Kohler Reports 98% Copilot Adoption in Just 6 Weeks

Many companies hand artificial intelligence to a pilot group and wait for the results. Kohler started with a Microsoft 365 Copilot license for the whole company.

In 6 weeks Kohler reports 98% adoption, according to the customer story published by Microsoft^[1] in 2026.

Kohler has been making premium kitchen and bath products for 153 years. It also has a hospitality brand. Microsoft writes that a change management and training program helped the company get to that 98%.

Licenses for everyone, then the work on people

Microsoft's page lists four moves in the program.

• License: Microsoft 365 Copilot rolled out to the whole company.
• Training: sessions the page describes as very well attended.
• Champions: an internal network of colleagues who help others use the tool.
• Leadership: open backing from executives for the use cases in each area.

Eric Tabor is Kohler's Chief Digital Officer. He says CEO David Kohler puts AI among the top three priorities at every company town hall he has attended.

"For a change of this scale you need a change management team and you need support from the top," says Tabor. He and the other executives use AI every day and say so in company forums.

Matt Wilson is a Senior IT Manager at Kohler. Using Microsoft Copilot Studio, he built an agent called "Leadership Candor Coach". The agent combines Kim Scott's "radical candor" communication method with the company pillar "Bold Leadership".

"The Candor Coach agent helps leaders stay true to our principles," says Wilson.

The numbers Kohler and Microsoft report

All the figures below come from the Microsoft customer story on Kohler^[1].

• Adoption: 98% for Microsoft 365 Copilot in 6 weeks.
• Time: 5,000 employees save 2 to 5 hours a week.
• Code: releases improved by 25% with GitHub Copilot.
• Errors: defects reduced by 75% with GitHub Copilot.
• Agents: 3,000 created across the company so far.

These are results announced by the company and its vendor. There is no independent verification in the material I read. The numbers measure usage and operational efficiency, and the page is silent on revenue.

Where the story stops

I measure a return with a before, an after and a denominator. Kohler's page gives the after. On the before and the denominator, four points remain open.

• The 98%: the page is silent on the total number of licenses and on what counts as usage.
• The hours: 2 to 5 a week is a wide range, with an unknown measurement method.
• The 25%: "improved releases" could mean faster or more frequent.
• The 75%: the starting number of defects is missing.

A course correction along the way is missing from the text I was able to read. In a story like this, it is the first thing I look for.

Anyone looking for predictive maintenance and plant downtime cases will find a different story here. Kohler is a manufacturing company. The page talks about office work and code.

Six steps a company can copy

A company of any size can follow these steps.

1. Record today's release times, defects and hours spent on repetitive tasks.
2. Give everyone a license and open the training sessions right away.
3. Pick a champion for each department and give them time to help colleagues.
4. Ask the CEO to talk about AI at every town hall and to show how they use it day to day.
5. Let the people who know the process build the agents, as Wilson did.
6. Read Microsoft's page on AI credits and limits for Microsoft 365 subscriptions^[2] before widening usage.

The question for anyone running a company

Wilson works in office IT and built an agent for managers. How many people in your company are allowed today to build an agent for their own work?