An employee at a Japanese financial services company was given a labor-intensive task — modify an enormous spreadsheet of data manually, using a mind-numbing repetitive process. It would have taken days.

He chose instead to combine some AI and automation tools — available in the company for anyone to use — and completed the task in an hour.

Was he praised for his innovative ingenuity and initiative?

Far from it.

His colleagues were angered that he demonstrated how their jobs could possibly be replaced — or how he made it more difficult to justify overtime pay, which many rely on to supplement their income. His managers were annoyed that he did not seek permission before using a more efficient method. He was sternly reprimanded. He was so universally pilloried that he began to question his own mental stability and his perception of the world, because he could not understand what he did wrong. He sought the help of a mental health professional. The professional assured him he is not crazy.

So what the hell is going on?

Japan's generative AI usage rate stands at 26.7%. Compare that to 68.8% in the United States and 81.2% in China. In corporate Japan, AI strategy development rates sit at roughly 42 to 50%. In the US and China, they exceed 80%. A survey of 210 senior executives across Japanese automotive, manufacturing, financial services, and retail found that while 93% reported at least a general understanding of AI, only 8% had reached full-scale implementation. The rest were in pilot or partial stages — or had not started at all.

These numbers do not describe a technology problem. Japan has the technology. The tools are available, affordable, and increasingly adapted for Japanese-language business environments. Technical readiness among Japanese organizations outpaces organizational readiness. The constraint is not the tool. It is the leader holding it.

I have seen this pattern too many times to mistake it for anything other than what it is — and not just in AI.

An executive tells me she has bold ideas for how a new method could transform her function. When I ask why she hasn't proposed them, she explains that the organization is too conservative — in Japan and at headquarters. She is waiting for others to change their thinking first.

Yet her own tentativeness is part of the conservative inertia she is complaining about. For all she knows, others in the organization with equally progressive ideas are waiting for her to move first. She is the problem she is blaming on everyone else.

A CEO describes his organization as slow to adopt a new technology. When I push, it turns out his people are not as resistant as he thinks. They simply don't believe he will back them if something goes wrong. So they wait. He waits. The organization waits. And next quarter, the usage rate at his company is still abysmally low.

This is what I call the Refraction Layer.

Organizations don't reject AI. They refract it. A technology with genuine capability enters the organization and gets bent — through hierarchy, through accountability aversion, through the need for consensus — until it becomes a pilot project, a study group, a committee to evaluate feasibility. The tool doesn't change. The organization bends it until it is ostensibly safe. And therefore useless.

I wrote some time ago that conservatism in organizations is behavioral, not cognitive. The executives I meet are not confused about whether AI can add value. They know it can. They have read the reports. They have attended the workshops. They have nodded at the impressive demos. The question is not whether they understand AI. The question is whether they have the discipline to act on what they know.

I define discipline as the ability to do what you know is right — even when the immediate consequences are uncomfortable and the future payoff is not certain. Discipline is what separates the CEO who reforms his Japan operation — who fires underperforming executives, promotes on merit, and ignores the warnings of local staff and putative Japan experts — from the one who hesitates at every step and watches his early progress stall.

AI adoption in Japan follows exactly the same pattern.

The executives citing legal uncertainty, data privacy risks, and the unreliability of AI outputs are not wrong about the concerns. Those concerns are real. But they are using real concerns as reasons to delay rather than reasons to act carefully.

Your legal team's job is to help you act safely — not to give you permission to wait indefinitely. I have met company legal managers who would gladly chain the front doors of your office building closed to eliminate all potential risk.

Your IT team's job is to build governance around what you are doing — not to veto from the outset what you have yet to start.

Japanese technology companies have imposed restrictions on generative AI tools within their operations, primarily due to data security concerns. This is understandable. It is also, in too many cases, the end of the story rather than the beginning of a response. The correct reaction to a data security concern is a governance framework and an enterprise-grade deployment. The incorrect reaction — the common one — is to ban the tool and declare the matter closed.

When SoftBank, Sony, Honda, and NEC formed a consortium to build Japan's own domestic AI foundation model, they gave a specific reason: they did not trust foreign platforms with Japanese data. The instinct is legitimate. But notice the form the response takes. Rather than building governance robust enough to use what exists, Japan is spending a trillion yen to build something it can trust instead. Bold at the national level. Avoidance at the organizational one.

The problem with AI adoption in Japan is not about technology. It is about trust. Luddism is a trust problem in disguise.

The bottleneck is not AI. It is not trust in AI. It is trust in you.

So what do you do on Monday?

Stop forming committees to study AI feasibility. If you have one, disband it. A committee is proof you already know AI has value and lack the discipline to act on it.

Pick one decision your organization makes repeatedly — a forecast, a customer prioritization, a content output, a diagnostic — and assign an AI-assisted process to it this quarter. Not a pilot. A process. The distinction matters. Pilots are designed to be abandoned. Processes are designed to produce results.

Find the person on your team who is already using AI to produce better work. Promote them or give them expanded scope. Do it visibly. This is how you signal that the permission structure has changed — not through a policy document or a town hall, but through who gets rewarded.

Accept that the first AI-assisted decision that goes wrong will be uncomfortable. It will happen. The question is whether you treat it as a reason to learn and improve, or as permission to retreat. Leaders who retreat at the first failure were never really committed. They were waiting for an excuse.

And stop listening to the people who are telling you it isn't safe yet. They are not protecting you. They are protecting themselves — or at least they think they are.

So. Can you build an organization where acting on what you know is possible? Where your people trust that if they take a risk on AI and something goes wrong, you will stand behind them rather than making them the scapegoat? Can they trust that your objective as leader is to enable achievement of previously unachievable business outcomes? Do they view technology as a force multiplier to create great value — as opposed to a labor substitute and a cost-cutting tactic?

These are not AI questions. They never were.

They are leadership questions. And the only person who can answer them is you.