EP 081 · 117 MIN 37 SKILLS. HE FORGETS WHAT HALF OF THEM DO

How We Use AI to Run Our Businesses With Skills, Agents & MCPs

September 6, 2026117 min Rory Flynn & Drew Brucker

The short version

Episode 81 opens up Drew and Rory's personal AI operating systems for running solo businesses. Drew walks through 37 Claude skills, including a 21-layer voice-writing profile and a "diff" skill that trains on every email he actually sends, while Rory runs his whole business, from client CRM to LinkedIn-performance prediction to red-teamed contracts, out of Claude connected to Notion. They also cover Fable 5.1, the incoming Astra release, and the Riverside MCP that packages this podcast automatically.

Key takeaways

  1. 01

    Skills compound: Drew's "diff" skill compares every draft against what he actually sent, so his AI voice profile keeps getting closer to how he really writes.

  2. 02

    Before building a new skill, ask your AI to check what you already have: Drew's "adopt" skill compares a new idea against his existing system first.

  3. 03

    Route inbound email through an AI CRM: Rory's setup classifies leads twice a day and pushes real opportunities straight into a Notion pipeline.

  4. 04

    Have AI red-team your own contracts and operating agreements: Rory ran his LLC agreement through an adversarial "agent swarm" looking for holes before signing.

  5. 05

    Connect the tools you already have before building anything fancier: unapproved connectors, not missing AI skill, is the actual bottleneck at most companies.

  6. 06

    Give AI a business problem instead of a tool request: Rory replaced a paid NVIDIA analytics tool entirely once Claude could just build the same agent.

  7. 07

    Use AI to use AI: when you're stuck on a prompt, a workflow, or troubleshooting a tool, ask the model itself before giving up.

Receipts · said on the record

“When I'm doing things with Claude and I need to read the threads, or I'm doing something with Gemini and I need to read it, I feel like I read the equivalent of a book a day.”
Rory Flynn The daily reading load of directing AI agents Play from 16:39
“Some people, right, really believe, maybe maybe even myself included, that like biologically, humans actually have all the things necessary, generally speaking, to survive but also thrive, right?”
Drew Brucker Whether humans need AI integrated into their biology to thrive Play from 19:08
“It's almost like what's the what's the optimal place? It just injects it into your brain. Like there's a like a psychic transfer of information that is instantaneous, you know?”
Drew Brucker The mental cost of staying on top of long AI-agent threads Play from 16:59
“People are just not buying enterprise anymore because things change so quickly. I'd rather give ten people just the max plan on Claude than try to get a hundred seats for everyone.”
Rory Flynn Why he'd skip enterprise licensing for a handful of max-tier seats Play from 26:35
“We had it red team it for like different scenarios. So like I would have it like on negative scenarios, on positive scenarios, like run like an agent swarm just to find the holes against the agreement.”
Rory Flynn Red-teaming their own LLC operating agreement with Claude Play from 1:16:41

Chapters

  1. 0:00:00 Two sleep-deprived founders, no guest
  2. 0:01:16 Martinis, $25 cocktails and vintage Bud Ice
  3. 0:07:51 Football is back, and so are the AI models
  4. 0:09:09 Fable 5.1, Astra and the next AI model wave
  5. 0:11:43 Why 3D and spatial AI could be next
  6. 0:14:55 The hidden cost of AI: reading a book a day
  7. 0:18:09 Biology vs. technology and the Matrix problem
  8. 0:22:10 Why ChatGPT feels useless to some people
  9. 0:23:32 Why enterprise AI adoption stalls
  10. 0:28:16 Connectors turn scattered context into leverage
  11. 0:30:00 Inside Drew’s 37-skill AI operating system
  12. 0:33:48 “Diff”: train AI on every edit you make
  13. 0:39:52 “Adopt”: stop rebuilding what already exists
  14. 0:42:59 Riverside MCP automates podcast packaging
  15. 0:49:04 Let AI write your Nano Banana edit prompts
  16. 0:54:17 When AI skills become a routing system
  17. 0:56:24 Turn client research into repeatable reports
  18. 1:00:23 The orchestrator problem: too many AI skills
  19. 1:02:32 Rory’s Claude + Notion business operating system
  20. 1:07:32 Turn inbound opportunities into an AI CRM
  21. 1:13:22 Building an AI finance department
  22. 1:15:48 Using Claude to set up and run a business
  23. 1:16:30 Red-team contracts and decisions with AI agents
  24. 1:21:45 Can AI predict LinkedIn post performance?
  25. 1:30:56 Build a personalized AI learning system
  26. 1:32:19 Turn great ads into reusable AI templates
  27. 1:36:39 Build AI pipelines once, then strip them for parts
  28. 1:37:09 Claude interviewed Gemini for six hours
  29. 1:38:45 How an AI orchestrator reuses your entire stack
  30. 1:43:01 AI agents as your finance + ops department
  31. 1:44:31 Use AI to use AI: the biggest takeaways
  32. 1:48:14 Why people give up on AI too early
  33. 1:50:52 AI systems for your personal life
  34. 1:52:24 Using AI for fantasy football draft analysis
  35. 1:55:23 Final thoughts + what we should build next

Questions this episode answers

How many AI skills has Drew Brucker built for his business?

Drew has built 37 Claude skills, ranked S-tier through C-tier, including a 21-layer voice-writing profile, an email-drafting pipeline, and a "diff" skill that compares his sent messages against AI drafts to keep refining them.

How does Rory Flynn run his business with Claude and Notion?

Rory routes nearly everything, client CRM, invoicing, contracts, and daily briefs, through Claude connected to a single Notion workspace, with an "orchestrator" skill that knows his whole library of pipelines well enough to route any new request to the right one.

What is Drew Brucker's "adopt" skill?

It's a Claude skill Drew built for whenever he sees someone else's workflow or post online: he feeds it in and asks Claude to check whether he already has something similar, what's objectively better about the new one, and whether it's even relevant to him.

Can AI red-team a contract or business agreement?

Yes: Rory and Drew had Claude run their own LLC operating agreement through an adversarial "agent swarm" against negative and positive scenarios to find holes before signing, and Drew runs client MSAs through a similar check for risky clauses.

What's the biggest mistake people make when trying AI tools, according to Fast Hours?

Drew argues people give up too early: the first attempt at using AI for something new often takes longer, but every attempt after that gets significantly faster, and most people quit right before they'd see that payoff.

Show notes

To run a solo business on AI, build a system of small skills and agents that handle the admin, comms, and ops so your time goes to creating.

Every word Read the full transcript

If you liked this one