6 Lessons I Learned From Installing AI in a $10M Business

6 Lessons I Learned From Installing AI in a $10M Business

More

Summary

Liam Ottley shares six lessons from a week spent installing AI into an eight-figure custom-products business run by Isaac, known as The Sourcing Guy. The team built a system that generates custom sales pitches, with product mock-ups, spec sheets, and videos in a client’s own branding, that staff can trigger by messaging an AI in Slack.

The results were mixed. A real client was impressed by the output but also said parts felt like AI, pointing to the copywriting and overly flashy video effects. In two days the team also burned through $100 in AI credits. Ottley’s first lesson is that AI handles the first 80 percent while human taste decides the last 20 percent, so teams should document their best person’s rules and examples. He also advises fixing one bottleneck rather than everything at once, in this case pitching, which consumed roughly one full-time person’s hours.

Other lessons cover data quality, which sets the ceiling for any AI project. Only about 43 of roughly 1,000 catalog products had everything the system needed, and the AI sometimes used the wrong logo. The team consolidated data into a Supabase database, added a logo-checking step, and focused on teaching staff to maintain the system themselves.


📺 Source: Liam Ottley · Published November 20, 2026
🏷️ Format: Workflow Case Study

1 Item

Channels

1 Item

People