Skip to article

How to build an AI chief of staff

Build an AI chief of staff in a week: approval, a context file, rules, then calendar, notes and email. Only 12.5% of senior leaders connect AI to their tools.

How to Build an AI Chief of Staff: A Setup Guide for CoS
Short answer

An AI chief of staff runs on what you write: a one-page context file, written rules and saved jobs like a daily brief. An approved assistant such as Claude or ChatGPT reads your calendar, meeting notes and email and does your catching up. Once IT approves the assistant, a first version takes about a week, without code.

Key takeaways
  • Build it in this order: written approval, a one-page context file, rules for what it may do alone, then one connection and one recurring job at a time.
  • Send the request to IT on day one and write the context file while you wait, because approval often takes weeks.
  • Start with a daily brief and a meeting prep pack, run by hand for a week. Schedule a job only when three briefs in a row needed no more than a one-line fix.
  • Let it read, summarize and draft alone. Sends, invites and task assignments wait for your approval until each has a track record, and sending anything in your principal's name, or anything about a person, always stays with you.
  • Time your morning catch-up and 1:1 prep before you start, and judge the setup against those numbers after a month, because the gains come after weeks of tuning.

In our AI Readiness Diagnostic of senior leaders, 86% use AI every day, but only 12.5% connect it to their other tools.

Setting it up to do your catching up before you arrive is mostly writing: a request to IT, a one-page context file, a rules file and two prompts you correct for a week. The connections come after, one at a time.

How do you build an AI chief of staff in a week?

You build an AI chief of staff by giving an approved assistant your context, access to the places your work lives and standing instructions for the jobs you repeat. A useful first version takes a week, in five steps:

  • Get written approval for the assistant, the plan and the data that may go into it.
  • Write a one-page context file: your principal's priorities, who owns what, decisions already made, and how your principal likes information.
  • Write rules for what it does alone, what waits for your approval and what it never touches.
  • Connect your calendar first, then meeting notes, email and shared documents, a week apart.
  • Once the calendar is connected, build a daily brief and a meeting prep pack, run them by hand for a week (the week of tuning below shows what to correct), then schedule them.

The judgment calls stay with you. The calendar is the safest place to start, because a mistake in a brief is cheap to fix.

The build order. Send the request to IT on day one; approval often takes weeks. Write the context file (about 30 minutes) and the rules while you wait. After written approval, connect one a week: calendar, meeting notes, email, shared docs.

If you're the executive and want this for yourself, build it in the same order. Write the context file about your own priorities and team. The rules still apply to you, the "never" list included, because a message to your board in your own name carries the same risk. Our Claude guide shows how to swap the brackets in its templates for your own name and title, and the same swap works for the templates on this page. If you'd rather your Chief of Staff run it for you, the last FAQ covers that.

Alex Craig, Chief of Staff at PortSwigger, built his own and talked it through on our podcast. His morning catch-up and his 1:1 prep each shrank to a few minutes of reading, and board prep now starts with a draft in minutes.

What one Chief of Staff got back after building his setup

Alex Craig, Chief of Staff at PortSwigger, on two daily jobs before and after his build.

Rebuilding the picture each morning

about 30 min before
about 5 min after

Prep for a 1:1

about 15 min before
about 2 min after
Source: Alex Craig, "I'm a Chief of Staff. So I Built Myself One.", Chief of Staff Network blog

He also describes a few hours a week of tinkering for the first two months. When you're out, the briefs stop unless someone else can run them. Your principal will notice you know more, sooner, and start expecting it.

Here is the first week as a checklist, with a lane for the week IT hasn't answered yet.

