3.8% of Japanese companies say their DX talent is good enough

IPA's DX Trends 2025 contains one comparison that makes the rest of the report fade.

Asked whether they have adequate DX talent in terms of quality, 3.8% of Japanese companies said yes. Germany said 25.1%. The United States said 52.9%.

This is quality, not headcount. Effectively 96% of Japanese companies are saying that having people is not the same as having the right people.

Reporting a shortage while not hiring

The same survey holds a second number that refuses to fit. More than 80% of companies recognise a talent shortage. And 19.4% run no recruitment activity at all.

It looks contradictory. In practice it makes sense.

They cannot write the job requirement. If you post a role without being able to articulate what the person should be able to do, you cannot evaluate the applicants. If you cannot evaluate them, nobody gets hired. Eventually the posting is withdrawn.

The reason the requirement cannot be written is that nobody inside the company can assess that skill set. Hiring someone who can assess it requires someone who can assess it. That loop is where a lot of organisations are sitting.

The missing role differs by country

IPA also asked which role is most lacking. In Japan the answer is business architect. In the United States and Germany it is data scientist.

A business architect designs the join between the business and the technology. The person who decides what to build and where it enters the existing process.

That Japan's scarcest role is this one lines up exactly with the finding that only 6.0% to 11.9% of Japanese firms have redesigned a process around AI. The constraint is not people who can build. It is people who decide what gets built.

Relatedly, 56.4% of Japanese companies say AI researchers are not needed. For most operating companies that judgement is reasonable. The problem is that no replacement role was ever specified.

The shortage is not engineers. It is people who stand between the business and the technology and design the join. The job posting cannot be written because that function was never defined internally.

Three options besides hiring

3.8% will not be fixed through the recruitment market alone. Three approaches are realistic.

Convert existing staff. The people who understand the work already sit inside the company. Teaching technology to someone who knows the process is often faster than teaching the process to an external engineer. The reskilling course of the Human Resources Development Subsidy covers up to 75% of training cost.

Work with outside partners on handover terms. Contract for a state where your team can operate and improve the thing, not for a delivery date. Making the reasoning behind design decisions a documented deliverable changes what the handover is worth.

Fix the requirement first. Before recruiting, write out what this person will do in their first three months at the level of individual process steps. If you cannot write it, you are not yet at the stage of hiring for that role.

Start by defining quality

What does adequate quality look like concretely. At minimum you can say all of the following.

• Someone can decide what to build
• Someone can implement what was decided
• Someone can fix it once it is running
• Those three people share a vocabulary

The fourth is the one most often missing. You can assemble genuinely capable specialists and still watch the output drift away from the requirement, because the language of the business and the language of the technology were never reconciled.

There is no shortage of capable people in Japan. What is missing is a structure in which that capability does anything, and 3.8% is what that absence measures.