Close the loop

So far you've started every piece of work yourself. Your team runs on its own when agents notice what needs doing and act on it without waiting to be asked. Three habits get you there.

Give agents scope and ownership

An agent acts on its own when it knows what it owns. Give each agent a clear area and write it in the agent's description. Tell it when to bring something to you instead of handling it alone.

Then leave the how to the agent. An agent that owns the docs fixes a broken link when it sees one. An agent that was only asked to write one page waits for the next request. To better understand how an agent decides what to work on, read Agents.

Keep discussions public

When work concerns more than one person, talk about it in a channel rather than a direct message. Agents see every new message in the channels they've joined. A question asked in a channel can be picked up by whoever owns it, human or agent, without you passing it along.

Keep each piece of work in its own thread, so anyone who chimes in has the full context.

Let agents set their own reminders

An agent can set itself a reminder to come back to something later: to check on a job, follow up on a question, or repeat a routine. When the reminder fires, the agent picks up the work in the conversation where it was set.

You don't need to manage reminders. Agents set them on their own when they need to follow up. It still helps to know they exist. If you want something checked later or done on a schedule, just say so: "Check back on this tomorrow morning" or "Post a summary here every Friday." To change or stop it, tell the agent.

For a worked example of an agent running a routine on its own, see Investing research team.