I Have Some Questions...
Erik and Justin unpack what an “AI orchestration layer” actually means when agents move from experiments into day-to-day operations. They focus on the practical shift from building tools to managing systems: mirroring the org chart with digital agents, defining who maintains them, and creating an auditing layer so leaders can trust performance at scale. 🧭 Conversation Highlights * Teams are quickly moving from a handful of agents to managing 5 to 10 agents per person, and that forces org design questions, not just tooling questions. * Justin frames the orchestration layer as translating real job responsibilities into AI agents, then stacking the necessary “maintenance” role to keep them current and connected. * An agent’s basic structure includes channels (where it can communicate), instructions/persona (its “job description”), skills (step-by-step processes written in plain language), plus memory and access * Auditing is still emerging: some systems show activity and conversational logs, but companies will need better frameworks to measure outcomes, effectiveness, and risk across many agents. 💡 Key Takeaways * The big change is managerial: leaders (and future roles) will oversee two mirrored systems, humans in the physical org chart and agents in the digital one. * Maintenance becomes its own discipline because agents rely on specific workflows, skills, knowledge files, tool integrations, and ongoing updates. * Agent development can be lower-friction than people expect because “skills” and “instructions” can be described in natural language rather than requiring traditional software engineering. * Trust at scale will depend on auditing: what agents did, how well they did it, and whether changes (like tool updates or memory behavior) quietly degrade performance. ❓ Questions That Mattered * Who should own agent maintenance when one person might end up responsible for dozens of digital entities? * What does an agent need in order to operate reliably (channels, instructions, skills, knowledge, memory) and how do those parts change over time? * Where does visibility come from today: can you audit outcomes and correctness, not just view that the agent “worked”? * How do you measure agent effectiveness in a way that’s actually accountable, like tracking nudges, accept rates, and task completion? 🗣️ Notable Quotes * “You need to start thinking about how you manage two mirrored org charts where you have for every position 5 to 10 different digital entities.” * “What you're talking about is a new job, a new role. It lives sort of in IT, it lives sort of in HR.” * “Agents are fairly new. Last year, 2025, the infrastructure for agents to work was being worked on. Now that infrastructure exists.” 🔗 Links & Resources * Listen To Other Episodes Co-Hosted With Justin [https://www.google.com/url?q=https://podcast.languageofleadership.io/categories/i-have-some-ai-questions-with-justin-coats/&sa=D&source=editors&ust=1780528759104935&usg=AOvVaw1Ae6zodJyXV8bCEIrb8Qo1]
170 episodios
Comentarios
0Sé la primera persona en comentar
¡Regístrate ahora y únete a la comunidad de I Have Some Questions...!