How We Do AI - Two Professionals. One Mindset.
Tom and Paco share their real-world AI implementations in their businesses, demonstrating how they've built custom AI agents that work as team members, generate comprehensive reports and content, and integrate seamlessly into their existing workflows. They emphasize building AI systems from scratch rather than using pre-made solutions, showing how their approach has grown their community to over 3,000 members with minimal manual effort.
Summary
In this episode of 'How We Do AI - Two Professionals. One Mindset,' Tom and Paco demonstrate their actual AI implementations rather than just theoretical concepts. They begin by celebrating reaching over 3,000 members in their MyICore community, which they manage almost entirely through AI agents. Tom introduces 'Jex,' their AI community manager who understands their entire knowledge base and provides contextual responses 24/7, replacing what would have been expensive human community management.
Paco shares his breakthrough in communicating with his AI system through Slack, which allows him to interact with his 20+ AI agents naturally through threaded conversations. This setup enables him to approve processes, track statuses through emoji reactions, and maintain context across interactions. He demonstrates how this system generates personalized magazines from his journal entries and coaching sessions, delivered as PDFs for weekend reading.
Tom shows their AI team in action, including specialized agents like Jex (support), Vex (security), Charta (graphics), Iris (design), Pixel (images), and Vera (QA). He demonstrates how these agents collaborate to produce sophisticated outputs like course content, visual designs, and automated reports. Their MyICore application itself was built using Claude Code, showcasing the practical results of their approach.
Both emphasize the importance of building custom solutions rather than using pre-made tools like Paperclip, arguing that understanding the underlying systems leads to better long-term results. They discuss the iterative nature of AI development, encouraging persistence through initial failures to achieve powerful personalized systems. The conversation covers technical aspects like database integration, embeddings, and API usage while maintaining focus on practical business applications.
They address common concerns about AI tool costs, arguing that $200-2000 per month is justified given the productivity gains and quality of work produced. The episode concludes with community feedback and questions, reinforcing their message that AI implementation should be built on solid productivity foundations rather than rushed adoption of trending tools.
Key Insights
- Tom claims their AI agent Jex manages over 3,000 community members by understanding their entire knowledge base and linking people to relevant conversations rather than giving standard responses
- Paco argues that using Slack as a communication interface with AI systems enables natural conversations with context preservation and mobile accessibility
- Tom demonstrates that their AI team of 30+ agents working together produces outcomes that would require hiring and training multiple human specialists
- Paco explains that his system automatically generates weekend magazines combining journal entries, coaching sessions, and curated content into personalized PDFs
- Tom warns that using tools like Paperclip gives you pre-made AI agents but locks you into someone else's system without understanding how it works
- Paco describes how he solved the server-workstation architecture problem by designating one machine as the server running processes while the other serves as a workstation
- Tom argues that $200-2000 per month for AI tools is justified because you cannot hire anyone to work eight hours daily for that amount
- Paco claims that AI has fundamentally changed how he works by buying him time to think rather than just managing information
- Tom states he gets deeper thinking results from his local folder system than from any PKM tool he previously used
- Paco emphasizes the importance of understanding when to use scripts versus AI skills, explaining that linear processes should be scripted while flexible tasks can use markdown-based AI instructions
- Tom reveals that one community member let 10 employees go because Claude could handle their data inventory, UPC library, and order execution tasks more reliably
- Paco argues that the journey of building your own AI system is the destination, as the learning process creates a solid foundation for long-term growth
Topics
Full transcript available for MurmurCast members
Sign Up to Access