral colors as a base, and the specified color is used limitedly for:

• Major series
• Latest values
• Baselines
• Section numbers
• Emphasized ruled lines

Therefore, even if you change the main color, the overall design does not become too colorful.

In this example, I used green, but it can be changed to match the corporate color or the purpose of the document.

Usage: "Insert Image → Confirm Structure → Generate One by One"

Usage is also simple.

First, attach multiple slide images you want to re-edit and send them with the prompt.

At that time, if necessary, specify at the beginning:

Main color: Green

Then, instead of generating images from the start, the AI will check across all pages and propose in text:

• What to keep
• What to integrate
• What to move to another page
• How many pages to restructure into
• Whether to show each page as a table, graph, or text

Once you have confirmed the structure, input "1st sheet."

Then it will generate only the first sheet after re-editing.

By continuing to specify "2nd sheet," "3rd sheet," and so on, you can create them one page at a time.

Separating the structure and image generation instead of letting it make the whole deck at once is also a key point.

From Here On is Membership Content

Since it is designed so you can specify the main color, you can change it not only to the same green as the example but also to navy, bordeaux, blue-gray, etc.

Attach the image and just paste the prompt to use it.

Analytical Document Prompt is Here ↓

note: I Created a Prompt to Turn AI-Style Slides into "Analytical Documents"

If you are a note member, all the prompts and GPTs introduced so far are available, and more will be added in the future!

For the price of one lunch, slide creation will definitely become easier.

Membership: Fujikururi's AI Learning Room

https://note.com/fuji_kururi/membership

Please consider it.