← Writing

Case study · CrewTracks

The Marketer Who Stopped Guessing, and the Guy Nobody Coached

Six sessions at CrewTracks. Erin stopped waiting on the dev queue. Casey never got coached directly and cleared 1,900 junk CRM records in fifteen minutes.

August 23, 2026

·

5 min read

I don't feel like I am guessing at what to do, which has been a game-changer.

Erin Koets, twelve days after session one

Who this is about

Erin Koets

Erin Koets

Director of Strategic Partnerships, CrewTracks

Casey Black

Casey Black

Sales, CrewTracks

Erin Koets is the entire marketing department at CrewTracks. B2B SaaS, crew management for contractors, one person to market it. When I showed up in April she had never heard of Claude Code.

Casey Black sat in. He is on the sales team. Nobody asked him to be there and nobody coached him. He listened.

Six sessions later Erin runs five project workspaces she built herself and Casey has built two dashboards nobody asked him for. Here is what happened in between.

Session one: the guessing stops

Session one was two hours on site and ran thirty minutes long. We set up her machine and installed a memory system, which is a plain text file holding what the project is, what has been decided, and what is next. Claude reads it before it does anything, so you stop re-explaining your own business every morning.

Twelve days later Erin emailed to book the next session and said this:

"The NOW.md has been super helpful in keeping everything organized. I don't feel like I am guessing at what to do, which has been a game-changer."

That is the actual product. Not a tool, not a workflow diagram. She opens her laptop and knows what to do next.

Session two: thirty meta descriptions, under two hours

Session two we picked something boring and real. CrewTracks has a large pile of landing pages and the meta descriptions on a big chunk of them needed rewriting, the kind of task that sits on a list for a year because it is tedious and never urgent.

We connected the WordPress API so changes could be pushed programmatically instead of typed into a CMS one page at a time, then wrote thirty meta descriptions. Total time, under two hours, API setup included.

I asked Erin how long that would have taken her alone. Her answer: "at least a week."

The API connection matters more than the thirty. Once it exists, the next thirty cost almost nothing. Most of what looks like AI saving time is infrastructure that finally got built because AI made it cheap to build.

She booked session three before she left the room.

Session three: the observer goes first

Session three we ran a Search Console audit on their own pages. Both of them had been carrying a suspicion for months that something was wrong with their marketing and no evidence to hand anyone. The data confirmed it, including a well-trafficked landing page aimed at the wrong thing entirely.

Both of them said the same thing, out loud, at the same time:

"This is so validating."

Then Erin said the sentence that told me the coaching had landed:

"Do you think we could vibe code it?"

Three sessions earlier she did not know what Claude Code was. Now she was proposing to build the landing page herself rather than file a ticket and wait on the engineering queue. Casey, still mostly quiet, said his own version of it:

"It isn't as scary anymore."

An hour or two after that session ended, my inbox got this from Casey:

"This never could have been done by hand. And it was done in about 15 minutes!! Thanks for all your help with this. I am excited to see what else we can accomplish with AI."

He had gone back to his desk and cleaned the CRM. About 1,900 junk leads deleted across every rep. Another 108 spam records under the admin account, gone. Seventy three leads properly tagged with source, detail, and type. And eight real trade show leads that had been buried in the garbage, found and restored.

Fifteen minutes. From the guy who was never coached.

That is the most useful thing that happened in this engagement and it happened to someone who was not the client. Casey watched Erin learn a marketing workflow, extracted the shape of it, and applied it to sales operations. The fluency moved. Training decks do not do that.

Sessions four through six: they stop needing me

Session four we built voice profiles so Claude writes in their words instead of the beige committee voice everyone recognizes on sight.

Session five we built live dashboards. Both of them connected the Zoho API themselves, rolling ten minute secret authentication and all, which is fiddly work neither would have attempted in April. Casey moved himself off Claude Desktop into VS Code partway through, without being told to, because vibe coding made the dev environment finally make sense.

Session six was cleanup. Erin had five projects crammed into one workspace: a website rebuild, SEO and ads, customer communications, and two more, all bleeding into each other. We separated them and she ran the setup on each one herself while I watched. The idea that landed was treating a project like an employee. You do not hand one person your website, your ad account, your customer emails, and your reporting and expect clean work. Same with a project. She left with five clean workspaces and the ability to spot when a sixth needs to exist.

Casey, meanwhile, had built two dashboards on his own between sessions and was automating them. Still nobody coaching him.

What I told them on the way out

My objective is for you to be independent of me.

Erin is building the website rebuild herself now, far enough along that she can hand it to the engineers to implement instead of asking them to design it. Casey is automating dashboards. Neither of them needs a standing appointment with me.

The sessions that worked best were the ones where I mostly sat there. Fix the environment, point at the thing they are avoiding, get out of the way. The learning happened because they did the work in the room with somebody there to say "ask Claude to find that for you" the first ten times the old reflex kicked in. That is the job. Standing next to you until the new reflex is the default.

Evan Van Dyke

Let's figure out what to fix first.

30 minutes. No pitch. Tell me what's broken and I'll tell you what I'd fix first. I've done this for 100+ businesses. I know where the leverage is.

Work With Me