Template: Week-one AI setup plan

Week one: my AI setup

Day 1

  • ☐ Send the request to IT (approved assistant, plan, connections, data line)
  • ☐ Write my baseline: morning catch-up, 1:1 prep, commitments chased, decision wait times

Day 2

  • ☐ Write the context file (about 30 minutes, one page)
  • ☐ Ask [principal]: "Is this the right order for the quarter?"
  • ☐ If [principal] has an EA: agree who reads the calendar and inbox

Day 3

  • ☐ Write my rules from the template and paste them in; set sends, shares and deletes to ask me first
  • ☐ Before I connect: list the shared folders and delegate inboxes the assistant must not read
  • ☐ Once approved: connect calendar (read only)
  • ☐ No approval yet: run tomorrow's brief by pasting the calendar in by hand

Day 4

  • ☐ Run the daily brief; note every correction
  • ☐ Run the meeting prep pack for my hardest meeting tomorrow

Day 5

  • ☐ Tell my regular meetings how notes will be taken, if I add a notetaker
  • ☐ Fold this week's corrections into the instructions
  • ☐ Decide the next connection for week two

Every Friday after

  • ☐ Update the context file (20 minutes)
  • ☐ Add at most one new recurring job
# Week one: my AI setup

## Day 1
- [ ] Send the request to IT (approved assistant, plan, connections, data line)
- [ ] Write my baseline: morning catch-up, 1:1 prep, commitments chased, decision wait times

## Day 2
- [ ] Write the context file (about 30 minutes, one page)
- [ ] Ask [principal]: "Is this the right order for the quarter?"
- [ ] If [principal] has an EA: agree who reads the calendar and inbox

## Day 3
- [ ] Write my rules from the template and paste them in; set sends, shares and deletes to ask me first
- [ ] Before I connect: list the shared folders and delegate inboxes the assistant must not read
- [ ] Once approved: connect calendar (read only)
- [ ] No approval yet: run tomorrow's brief by pasting the calendar in by hand

## Day 4
- [ ] Run the daily brief; note every correction
- [ ] Run the meeting prep pack for my hardest meeting tomorrow

## Day 5
- [ ] Tell my regular meetings how notes will be taken, if I add a notetaker
- [ ] Fold this week's corrections into the instructions
- [ ] Decide the next connection for week two

## Every Friday after
- [ ] Update the context file (20 minutes)
- [ ] Add at most one new recurring job

What do you need before you connect anything?

You need three answers in writing before you connect a single account: which assistant your company has approved, on which plan, and what data may go into it. If your company has no answer yet, getting one is your first deliverable, and it's a visible one.

Ask IT or security which AI tools are approved for company data, and whether that approval covers connecting email, calendar and documents or only chatting. On a company plan, an admin turns the connections on and decides whether anything may send without asking. On Microsoft 365, an administrator usually has to consent once before people can connect or use some actions. In the UK and EU, ask whether your data protection lead needs to review it as well, since email and meeting notes hold other people's personal data. A personal subscription on your own card sits outside your company's controls, and connecting your work email to one is the risk the review exists to prevent.

Write down what stays out: customer data, people and compensation data, legal matters, and anything under a board confidentiality obligation. In our Chief of AI Report, security, compliance and ethics are the barrier to AI adoption named most often, so expect the review to focus there. For the company-wide policy, use the template in our AI guide. For the legal side, see AI and the Law.

Approval often takes weeks. Until it comes, paste tomorrow's calendar and your own notes into the approved assistant, with nothing sensitive in them, so you learn what a good brief looks like. If the answer is no, ask what would change it, and put the review date in your weekly update so the request doesn't quietly die.

Template: AI setup request to IT

Subject: Request: connecting [approved assistant] to my work accounts

Hi [name],

I'd like to set up [approved assistant, plan] to help with briefings, meeting follow-ups and status tracking for [principal].

What I'd connect, in this order:
1. Calendar (read only)
2. Meeting notes from [notetaker], for meetings where everyone has been told
3. Email (read and draft; nothing sends without my approval)
4. Shared documents in [folder or drive]

What I won't put in: customer data, people and compensation data, legal matters, board-confidential material.

Questions for you:
- Is this assistant and plan approved for company data?
- Does approval cover these connections, or only chat?
- Are there settings you want on (no training on our data, admin approval for send actions, retention limits)?
- Anything else I should keep out?

Happy to walk you through it. I'd like to start on [date].

Thanks,
[name]

Send the request this week. While you wait, write the context file, which needs no access at all.

How do you write the context file?

