How ServiceCore got 80%+ AI adoption across the whole company.

ServiceCore makes the software that runs portable sanitation, septic and dumpster rental businesses. It has about 150 people and is backed by private equity.

Like most software companies, it already used AI. The engineers had Cursor, Claude Code and Codex. Everyone else had whatever they had found on their own. There was no strategy that tied it all together.

Then Claude Cowork landed in February, and Michael Boord, ServiceCore’s President and COO, saw that it could change how every team worked, well beyond the engineering organization.

It was also the kind of moment that could’ve frozen a leadership team.

There were a thousand directions to go and no easy way to tell what was real from what was hype. Michael spent late February and early March talking to anyone who seemed further ahead.

We asked Michael what he was afraid would happen if ServiceCore waited another six to twelve months. He answered with his employees first.

He saw a platform shift. He knew most of his people wouldn’t work at ServiceCore forever, and he wanted them in demand wherever they went next. Reinventing how you work is scary. He wanted the company to be the place that made it powerful instead.

His second answer was the business. If ServiceCore didn’t lean in, it could fall behind the other software providers in its market.

Michael started on LinkedIn. Most of the free AI content there was something he could have written himself with an afternoon to spare.

A small group felt like a step ahead. MasteringAI was one of them.

He also did something most executives skip. Before he brought anyone in for his team, he enrolled in our AI Builders Bootcamp himself. If the executive team couldn’t show the company they were learning too, he knew the rollout would fall flat.

He talked to independent consultants and to very large firms that train Fortune 500 companies. The large firms came with a method and little room to bend it. What he wanted sat in between: a real training catalog and an executive program that matched what he had in mind.

MasteringAI didn’t ask: “Which AI tools should you buy?”

We asked: “What does your leadership team need to decide, and who owns each decision?”

How the engagement ran

1. Pre-work survey: The executive team and key team leads told us where AI was working, where it wasn’t and what worried them.
2. The audit: One hour mapping the skills and agents already scattered across teams, and deciding what should be shared and what should stay with one team.
3. The executive session: The full leadership team in one room in Denver, working hands-on in Claude on real ServiceCore scenarios, then deciding.

By the end of the session, the team had made ten decisions, each with an owner. Among them:

1. Claude becomes the AI system for every department. The vote was unanimous.
2. Governance and managed IT come before a broad rollout.
3. Nobody gets access before they finish the training.
4. Everyone signs an acceptable use policy first.
5. One shared registry of approved workflows, agents and skills.
6. A spending limit on every account, starting small.
7. One standard set of AI tools for engineering.

The rule behind all of it: give people the right AI tools, with guardrails, instead of banning the rest.

Afterwards the team had the session slides, an adoption plan, a decision-making cheat sheet and a template for adding new workflows to the registry. Michael describes what came out of it as a framework built on a few core pillars:

• Fluency ladders, so people know what good looks like in their role
• Governance and security
• An AI council to steer the program
• Outcome metrics

The team treated it as a direction to commit to and adjust along the way.

Early in 2026, plenty of companies were posting AI vanity metrics: tokens used and dollars spent.

When the team brainstormed how to measure success, the first ideas were the same kind of activity metrics. We pushed back: activity shows who is trying, and that’s all it shows.

So we went back to the metrics ServiceCore had always cared about.

Months later, Michael put it the same way: the same metrics, improving faster. Quota per sales rep. The customers and revenue each support and success person can handle. The speed of customer implementations.

1. February

   Claude Cowork lands. AI lives with the engineers, and there is no company strategy.

2. March 24

   The executive session in Denver. Ten decisions, each with an owner.

3. April

   Claude goes to the whole company, behind training, a signed policy and spending limits.

4. August

   80% of employees use AI every week. Revenue is up 30% in a few months, with fewer people. Customer support is “totally transformed.”

5. October

   Almost every employee uses AI every day. ARR per employee, gross margin and EBITDA margin are all “dramatically up.”

Michael is careful about credit, and so are we.

ServiceCore was also disciplined about hiring this year, so these numbers would’ve improved even without an increase in AI adoption. Michael’s own estimate is that 30 to 40% of the jump in revenue per employee came from AI adoption: people taking on more because the tools carried part of the load. Product and engineering are clearly much more productive. Customer support, marketing and several other teams have changed their operational flows and output completely.

While attribution can be tricky, the numbers ServiceCore already tracked improved meaningfully this year: ARR per employee, gross margin and EBITDA margin.

In a June pulse survey, employees asked for one thing above all: training for their own role, down to the best use of AI in their day-to-day job.

Now ServiceCore’s new operations team is going back through its list of agent ideas and ranking them for 2027. Its director of operations and a new go-to-market engineer are among the strongest AI users in the company outside of engineering.

That’s the work ahead: customer onboarding agents, pipeline analysis agents, the repeatable jobs each team does every week. Shared across each team, and owned by people who can build the next one themselves.

A second question comes with it, and Michael hears it from the other CEOs and CFOs in his private equity portfolio: what should AI cost? It’s now a real line on the P&L and hard to predict for 2027. Nobody wants to cap the experimenting that makes the tools pay off.

Most companies of 100 to 500 people are where ServiceCore was in February:

1. A few power users, mostly in engineering, while everyone else waits.
2. A board or investors asking what the AI plan is.
3. An AI bill that keeps growing, with no clear way to tell what it produces.

The fix is a plan your leadership team believes in, guardrails that let people move, and metrics you already trust.

That is the work MasteringAI does.

We start with your pain points. We sit in the room with your executive team, help you make the decisions, and tie every AI effort to a number you already report. Then we help each team build the agents that do part of its job.

ServiceCore started with one session in March. By fall, almost every employee used AI every day.