If your employees don’t know how to build AI skills, fire them.
Ok, maybe that's a little harsh. But hear me out.
Knowing how to build Skills with AI is the highest-leverage "skill" you can learn right now.
And if you don't know what Skills are, they're kinda like recipes for how you do a task in your business. You turn the task into a Skill file and hand it off to an AI agent like Claude Cowork. Now, Claude Cowork executes the task the same way every time. You never have to do it again.
You can probably see how powerful this can be which is why you (and your employees) need to learn this stuff ASAP (before your competitors do).
Two weeks ago my executive assistant had never written a single AI skill. She’s non-technical. This week she shipped six polished skills in Claude Cowork that run together as one connected pipeline.
I asked her to build a full podcast guesting engine from scratch. It scrapes every guest from every show I’ve been on and turns them into a warm collaboration outreach list, tiered and sorted, written straight into a Notion database.
I could have easily built this myself in about 30 minutes but I held off on purpose.
Because if I build it, I get one system. If she builds it, I get the system AND an EA who can build the next ten without me in the loop. That’s the whole reason I made her do the work instead of doing it myself.
Here’s the exact build she shipped so you can run the same play on your team this week.
Step 1: Pick one real workflow that’s manual today.
Something with stakes that someone actually depends on. Ours was the podcast guesting pipeline. Two jobs in one: find collaboration partners by scraping guests from every show I’ve been on, and pitch me for new shows using the same engine.
Output is a Notion database with columns for guest name, anchor podcast, episode link, platforms, audience, tier, and collab angle.
Step 2: Map the data flow on paper before anyone writes a skill.
What goes in, what comes out, what each stage hands to the next. If you can’t draw it, you can’t build it. This is the clarity-of-thought work, and it is most of the job.
Step 3: Build the context files first.
Context files are the rules and reference data every skill consults while it runs. Skip them and your skills produce inconsistent garbage. She built five context files before writing a single skill:
- Anchor podcast list
- Which platforms qualify
- Minimum thresholds per platform
- Tiering rubric
- Outreach angle per platform type
Step 4: Build one skill at a time. Test each in isolation.
She built 6 skills and tested each one before moving to the next.
- RSS scraper
- Guest extractor
- Platform researcher
- Outreach builder
- Notion database writer
- Orchestrator that chains the first five in sequence
Build it, test against real data, confirm the output is correct, then move to the next. Composing skills that were never tested alone is how you end up debugging five things at once.
Step 5: Make them record a Loom on every skill.
This is a critical part of the process for confirming your team understands what they built. For each skill, she recorded a short Loom walking through it running on real data.
Explaining what you built forces you to actually understand it. I watched all six on 2x and either approved or sent back for fixes in minutes. The Loom is the verification layer and the learning layer at the same time.
And there you have it.
In just a few days she built 6 skills that run without me. I now have an EA who can now build the next 10 the same way.
Pick one manual workflow this week and hand it to someone on your team instead of building it yourself. Reply and tell me which workflow you’re handing off first. I read every response.
- Corey
A few notes/tweets/cool things sourced from the AI community.
1)
A wise reminder.
2)
Someone is going to turn one of these into a billion-dollar company.
3)
Build things that save or make businesses money. Simple.
Thanks for reading the Corey's Notes newsletter. I'd appreciate it if you sent it to a friend that might find it interesting.
Be sure to check out t he Build With AI podcast on Apple, Spotify, and YouTube. If you're a non-technical entrepreneur who wants to learn how to integrate AI into your business, you'll love it.
Be back next week.
-Corey
