I just finished watching Nate’s AIOS course For those of you who’ve made the leap and implemented a Company AIOS 1) What did you have to redesign so it could scale across roles and teams instead of staying a personal system? 2) What are the 2–3 best practices that made it work at team scale? Looking for concrete setups / resources / examples.

If you could automate ONE thing in your business tomorrow — What would it be? (Mine changed everything when I finally set it up)

Every morning at 8AM, this workflow runs silently in the background. Here's what it does: It pulls every active case from Google Sheets. If a client was already contacted today — it skips them. No double emails. Then it checks three things: Filing deadline in 7 days? The attorney gets a Slack alert. Filing deadline in 3 days? Critical Slack alert. Escalate now. No contact in 14+ days? An AI writes and sends a check-in email on behalf of the attorney. Client ghosting on documents for 7+ days? A separate AI-written reminder goes out — polite, firm, urgent. Every email is generated by Llama 3.3 via Groq. Warm tone. Professional. Sounds like the attorney actually wrote it. After each send, the sheet logs "Contacted" + the exact timestamp. The firm stops chasing. The system chases for them. Tools: n8n · Groq (LLaMA 3.3) · Gmail · Slack · Google Sheets

I am building an AI native restaurant management OS and one of the AI automation use cases I thought to try was to automate the creation of weekly blog posts to build up my SEO - and man what a feeling this was! Here's how it works, every Monday, it kicks off a 6-stage agent sequence: 1. Researcher Agent — generates keyword candidates, pulls live Google SERP data (serper.dev), picks the best opportunity, and drafts an outline 2. Writer Agent — writes a 1,400-word post in our brand voice, with hard guardrails against competitor mentions and clichéd phrasing 3. SEO Optimizer Agent — generates meta title, description, schema markup, and internal link suggestions automatically 4. Illustrator Agent — plans and generates custom images using GPT Image, uploads them to storage. This is the area that I think needs most work because the images generated are very much AI looking. 5. Compiler Agent — splices the images into the article and renders the final HTML 6. Publisher Agent — goes live At 3 points in the pipeline (brief, draft, final), I get an email with an approve/reject link (served by Resend). One click either advances the post or kills the run. The whole thing runs on pg_cron in Postgres. One SQL line schedules the entire weekly content operation. One post per week, basically on autopilot. Now could this be done simpler? Probably, but I wanted to try out the agent hand-offs and human in the loop. Also this was built on Supabase Edge Functions not Claude routines. If you're curious to look at the output here is a link https://www.withbagel.com/blog

I sat down with Devin Kearns, co-founder & CEO of Custom AI Studio, to break down what it actually takes to build an AI agency with real enterprise value, not just another lifestyle business. We get into why most AI work being sold today won't survive 2027, why the mid-market is the prime opportunity (not SMBs or enterprises), the 11 ways AI experts are actually making money right now, how to position with frameworks instead of being just another vendor, and the five things Devin wishes he knew sooner. If you're building, running, or thinking about starting an AI agency, this is the strategic conversation I wish I'd had two years ago.