AI Agents

    What Does an AI Chief of Staff Do? A Founder's Guide

    James KillickJames Killick22 Nov 20267 min read
    AI chief of staffAI agentsfounders

    Short answer
    An AI chief of staff is an AI setup that does your admin and prep work: it triages your inbox, manages your calendar, writes your briefs and chases loose ends. It prepares. You decide. A person should approve anything that goes to a client, moves money or gets signed.

    If you're a founder, you probably do the work of three people. You also do the admin of two more.

    An AI chief of staff takes the admin off you. Mine does. Here's what that looks like on a normal day, and the lines I never let it cross.

    What is an AI chief of staff?

    It's the AI team that sits between you and your day. It reads what comes in, organises it, prepares what you need and puts decisions in front of you. It doesn't replace your judgement.

    A human chief of staff runs the founder's day so the founder can spend time on the big calls. An AI version does a slice of that work. It's not one chatbot. It's a set of routines and instructions that run inside your AI operating system.

    What does an AI chief of staff do every day?

    It handles five jobs: inbox triage, calendar, meeting prep, follow-ups and daily briefs.

    Here's the day in order:

    • Inbox triage. One engine reads everything that arrives and sorts it once. Does this need me? Can someone else own it? Can the AI handle it? Should it just be filed? Is it junk? Every sort leaves a record I can check and mark right or wrong, so it gets better. Important things always come to me.
    • Calendar. It checks your diary before anything gets offered or booked. It knows your blocked time.
    • Meeting prep. Before a call, it pulls what you've said to that person before and what's still open.
    • Follow-ups. It tracks promises, yours and theirs, and flags what's slipping.
    • A morning brief. Short and clear: what's on today, what needs a decision, what's waiting on someone else. Mine is a video. Yours can be one page of text.

    None of that is magic. It's the work a good assistant does. The difference is it runs while you sleep and doesn't forget.

    Here's what a real one looks like. Mine is a short video. Every weekday at 5am it lands on my phone, narrated by an AI voice, and it runs about a minute or two. A second one lands at 4:30pm.

    The morning video covers three things:

    1. What needs me today. The few items only I can decide, up front.
    2. What the system already handled overnight. So I know what's done and don't go looking.
    3. What my day looks like. My meetings, any time I'm away, today's goal and how many hours I have.

    The afternoon one is about wins, meeting outcomes, what's still open and tomorrow. It counts goals moved, not tasks ticked.

    I know my day before I open anything else. It works because it's short and it leads with the decisions.

    What does a morning brief look like?

    A good brief is short and leads with the decisions you need to make. Here's a made-up layout you can copy, whether you read it or watch it.

    This is a template, not a real brief. Keep each part to a few lines.

    1. Today. Your meetings and blocked time, in order.
    2. Decisions. The two or three things only you can decide, each with a draft answer ready.
    3. Replies waiting. Emails that need you, with a drafted reply for each.
    4. Waiting on others. What you asked for and haven't got yet.
    5. Slipping. Anything you promised that's close to late.

    Say you run a ten-person design studio. Your brief might show a client call at 10, a quote that needs your yes and a supplier who hasn't replied in five days. You read it in a few minutes and know your day.

    The test of a good brief is simple. Can you act on it without opening your inbox? If not, tighten it.

    What should an AI chief of staff never do alone?

    Never let it send to clients, move money or sign anything on its own. Prepare, yes. Act on anything important, only with a person's yes.

    This is the most important part. My rules are simple:

    • It can draft. A person approves anything that goes to a client.
    • It can queue a payment task. It can't pay. I do that myself.
    • It can suggest a time. It has to check my diary first.
    • It can flag a mess. It can't permanently delete anything.

    Approval scales with how much the item matters. The more important it is, the more certain it is that I see it first. Anything about money, anything new and anything sensitive always waits for me.

    Australia's Guidance for AI adoption calls this maintaining human control. It's one of six essential practices. I'd put it first.

    What mistakes do founders make with one?

    Most founders give it too much power too fast, then stop checking its work.

    • Letting it send on day one. Start with drafts. Earn trust first.
    • Vague instructions. "Handle my inbox" isn't a job. "Sort new email into reply, wait and noise" is.
    • No review. Read what it did for the first few weeks. Fix the instructions when it gets something wrong.
    • Starting with five jobs. Start with one.
    • Forgetting the rules change. As the work proves itself, you can loosen a gate. Do it on purpose, not by drift.

    How is an AI chief of staff different from an AI agent?

    An agent is a system that decides its own steps to reach a goal. A chief of staff is a set of agents and routines with clear lanes and clear limits.

    Anthropic's guide on building effective agents separates workflows from agents. Workflows follow a set path. Agents choose their own path and tools as they go. A good chief of staff uses both. Routine jobs run as workflows. Open-ended jobs use an agent, and a person checks the result.

    If you want a plain-English view of agents, read what is agentic AI.

    What do you need to build one?

    Three things: access to your tools, written instructions about how you work, and a rule for what needs your yes.

    • Access. Connectors to email, calendar and files. The Model Context Protocol is the open standard many tools use for this. My plain guide is MCP for non-developers.
    • Instructions. How you write, who matters, what's urgent. Write it once and keep it current.
    • Gates. What needs approval. Start strict. Loosen it only when the work proves itself.

    Start with one job. I'd start with the morning brief, because it only reads. It can't break anything. That's the Power of One at work: one focused win at a time.

    What if I want one built for me?

    If you're a founder who's too busy to build this, The AI Orchestrators helps founders build an AI chief of staff from their own method.

    FAQ

    Can an AI chief of staff replace a human assistant?
    It does a slice of the work: sorting, drafting, prep and reminders. It doesn't replace judgement or relationships. Think of it as a first pass that a person finishes.

    Is it safe to give an AI access to my email and calendar?
    It's a real risk, so treat it like one. Give access only to what the job needs, keep a human yes on anything that goes to a client or moves money and read the terms of the plan you use.

    What's the first job to give it?
    The morning brief. It only reads and summarises, so it can't send, pay or delete anything.

    Do I need to code?
    No. You need clear written instructions, access to your tools and a rule for what needs your approval.

    Your next step

    Write down your five most repeated admin jobs. Then watch my video on building your AI operating system with Claude. It shows the four parts I build on: the brain, the team, the hands and the heartbeat. More at youtube.com/@james-killick.

    About the Author

    James Killick
    James Killick

    The AI Orchestrator

    AI Orchestrator and entrepreneur with 10+ years building digital products. Helping $1M+ business owners scale with AI systems, automation, and implementation through The AI Orchestrators, Devwiz, and Njin.

    Related Posts