Work / Case study

A whole agency working with AI, not just the two people who were curious.

A 25-person marketing agency had a couple of people getting a lot out of AI and everyone else watching. I built the shared setup, the client dashboard and the training that brought the rest of the team in. Today 22 teammates are set up, and the team writes and shares its own AI processes.

Client: 25-person marketing agency
What I did: Client dashboard, team AI workspace, training
Who uses it: The whole team, on Mac and Windows

Where they started

The agency wasn't short on interest in AI. It was short on a way for everyone to use it.

Two teams ran on two different systems, one in spreadsheets and one in a database tool. Nobody could see a client's whole picture in one place.
Client reporting was pulled together by hand.
The site clients used to check on their work was brittle.
The "shared" AI workspace was really one person's personal setup. Teammates on Windows couldn't get it running at all.

What I built

1. A dashboard every client can log in to

It syncs every hour from the sheets the team already works in, so nobody had to change how they work. If a sync reads bad data, it keeps the last good version instead of overwriting it. Each client sees only their own work. A "waiting on your review" queue, with email alerts, stops approvals from sitting. The team adds a new client themselves in under 10 seconds.

It was live for the first client about seven weeks after we started.

A simplified drawing of the client's view, not a screenshot.

2. A team AI workspace anyone can install

One command takes a teammate from a fresh laptop, Mac or Windows, to a working setup. A health check says what's wrong in plain words. API keys get saved safely and never pasted into a chat. Updates are one command too.

A simplified drawing of what a new teammate sees.

3. A safe way to share what works

When someone builds something useful, they say "share this with the team." The change gets checked automatically: fixed checks first, then an AI review. Safe changes merge on their own. Anything risky waits for a person.

That matters because most of the team isn't technical. They never see a merge conflict. They just see that their work is now available to everyone.

4. Training, until it stuck

Onboarding sessions for each team, a weekly check-in, one-to-one setup help, and written best practices people can go back to.

Who did what

Most of the processes in that shared library were written by the agency's founder and team, not by me. That's the point. My job was the dashboard, the plumbing and the training that made it possible for non-technical people to build their own, share them, and trust them.

Results

22: teammates set up and working
16: shared team processes, across sales, client delivery, reporting, content and operations
5+: teammates besides the founder who have contributed their own
~10 weeks: to the first share that was checked and merged with no human involved
2,282 → 414: files in the workspace after slimming it down

We didn't measure hours saved, so I won't quote any. What changed is who uses AI at the agency: it went from two people to the whole team, and the team now adds to it without me.