Before you give a prompt to AI, dive into the field. — Work techniques for humans to become 'Editors-in-Chief' in the generative AI era

Using generative AI makes work faster.

Sentences are polished.
Drafts for materials are generated.
Ideas are produced in bulk.
Summarization and classification can be done in an instant.

However, I have been thinking about this again recently.

What is truly being tested in the AI era is not
whether or not you can use AI.

Rather, what is being tested is

which is the human side of the attitude.

The book I read this time, 'The Strongest Organization is Born with Generative AI,' was not just a book on AI utilization techniques.

It was about how to connect management and the field.
How to turn chaotic primary information into decision-making.
How to make AI not just a work tool, but the nervous system of an organization.

It was a book that faced those questions.

There is no point in building a 'faster carriage' with AI

What was impressive at the beginning of this book was
the idea of not building a 'faster carriage'.

This means that just speeding up work with AI is not enough.

For example, suppose that document creation that used to take 3 hours now takes 30 minutes.
That in itself is amazing.

But what if that document is answering the wrong question in the first place?

It looks polished.
The sentences sound plausible.
The structure is not broken.

But it doesn't reach the management issues that really need to be solved.
It doesn't touch the pain of the field.
It doesn't capture the true feelings of the customer.
What the president really wanted to say has been diluted into clean, general arguments.

If that happens, you will just be 'making mistakes faster'.

What is needed in the AI era is not just an improvement in work speed.
It is to rewrite the OS of decision-making itself.

Work in the AI era becomes editing

In this book, AI is viewed as an 'excellent investigative reporter'.

This resonated with me quite well.

AI gathers information.
It summarizes.
It classifies.
It lays out arguments.
It compares.
It proposes outlines.

It is like a reporter who can organize vast amounts of interview notes in an instant.

However, the one who ultimately decides the layout is the editor-in-chief.

What to pursue.
What not to pursue.
Which information to put on the front page.
Which arguments to discard.
Where to place your bets.

It is the human who decides these things.

Work in the AI era is shifting from being an operator to being an editor-in-chief.

And there are three main jobs for an editor-in-chief.

The first is to define the question. Deciding what to pursue and what not to pursue.

The second is to articulate the criteria for judgment. Giving the AI what it should consider important.

The third is to commit. Deciding where to place your final bet based on the information obtained.

No matter how much AI evolves, these three will remain as human tasks.

In fact, the more AI evolves, the more the value of these three will increase.

Primary information is a fact with context

What I felt was particularly important in this book was the definition of primary information.

Primary information is not just raw data.
It is not just the facts that occurred on the ground either.

What matters is what kind of context is attached to those facts.

For example, suppose a customer says, 'The price is high'.

On the surface, this looks like a pricing issue.
Therefore, it easily becomes a discussion about lowering the price.

But is that really the case?

The customer might have said the price was high, but perhaps they just didn't understand the difference in value.
Perhaps they were only comparing it to something cheaper.
Perhaps they were just using price as an excuse because it was difficult to explain internally.
Or perhaps our proposal failed to resolve the client's concerns.

Even with the same words, 'the price is high,' the meaning changes depending on the context.

Therefore, I believe primary information is not just the fact that 'the customer said this,'
but information that includes 'in what context and what does that statement mean.'

The same applies to data that is valuable to AI.

It is not enough to just feed in large amounts of data.
What does that data mean?
From what context should it be viewed?
What kind of judgment will it be used for?

Only with those labels can AI begin high-level analysis.

The job of context engineering

This is where the concept of context engineering becomes important.

Simply passing information to AI is not enough.
You need to teach the AI the company's perspective.

What does the company value?
What does it see as a winning strategy?
What are the landmines to avoid?
What kind of customers does it want to be chosen by?
What failures has it experienced in the past?
Which expressions are tacky, and which ones are characteristic of the company?

Only after providing these will the AI begin to function as a 'staff officer for the company.'

In this book, Role, Focus, and Goal were presented as the axes for context definition.

Role is what personality or role you want the AI to think from.
Focus is what to pick up and what to discard.
Goal is what it will ultimately be used for.

I believe these three will become the basic OS for future AI utilization.

Just asking AI to 'create a project proposal' is weak.

What is needed is to clarify:

'Who are you thinking as?'
'What do you prioritize, and what do you discard as noise?'
'Who will ultimately use this to make what judgment?'

clarifying these points.

In other words, a prompt is no longer just a request.
In practical work, a prompt is closer to a specification document.

A prompt is a specification document.

Make it look good,
make it easy to understand,
make it cool.

With requests like these, AI can only provide average answers.

AI truly demonstrates its power when your vision of the finished product, criteria for judgment, and constraints are clear.

What role should it think from?
What is the background?
What should be included and what should be discarded?
Who is it for, and what should it decide?
What is the output format?
What are the criteria for good output?
What expressions should be strictly avoided?

A prompt defined to this extent is no longer just text; it is a blueprint.

This is, even before AI utilization, the very definition of work requirements.

And the ability to write this specification document is exactly what planners, editors, and marketers will need from now on.

AI does not go beyond your thinking framework.

However, there is something you must not forget here.

AI is excellent.
But AI will not go beyond your own thinking framework.

If your question is shallow, the AI's answer will also be shallow.
If your hypothesis is distorted, the AI will beautifully reinforce that distortion.

This is what is scary.

AI can provide quite persuasive answers even to the wrong questions.
That is precisely why humans need metacognition.

In the first place, is this question good?
Is this hypothesis correct?
What do I want to see?
Conversely, what do I not want to see?

What you make AI talk about reflects your management stance itself.

From a distorted lens, only distorted primary information is born.

