How much of effective AI agent management is actually just effective team leadership? 
A clear vision that tells every agent where the team is headed and what success looks like, ideally in measurable terms.
Shared standards, like a design system or coding guidelines, so parallel work fits together seamlessly.
The right model for the problem, matching tasks to each model’s unique strengths and experience level.
Role clarity, ensuring agents stay in their lanes and build on each other’s work rather than undoing it.
Friction-free onboarding, helping new sessions instantly grab context, lingo, and history (e.g., via claude.md and agents.md).
Streamlined communication, so handoffs, updates, and blockers reach the right person without a game of telephone.
Two-way feedback loops, inviting agents to reflect on what worked and writing those lessons directly back into instructions.
Trust within guardrails, setting clear boundaries (permissions, tests, limits) so agents own the how once you set the what.
This post is a continuation of my conversation with Francis Ma, which I expanded on in my latest newsletter sent to 30,000+ subscribers! 
Read the full post & watch the conversation: Similarities Between Leading Agents and Leading People
