In episode 2 of Redacted, David and Taylor get into the messy middle of building with AI inside a real business.
After compressing Offline from a 34-person team to a much smaller operating crew, AI stopped being a fun experiment and became a necessity. This episode is about what that actually looks like: rebuilding lead-gen workflows, trying to make HubSpot reflect reality, keeping AI agents alive like Tamagotchis, and testing whether Claude Code can help generate a real shareholder update from scattered company data.
The takeaway: AI can do a lot more than send one email, but only if you teach it where the business actually lives.
Timestamps
00:00 — Welcome back to Redacted
00:36 — “Why should people even listen to us?”
02:07 — How Offline compressed from 34 employees to a tiny team
03:58 — The original AI lead-gen and CRM automation experiments
06:27 — Translating complicated human workflows into AI systems
07:00 — AI-powered inbound lead classification and HubSpot automation
08:09 — Using RSS feeds and AI to discover restaurant leads
08:52 — Where CRM automation becomes extremely difficult
10:16 — Why AI workflows become “Tamagotchis”
11:10 — Taylor’s multi-agent HubSpot cleanup system
12:18 — Why clean CRM data matters more than people think
13:08 — The tradeoff between API costs and AI workflow complexity
14:11 — The “unlock” of passing reasoning between LLMs
15:11 — Turning AI reasoning into actual HubSpot actions
16:41 — The AI existential crisis: “This will never work”
17:50 — Wanting AI systems that can simply ask questions when stuck
18:12 — PTSD from n8n and broken workflows
19:14 — Teaching AI systems to learn from mistakes
20:16 — The tradeoffs between local AI systems and n8n
21:59 — “Every CRM is chronically out of date”
23:57 — Why clean data is foundational for AI outbound sales
25:11 — Bottom-up vs top-down AI automation strategies
26:40 — The challenge of defining “objective reality” in business data
27:14 — David’s AI-generated shareholder update workflow
28:08 — Building “super skills” with Claude Code
29:18 — Mapping every data source needed for shareholder updates
31:00 — AI reading financials, GitHub commits, payroll, and board notes
32:14 — “I could’ve just written the shareholder update myself”
33:08 — How the shareholder update skill is structured
34:03 — The first AI-generated shareholder update draft
35:00 — AI recognizing profitability and company milestones automatically
35:40 — AI analyzing GitHub commits and engineering work
36:35 — Why this kind of context-heavy AI work matters
37:16 — Final thoughts and what’s next for Redacted
Where to Find David:
LinkedIn:
https://www.linkedin.com/in/davidshaner/Where to Find Taylor:
LinkedIn:
https://www.linkedin.com/in/taylorcotner/More about Offline:
https://www.linkedin.com/company/offline-media-inc-/---
This episode of Redacted is hosted by David Shaner and Taylor Cotner, and presented and produced by NC Tweener Fund.
We couldn’t share posts like this without our amazing sponsors:
Platinum:
NC IDEA: https://ncidea.org
Gold Sponsors:
- Balentine: https://www.balentine.com/triangle-entrepreneurs
- EisnerAmpner: https://www.eisneramper.com
- Robinson Bradshaw: https://www.robinsonbradshaw.com
Silver Sponsors:
- Automated Consulting Group: https://automated.co
- Bank of America: https://business.bofa.com/en-us/content/technology-industry-group.html