What Does a Chief of Staff AI Agent Do?

On October 7, 2026, I built a chief of staff AI agent live with Women Applying AI, and the builder asked what the AI agent should take care of. I answered with the jobs below, in one pass. The chief of staff that audits a roster of 25 AI agents every Friday is a different one, covered in Chief of Staff AI Agent Friday Audit.

From the recording, 9:01. So if you're looking at my screen, you'll see something that says, what should this agent take care of, question and mark.

Which jobs did I hand to my chief of staff AI agent?

I dictated the list at 11:54, and the article on how to dictate goals to an AI agent prints the full block, so I don't reprint it here. These are the jobs, one at a time.

• Relationships (12:02). My quarterly goal was to "maintain great relationships with the Women Applying AI community," so I asked the AI agent to follow up on Slack messages and emails and send useful information through Slack. In the demo I paused connecting Slack and kept building (38:05).
• Calendar (12:17). I asked it to "manage my calendar and not miss anything." In the demo I gave it calendar access with a one-click login (37:54).
• Replies before the day ends (12:23). I asked it to help me "not miss replying to critical emails at the end of day." Later in the build it wrote a skill called an end of day sweep (40:41).
• Dashboard (12:27). I asked for "a dashboard so that I can visually see things that I need to take action on right away." I did not tell it how to give me that information, and I said it would ask me clarifying questions in a moment (13:05).
• Proactive (12:44). I told it to "be proactive and make smart decisions."
• Honest (12:46). I added, "I shouldn't have to say this, but I will be honest with me." That is how the transcript has it.

From the recording, 13:05. Now you'll notice I didn't tell my agent how to give me this information outside of a dashboard. So it'll ask me clarifying questions in a moment.

The page on seeing what needs your action today picks up that dashboard question.

What does each of those jobs need the AI agent to reach?

Every job above needs the AI agent to reach something, and connectors are where I give that access.

From the recording, 26:07. Connectors are what make agents work. This is what helps them to move from one part of a workflow to the next part.

I said that if my AI agent is going to decide to check the calendar, check email, and create an agenda for a meeting or send a follow-up recap, it needs to have access to systems (26:39).

Meeting prep came later, when I named the connectors I like. I said I typically have my chief of staff AI agent reviewing my calendar and helping me decide which meetings need more prep and which need follow-up actions (30:04). For an organization I am meeting, I ask these:

From the recording, 30:19. If I'm meeting with an organization, have they been in the press recently? Have they shared a press release? Is there anything they're celebrating?

The three research connectors I named for that are Firecrawl, Exa and Apify, which I called "three that I typically use" (30:46).

If you do not want to connect your real inbox yet, I said you can select a dummy account to practice with (20:06). A way to test with a made-up inbox is in How Do I Test a Chief of Staff AI Agent With a Made-Up Inbox?.

Which skills did my AI agent write for the list?

Partway through the build, I asked for two skills and told the AI agent to ask me three or four questions first (36:56).

From the recording, 36:56. What I typically do is I start off by saying I want to create two skills for the COS agent.

From the recording, 37:10. Ask me three or four questions to customize my agent's skills.

From the recording, 40:41. It's created two skills for me, an end of day sweep as well as a relationship tracker.

I did not read out what is inside either skill. What I showed was where to click into agent skills to see step by step what they are (41:14). A page about reading and editing skills is How to Create Skills for an AI Agent.

What the AI agent can reach in Gumloop

The builder in this session is Gumloop, and On my relationship to it, I said this:

From the recording, 33:22. I don't work for Gumloop. I found this tool over the summer through Lenny's podcast, and I started using it and have since made myself an ambassador for them. I've created a relationship with Gumloop. I am not an employee.

More on that is in Do you work for Gumloop? (the answer from the session)

Checked on 2026-11-05 against Gumloop's agent documentation, which describes connectors as the integrations an AI agent uses, such as Gmail, Salesforce and Slack, and says the connectors you add determine what it can access. On October 7, 2026 I said there are "over 300 connectors" you can use inside Gumloop (30:55).

What goes wrong

The one thing I named going wrong is leaving the AI agent short, and I used a new hire to say it. I said that if I leave out critical information and restrict access, that new hire isn't going to achieve as much (6:05).

What the recording does not show: the contents of the two skills, the finished dashboard, or my own day-to-day chief of staff AI agent, which I chose not to open on screen (55:15).

Where this comes from

This article grows from the session I taught live with Women Applying AI on October 7, 2026. The quoted passages are my own words, taken verbatim from the recording. The lists and the steps around them were drafted by my team's AI agents from those words and my published pages, and checked against the recording before publishing. Platform details that have a public source carry a dated Checked on line.

Build Your First AI Agent in One Hour: the full session with the clips, the prompts and the transcript.

Related reading: How Do You Dictate Goals to an AI Agent?, Chief of Staff AI Agent Friday Audit, What Should Your First AI Agent Be?, and the Chief of Staff session page.