How a UK Construction Firm Got 11,000 Employees to Actually Use AI

Key points

• Kier Group, a UK construction firm with around 11,000 employees, began rolling out Microsoft 365 Copilot roughly two years ago.
• About 40 percent of Kier staff now use Copilot daily; the remaining employees have Teams Premium, which adds AI features to the firm's chat and meeting tools.
• The company ran "prompt-a-thons", small peer group sessions where colleagues tested AI tools together, to reduce anxiety around new technology.
• Kier is building AI agents, software that can carry out multi-step tasks on its own, to collect safety data and analyse drone footage from construction sites.
• A site manager who is dyslexic told Kier's chief operating officer that Copilot had been life-changing for his day-to-day work.

Kier Group, which builds schools, prisons and large water and nuclear energy projects across the UK, had a problem two years ago. Senior leaders were interested in AI but nervous about looking foolish in front of their teams. Nobody wanted to ask the basic questions.

Louisa Finlay, Kier's chief operating officer, had a simple fix. She invited executives and colleagues to what the company called prompt-a-thons: small group sessions where people tried out Microsoft Copilot, an AI assistant built into Microsoft's office software, together. One team answered questions unaided; another answered the same questions with Copilot's help. The difference was immediate and visible.

What actually happened at these sessions?

They worked because nobody had to perform. Small groups meant people felt safe asking questions they would never raise in a company-wide webinar.

In a separate exercise, participants gave Copilot access to their emails, meeting notes and documents from the past six months. The tool then told each person which character from their favourite film they most resembled, and flagged their leadership strengths and blind spots. Finlay says the engagement after that exercise was striking. People started talking.

Since then, she says, adoption has "taken off." Around 40 percent of Kier's roughly 11,000 staff now use Microsoft 365 Copilot for everyday tasks. The rest have Teams Premium, which brings AI tools to the firm's chat and meeting platform. Finlay herself uses Copilot inside Outlook, Excel, Teams and PowerPoint every day, asking it to prioritise emails, draft job descriptions and surface connections in data that she might otherwise miss.

What does this mean for construction workers on site?

The bigger ambition goes beyond desk tasks. Kier is in the early stages of building AI agents to improve physical safety on building sites.

One agent is being trained to gather and organise safety data so human supervisors can spot risk patterns before incidents happen. A second is being trained on Kier's own safety standards and linked to photographs and drone video from its construction sites. Drones already reduce the need for workers to operate at height. The agent would help humans interpret what the drones find.

Finlay calls the guiding principle "Human at the Helm." The AI handles the data gathering and pattern-spotting; a person makes the call.

The company also plans to use AI in its facilities management work, predicting when pumps or lifts are likely to need servicing before they fail.

What should readers take away from this?

Kier's approach offers a practical model. The firm did not mandate AI from the top and hope staff followed. It created low-pressure spaces where curiosity was the only requirement.

One detail from Finlay stands out. A site manager who is dyslexic wrote to thank her after the Copilot rollout, saying the tool had been, in his words, "game changing" for him. That is the kind of outcome that gets buried in productivity statistics.

Microsoft AI first reported the story. Finlay's own summary of what she wants colleagues to understand: "This is about people solving their own problems. It's not being done to them."

What to watch for if your employer is rolling out AI tools: Ask whether training happens in small groups or only through self-paced online modules. Peer sessions, where everyone is learning together, tend to build confidence faster. Check whether the AI has access to your personal documents and what the firm's data storage policy is. Kier stores its data on Microsoft's Azure cloud platform. And look for a clear statement of who is responsible when the AI's output is acted on. "Human at the Helm" is a useful test: if you cannot answer who the human is, ask.