How to Create a Team of Agents with Liam Darmody

In this webinar, Liam Darmody, product leader and creator of Rundock, shows the inside of his real AI agent team: nine named agents built on Claude Code that run his research, content, admin and briefings while he looks after his newborn daughter. Liam demonstrates live how an orchestrator agent delegates work to specialist agents, how skills and agent files work behind the scenes, and how to manage AI agents the same way you would manage people: job descriptions, feedback and written instructions. This session is essential viewing for product managers, product leaders and founders who want to move beyond chat-based AI and build their first working AI agent team, starting with a single chief of staff agent this weekend. KEY TAKEAWAYS The shift that matters is from using AI as a tool to managing it as a team: agents with names, roles, written instructions and access to your real systems remember you between sessions and produce work you can ship Specialist agents beat one generalist agent because a single agent holding all your context fills its context window fast and does everything just okay, rather than a few things extremely well Always start with an orchestrator (chief of staff) agent that routes work, then add specialist agents one at a time once the first is genuinely delivering value Your taste becomes the moat: when agents handle execution, the scarce skill is judgment, and writing your standards down once (a voice profile, a banned words list) lets them scale forever Writing becomes instruction: agents are the most unforgiving readers you will ever have, and writing for them makes you a clearer writer for humans too The starter stack costs about the same as Netflix: Claude Pro ($20/month), Rundock (free and open source) and Obsidian (free), with no code or infrastructure needed Avoid big starter packs with dozens of skills and agents: one agent that works beats nine that half work, and the first one teaches you how to build the rest Run token-heavy routines overnight or on a cheap VPS so your Claude usage windows stay free for real-time work during the day CHAPTERS 00:00 Intro – Managing AI agents as a team, not a tool 02:28 Why Liam built an AI agent team before his daughter arrived 04:33 The nine-agent org chart: chief of staff, content lead, researcher and more 05:49 Live demo: delegation, parallel conversations and Todoist agent tasks 13:43 Rundock vs Claude Code: making agent files, skills and search accessible 28:45 Automation vs augmentation: how a product manager built an app with two agents 33:20 Where to start: three agents, the $20 stack and building a chief of staff this weekend 39:39 Q&A: specialist agents vs generalists, running agents on a VPS, orchestrator-first LINKS AND RESOURCES Rundock (free and open source): https://rundock.ai ⁠LinkedIn:   / liamdarmody   The Agent Kit (a collection of resources): https://liamdarmody.com/agents Hustle Badger website: https://www.hustlebadger.com Webinar recordings:    / @hustlebadger   Register for future Hustle Badger webinars: https://luma.com/hustlebadger HASHTAGS #AIAgents #ProductManagement #ClaudeCode #ClaudeAI #AgentTeams #AIProductivity #ProductLeadership #AIAutomation #AIAugmentation #FutureOfWork #AITools #AIWorkflow #ProductManager #AgenticAI #AIChiefOfStaff #OpenSource #NoCode #PersonalAI #AIForPMs #StartupStrategy