That is why, in the AI era, there is meaning in reading books.
There is meaning in meeting people.
There is meaning in going to the field.

Reading books means increasing your thinking frameworks.
Meeting people means encountering expressions, pauses, enthusiasm, and discomfort that AI cannot yet pick up.
Going to the field means touching pain before it becomes numbers or data.

The more AI develops, the more humans need to train their human-like sensors.

The 1000-hour analysis is the resolve to find the right context.

This book also discusses how gritty, hands-on analysis is necessary for valuable AI design.

Do you have the resolve to perform a 1000-hour analysis?

This is not just about grit.
I believe it is the time required to find the correct context.

Listen to the voices on the ground.
Read customer complaints.
Look at past failures.
Pick up on the quirks in the president's language.
Face the reasons why things didn't sell.
Deconstruct the discomfort in documents.
Confront the reasons for losing to competitors.

It is through such gritty analysis that things finally become clear.

“Was this the real problem for this company?”
“Was this customer seeking peace of mind rather than features?”
“Should this product have been talked about as a source of pride rather than performance?”
“Was this proposal not about site renovation, but about redefining the company?”

AI runs fast.
However, it is humans who must find the terrain it should run on.

Elegance and AI utilization are connected.

This book is also connected to 'An Introduction to Elegance,' which I read recently.

Elegance was not just about behaving gracefully.
It was a technique for choosing distance, words, silence, and rejection so that mutual dignity is not compromised when in the presence of others.

I believe work in the AI era is the same.

Efficiency through AI is not the goal itself.
You must not use AI to carelessly process the pain of the front lines.
You must not use AI to consume customer voices as mere data.
You must not use AI to turn the president's philosophy into thin slogans.

Rather, AI should be used to protect human dignity.

To ensure the voices of the field are not erased.
To pick up on customer discomfort.
To put the president's philosophy into words.
To connect primary information buried within the organization to management decisions.

The more you use AI, the more elegant your work should be.

Mastering technology and what to deliver to the world.

In the end, this is what is asked of us.

What will you deliver to the world by mastering technology?

It is not about making documents faster with AI.
It is not about polishing text with AI.
It is not about replacing human work with AI.

What is truly being asked is,
whose pain will you alleviate with that technology?
What voices from the field will you visualize?
What company pride will you put into words?
What customer anxieties will you resolve?
What future decision-making will you support?

What is needed is the objective design capability to remain unconstrained by entanglements.
And the gritty implementation capability to continue standing by the pain of the field.

When these two come together, AI ceases to be a mere efficiency tool.

AI becomes the nervous system of the organization.
It becomes a translator connecting the field and management.
It becomes a strategist that expands human editorial power.

Therefore, what is important in future work is not using AI.

Before passing it to AI, dive into the field.
Before throwing it to AI, refine the question.
Before leaving it to AI, doubt your own lens.

And finally, as an Editor-in-Chief, commit yourself.

Technology evolves.
Tools change.
However, what humans want to see,
what they do not want to see,
and what they want to deliver to the world,

that is the one thing AI cannot decide.

Writing Exercise

1. What work are you currently leaving to AI?

________________________________________________________

2. Before that, what is the question that humans must define?

________________________________________________________

3. Where is the primary information in your work?

Example: Sales meeting notes, customer voices, field photos, casual chats with the president, social media reactions, reasons for failure, etc.

________________________________________________________

4. What context labels are needed for that primary information?

Example: Not a price issue but a value translation issue / Not a recruitment issue but a pride verbalization issue, etc.

________________________________________________________

5. Try writing down the Role / Focus / Goal to pass to the AI

Role: Who should the AI think as?
__________________________________________________

Focus: What to pick up and what to discard?
__________________________________________________

Goal: What will it ultimately be used for?
__________________________________________________

6. What do you want to deliver to the world using technology?

__________________________________________________

References and Bibliography

• Kei Nakano, 'The Strongest Organization is Born with Generative AI', KADOKAWA

• 'Introduction to Elegance'

📝
Consultation window for Masaomi Soda
My Vision & Goal
'Together, we create a place where hardworking people can earn money while fulfilling their self-esteem.'
'Managing a work space where I and those important to me can fulfill our self-esteem.'
'Management that fulfills self-esteem together while loving those around us.'

For those who tried the workshop while reading this note and felt:
'I want you to design it together with our company version'
'I want the entire management team to do this metacognition workshop'
'I want to compile it into a book, so I need a partner for editing and structuring'
1. For those who want to get a feel for the atmosphere first
I post daily reading notes and fragments of thought experiments on Instagram.
Please take a peek at 'what kind of vibe metacognition is blended into daily life'.
➡ Get to know the author

https://www.instagram.com/masaomisoda_dokusho/2. For those who want to have a specific consultation
For workshop design, in-house training, management retreats, book planning, etc.,
please feel free to contact me here if you think, 'Can we create this kind of space together?'
➡ Want to inquire immediately

masaomi_soda@garb.gr.jpPlease send an email to this address.
It will be smoother if you add a note saying, 'I contacted you after seeing your note.'
3. For those who want to know about the company's overall initiatives

Organization of management issues, business issues, and sales issues
Hands-on marketing and sales support
Creative production such as WEB / SNS / printed materials
We provide end-to-end support.
For those who say, 'I want to change the business itself starting from metacognition training,'
please see the company-wide service overview from here.
➡ Get to know the company (inquiries can also be made here)
GARB JAPAN, INC.
GARB JAPAN Co., Ltd.
https://garb.gr.jp/

#GenerativeAI
#AIUtilization
#PrimaryInformationManagement
#ContextEngineering
#ReadingNotes
#WorkTechniques
#Marketing
#Branding
#Management
#PromptDesign