Build an AI Chief of Staff for Your Business

You do not need another chatbot. You need someone who already knows the business and tells you what matters before you ask.

It is 6:40 in the morning and you are already behind

Lena runs an eight-person interior fit-out firm in Manchester.

She opens her laptop before coffee. Forty-one unread emails. A supplier wants to move Thursday's delivery. A client who was keen last week has gone quiet. There is a quote she promised yesterday and did not send. A fitter needs a decision before 9 or the whole day stalls.

She is the founder. She is also the sales lead, the project manager, the accounts chaser, and, without ever applying for the job, the chief of staff.

That last role is the one nobody sees. It is the work around the work. And it is quietly eating her best hours.

Here is how your morning actually runs

You already do this job. You just do it by hand, in a dozen tabs, before you have woken up.

• Open the inbox. Scan for what is on fire.
• Jump to the calendar. Work out what you are walking into.
• Flick to the chat app. Catch the message you were tagged in.
• Open the task list. Remember the thing you promised.
• Hold all of it in your head. Decide what to do first. Start typing.

Then a call comes in and the whole stack falls over.

You are not disorganised. You are doing a second full-time job that has no title and no salary.

Every growing company eventually hires a chief of staff to carry it. The person who reads the room before the meeting, remembers who is waiting on what, and tells the boss the one thing that matters today.

In a small business, that person is you. At 6:40 a.m. Unpaid.

What changed in 2026 is that you can finally hand that job to something else.

What is an AI chief of staff?

An AI chief of staff is an agent that connects to your email, calendar and tasks, remembers how your business runs, and works on a schedule, so it can hand you a short briefing each morning instead of leaving you to dig for it.

Not a search box you have to visit. Something that shows up on its own with the three things that actually need you today.

The difference from a normal chatbot comes down to three things it has that a chatbot does not.

• Memory. It remembers your clients, your suppliers, your last decision, and how you like to reply. It gets sharper the more you use it. (We wrote about why this matters in the AI agent that remembers your business.)
• Tools. It can actually read your Gmail, open your Google Calendar, check your Slack or ClickUp, and draft a reply. It does the work inside the apps you already use, not in a separate window you copy and paste from.
• A schedule. It runs on a timer, a scheduled task some builders call a cron job, so it works while you sleep and reaches out at the right moment. This is the same idea as putting your AI in a loop, pointed at your morning.

A chatbot waits for you to ask. A chief of staff already knows the business and speaks first.

Chatbot or chief of staff? The gap is bigger than it sounds

Ask a plain chatbot "what should I do today" and it gives you a tidy, generic list, because it knows nothing about your Thursday delivery or your quiet client.

A chief of staff has already read the inbox and the calendar. So it says: the client from last week has not replied in six days, here is a draft nudge; the supplier moved Thursday, that clashes with the site visit, you need to call them; and the quote you promised is still in drafts.

Why this suddenly works in 2026

The idea of a robot assistant is old. What is new is that three things finally lined up.

Agents can now hold real memory. They can connect straight into everyday tools like Gmail, Google Calendar, Slack and ClickUp. And they can run on a schedule instead of only when you prompt them.

The last piece is cost. The frontier models that power these agents keep getting cheaper to leave running. Anthropic's top model, Fable 5, is being folded into its standard Max and Team plans, which is exactly the shift that makes an always-on assistant affordable for a small firm rather than a luxury for a funded startup.

You can see how fast this is spreading in who is teaching it. A single 45 minute walkthrough on building a "24/7 AI chief of staff", by the product writer Peter Yang, has already been watched more than 36,000 times. This is not a fringe experiment. It is becoming the default way owners run their day.

Isn't this just a fancier chatbot?

No, and the gap is the whole point.

A chatbot autocompletes. A chief of staff uses judgment built from memory.

Here is a real example that made the rounds this week. The product writer Aakash Gupta shared a chief-of-staff agent built by Jyothi, an AI builder, that went well past summarising email. It told its owner to make a specific colleague an ally. Separately, it warned her to loop certain people in before she sent a sensitive message, because it had already learned who those people were.

That is not a smarter inbox. That is an assistant with a memory of your working relationships, giving you the kind of advice a good human chief of staff would. The post has been bookmarked more than 2,000 times, which tells you how badly owners want this.

You will not start there. But that is the direction of travel, and the first version you build this week is the first step toward it.

How do I build an AI chief of staff for my small business?

You do not need code, and you do not need a framework. You need one clear job, your real tools connected, and a schedule. Here is the path most builders follow.

1. Give it one job first. Do not try to automate your whole life. Start with the morning briefing. One job, done well, that you feel the value of on day one.
2. Connect it to where the work lives. In Claude, that means Projects and Connectors for Gmail and Google Calendar. ChatGPT and Claude Cowork work the same way. Start read-only, so it can see your inbox before it can ever send from it.
3. Write its brief and its boundaries. Tell it who you are, what your business does, what you care about, and how you like to reply. Then tell it what it must never touch. Peter Yang keeps this in a short profile and a "soul" file with rules like "be resourceful before asking" and "do not talk about private things."
4. Build one skill and test it by hand. The rule from experienced builders is simple: build the skill, run it once, check the output is actually good, and only then automate it. Never schedule something you have not seen work.
5. Put it on a schedule. Once the briefing is good, set it to run every morning at, say, 7 a.m. And message you. This scheduled task is what turns a clever prompt into a real member of staff.
6. Give it memory so it compounds. Let it save what it learns, your regulars, your suppliers, the reply you approved, so next week's briefing is sharper than this week's.
7. Keep a human on anything that sends or spends. Let it draft, flag and suggest freely. Make it ask before it emails a client or moves money. That one boundary is what lets you trust it.

The actual morning-brief instruction can be this plain: "Review my calendar and recent emails and notes, then send me a short briefing with the top three things I need to act on today, my meetings, and anything waiting on a reply." That single line, connected to your real tools and run on a timer, is a working chief of staff.

If you already publish a weekly report with an agent, this is the daily cousin of the same idea.

What you will have in a week, a month, six months

Picture Lena again, but building this instead of drowning in it.

In a week, she has one thing: a calm briefing waiting at 7 a.m. No more scanning forty-one emails to find the three that matter.

In a month, it drafts the replies she keeps putting off, flags the client who has gone quiet before the deal dies, and reminds her about the quote still sitting in drafts.

In six months, it knows her suppliers, her repeat clients and her patterns. It tells her what is about to slip before she feels the panic. The 6:40 a.m. Second job is gone, handed to something that never forgets and never sleeps.

Same inbox. Same calendar. Same Lena. A completely different morning.

Your first step costs you fifteen minutes today: open Claude or a tool like Hermes, connect your calendar, and ask it for tomorrow's briefing. Read what it gives back. That is the moment the light goes on.