NEC Just Rolled Claude Out to 30,000 Employees. Here's What They Got Right.

When AI tools land without a task definition attached, they get used for low-stakes tasks or not at all. NEC's deployment of Claude to 30,000 employees tells a different story — and the part worth paying attention to is not the scale, it is the structure behind the first login.

What NEC did differently

On April 23, Anthropic announced NEC — Japan's largest IT company — as its first Japan-based global partner (Anthropic blog, April 23, 2026). NEC is deploying Claude, including Opus 4.7 and Claude Code, to roughly 30,000 employees. Two elements separate this from a standard enterprise license: a dedicated Center of Excellence and structured training developed with Anthropic's direct support.

The Center of Excellence is not an IT committee. It is a permanent function whose job is to help employees find real, recurring work that AI can take on — and to make sure those first tasks stick before more ambitious use cases are added. Training is not optional or self-serve. It is part of the deployment, not something appended to it.

What this means for anyone rolling out AI to a team

The gap between AI tools that get used and AI tools that get abandoned is almost always located in the same place: before the first login. People need to know what their first task is before they open the tool. Without that, most will either use it for something trivial — formatting a document, rewording a sentence they were happy with — or avoid it entirely.

The NEC model makes task identification a structured event. You do not have to build a Center of Excellence to apply the same principle. You need one structured conversation per person before they start: write down two recurring tasks that take over 30 minutes and produce something someone else reads or uses. Pick one. That is the first task.

When a rollout goes without that conversation

A learning and development manager at a 90-person insurance agency gets budget approved for an AI platform in Q1. She sends a welcome email with login details and a link to the vendor's getting-started guide. After six weeks, her usage dashboard shows 31% of staff have logged in at least once. Three people have used it more than twice.

The gap is not motivation. It is task definition. No one told her team what to use it for in their specific role — not in general terms, but concretely. "Use AI for writing" lands differently when the task is "draft the weekly risk summary you write every Monday by hand."

The structured version of what she could have run: a 45-minute team session where each person writes down their two most repetitive weekly tasks, picks the one that feels most mechanical, and spends 20 minutes building a first prompt for exactly that task. The L&D manager does not need Anthropic's support to run that session. She needs the structure.

When the structure is built one person at a time

A team leader at a 250-person logistics operator had Copilot available to his team for four months. When he asked in a team meeting who had used it for something work-critical in the past week, three out of twelve hands went up.

He changed his approach. In the next round of one-to-ones, he asked each person to bring their most repetitive weekly task. He then spent 20 minutes with each person building a single-prompt workflow for exactly that task — not a general guide, a workflow for their specific work, with their specific inputs and outputs.

After six weeks, all twelve were using it for something that saved them measurable time. Not all of them found it transformative. But all of them had a workflow that ran reliably enough to use again.

The NEC Center of Excellence is that one-to-one, institutionalised at 30,000 people. The principle scales down to five.

The only thing that matters before the rollout is complete

The tool is ready before the team is. The investment in getting the team ready — one task, one workflow, one person at a time — is the part NEC built in from the start. That is the part most rollouts skip.