CODE IN TOKYO

Experts in the AI Era: Maybe the People Who Know the Company Will Become More Valuable Than the People Who Can Change Jobs

The Big Stay and a Shift in Career Value

Recently, I heard that the phrase Big Stay has started to take root in the United States.

It apparently means staying at the same company for a long time. In the United States?

The term existed before the rise of AI, but AI seems to have made it much more decisive.

Apparently, the portable skills that job changers bring with them are no longer valued as highly as before, and people can no longer expect the same salary increases from changing jobs.

The explanation was that, as a result, people’s motivation to change jobs is declining.

Until recently, career advice was simple.

In short, the message was:

Become someone who can succeed anywhere.

For a long time, I also thought this was correct.

The IT industry worked this way too.

Rather than knowledge of one company’s specific operations, market value came from general skills such as:

You could use them even if you changed companies.

You could use them even if you changed industries.

So the correct answer seemed to be: improve your portable skills.

AI Is Good at Portable Skills

But since AI began to spread, the scenery seems to have changed a little.

That is because the area AI is best at is precisely the area of portable skills.

General programming knowledge.

Marketing principles.

Accounting knowledge.

Contract review.

Knowledge that once required years of experience can now be produced in seconds by asking AI.

Of course, experts are still necessary.

But the value of people who only know general theory is clearly starting to decline.

Then what is becoming more valuable?

At first glance, the answer might seem to be industry knowledge or domain knowledge. But I think even that will eventually be substituted by AI.

In industries of a certain size, that expertise will be organized, and AI will be able to understand it.

Personally, I think the people who will have an advantage are those who know the context.

AI Knows Best Practices, Not the Reasons Behind Exceptions

AI knows best practices.

But it does not know why our company did not adopt them.

AI knows design principles.

But it does not know why this system ended up with such a distorted structure.

AI can talk about ideals.

But it does not know that this customer had a serious conflict with the company ten years ago, and that the current operation was created as a result.

Real work is made of stories like this.

I sometimes get asked to investigate systems.

I read the design documents.

I read the source code.

I check the logs.

Then I encounter something that makes me think, no one would normally build it this way.

At first, it looks like a design mistake.

But when I ask people who have been there for a long time, a reason appears.

It was an old customer request.

It was a specification from before the law changed.

It was the result of handling a past system failure.

In other words, the system was not simply wrong.

It existed on top of the context of its time.

AI can read code.

But it cannot read that history.

More precisely, it could do so if meeting minutes had been turned into a database and tacit knowledge had been carefully converted into explicit knowledge. But there are not many companies like that.

It might happen at the project level. But if we look across a long span of twenty years, it is almost nonexistent.

The Expert’s Role Is Moving Toward Context

That is why I think the value of experts in the AI era is shifting away from knowledge itself and toward connecting knowledge with context.

AI provides general theory.

Experts provide the reasons for exceptions.

Why was the ideal design abandoned?

Why was an inefficient operation kept?

Why was the textbook approach not followed?

Surprisingly few people can explain these things.

When you think about it, this is natural.

A company is a mass of successes and failures accumulated over many years.

Inside it exists a huge amount of tacit knowledge that cannot be fully turned into manuals.

And that tacit knowledge is exactly the part AI has the hardest time learning.

Ironically, the smarter AI becomes, the less human value may depend on the amount of knowledge someone has.

AI has knowledge.

AI searches.

AI creates documents.

AI writes code.

Then what do humans do?

Probably, they explain why this company became this way.

In the past, people were told to discard company-specific knowledge and build portable skills in order to be valued in the job market.

But in the AI era, company-specific knowledge may start to have value again.

An Era Where Changing Jobs May Become a Loss

If that is true, it is interesting.

Twenty years ago, people might have said:

Someone who only knows internal company circumstances has low market value.

Twenty years from now, people may say:

We are in trouble because no one knows those internal circumstances anymore.

In the end, AI is good at general theory.

And real work usually does not run on general theory.