AI School · Free replays

AI Webinars

Every live AI webinar we have run with Benza Maman and Max Marine: the full recording, every question from the audience and the answer, and the slides.

Free to watch. Two quick steps.

Sign in, then opt in to occasional AI School updates. That unlocks every replay, the full Q&A and the slides.

Unlock above to watch

What's covered

  • Context is the difference between a generic answer and a useful one. Benza's analogy: a genius you meet on the street can write your marketing email, but only well if you hand them a sheet covering who you are, your strengths and weaknesses, your tone, and past campaigns.
  • Build in layers. Action skills are written instructions for one task, orchestrator skills run several action skills in sequence, department heads own a vertical (marketing, content, sales, finance), and one CEO skill delegates to the right department. You only ever talk to the CEO.
  • Given full context, a vague prompt like "let's make more money, make it easy" produces a real plan. The demo CEO agent picked an untapped audience (spirituality and meditation), designed a digital offer, and built the GoHighLevel automation, email copy in Benza's voice, a landing page, and the Stripe product and link.
  • Run recurring work as scheduled routines. A "Butler" orchestrator scrapes iMessage, email, WhatsApp, and Slack three times a day, checks them against his goals and calendar, blocks out his schedule, and texts him only when something is pressing.
  • Perfect each action skill before you chain them. Get the WhatsApp scrape, the email scrape, and the calendar step each working alone, then let the orchestrator run them. Go easy on WhatsApp outreach volume, since bulk messaging risks the account.
  • A separate review agent with fresh context double-checks every output. Benza says this cuts hallucinations by roughly 70%, and you can raise the quality-assurance level for higher-stakes work at the cost of speed and usage.
  • Costs stay predictable if the system runs inside a Claude subscription rather than API tokens. Benza uses the $200 plan, suggests $100 as the sweet spot, and says starting on $20 works before you upgrade.
  • The system is not locked to one vendor. The same skills setup runs in Codex or other agent tools, though Benza prefers Claude Code as the harness and finds Claude's reasoning stronger for tasks like writing Google Docs.
  • Framing: context engineering, and why agents are "motivated interns" that need context
  • Benza's background and why adopting agents now matters: the before and after of a solopreneur's week
  • What an AI operating system is: context as the foundation, 10 Claude instances as 10x output, and L3Harris time-savings numbers
  • Architecture: action skills, orchestrator skills, department heads, and the CEO agent
  • Demo: the Butler comms and schedule agent, the agentic dashboard, verticals, and the review agent
  • Demo: the CEO agent with context turns "make more money" into an offer, GoHighLevel workflow, emails, landing page, and Stripe checkout
  • Self-improving ad loops, AI UGC video ads through Higgsfield, and Claude running content and YouTube
  • Common objections: course quality, "I'm not technical," and "I don't have time"
  • Q&A opens: orchestrating skills, the Butler prompt, Outlook, Claude Code versus Cowork, and the video editing tool
  • Q&A: Claude versus GPT and Codex, monthly cost and ROI, Gemini
  • Q&A: who this suits beyond content creators, starting a business agent-first
  • Q&A: which Claude plan, Grokbot, letting Claude reply for you, and rising costs
  • Q&A: supervision and quality-assurance layers