A context file is a one-page document your assistant reads before every serious task: your principal's priorities in order, what's off the list, who owns what, the decisions already made and why, the words the company uses, and how your principal likes information.

Write the first version in about 30 minutes, then ask your principal: "Is this the right order for the quarter?" The excerpt below shows the shape. The full template, with the decision log and commitments list that sit beside it, is in our context management guide.

Template: Context file (excerpt)

Context file: [Principal name, title]

Last updated: [date] · Owner: [Chief of Staff]

This quarter's priorities (in order)

  1. [Priority]: why it matters, what "done" looks like, owner
  2. [Priority]
  3. [Priority]

Off the list this quarter

  • [Item]: why it was deprioritized, who agreed

Who owns what

AreaOwnerNotes
[Area][Name, title][Current focus, how they like to work]

Decisions that shape everything else

DateDecisionWhyRevisit when
[Date][Decision][Reasoning][Trigger]

How [principal] wants information

  • Format, tone, and what never to do

Rules for AI working with this file

  • Flag anything that conflicts with the priorities above.
  • Never state a number, name or date that isn't in the source material. Mark gaps as [CHECK].
  • Anything about a person is a draft for me to verify, never a fact.
# Context file: [Principal name, title]
Last updated: [date] · Owner: [Chief of Staff]

## This quarter's priorities (in order)
1. [Priority]: why it matters, what "done" looks like, owner
2. [Priority]
3. [Priority]

## Off the list this quarter
- [Item]: why it was deprioritized, who agreed

## Who owns what
| Area | Owner | Notes |
| --- | --- | --- |
| [Area] | [Name, title] | [Current focus, how they like to work] |

## Decisions that shape everything else
| Date | Decision | Why | Revisit when |
| --- | --- | --- | --- |
| [Date] | [Decision] | [Reasoning] | [Trigger] |

## How [principal] wants information
- Format, tone, and what never to do

## Rules for AI working with this file
- Flag anything that conflicts with the priorities above.
- Never state a number, name or date that isn't in the source material. Mark gaps as [CHECK].
- Anything about a person is a draft for me to verify, never a fact.

Example: Context file excerpt for a Series B CEO (illustration)

Context file: CEO, Series B fintech, 120 people

Last updated: Nov 3 · Owner: Chief of Staff

This quarter's priorities (in order)

  1. Hit the Q4 revenue plan. Done means plan met in the January board pre-read. Owner: CRO.
  2. Move the top 30 customers to the new platform before the old one switches off. Owner: VP Customer Success.
  3. Hire a CFO. Done means a signed offer. Owner: CEO.

Off the list this quarter

  • Small-business product line: deferred to Q2, agreed by the CEO and the CRO

Who owns what

AreaOwnerNotes
Revenue planCROForecast call every Monday
Platform migrationVP Customer SuccessWants issues raised in writing

Decisions that shape everything else

DateDecisionWhyRevisit when
Oct 2Old platform switches off March 31Running two systems ties up the support teamA top-10 customer can't move in time

How the CEO wants information

  • Decision first, one page, numbers in a table
  • Never an "FYI" without an ask

Rules for AI working with this file

  • Never state a number, name or date that isn't in the source material. Mark gaps as [CHECK].
# Context file: CEO, Series B fintech, 120 people
Last updated: Nov 3 · Owner: Chief of Staff

## This quarter's priorities (in order)
1. Hit the Q4 revenue plan. Done means plan met in the January board pre-read. Owner: CRO.
2. Move the top 30 customers to the new platform before the old one switches off. Owner: VP Customer Success.
3. Hire a CFO. Done means a signed offer. Owner: CEO.

## Off the list this quarter
- Small-business product line: deferred to Q2, agreed by the CEO and the CRO

## Who owns what
| Area | Owner | Notes |
| --- | --- | --- |
| Revenue plan | CRO | Forecast call every Monday |
| Platform migration | VP Customer Success | Wants issues raised in writing |

## Decisions that shape everything else
| Date | Decision | Why | Revisit when |
| --- | --- | --- | --- |
| Oct 2 | Old platform switches off March 31 | Running two systems ties up the support team | A top-10 customer can't move in time |

## How the CEO wants information
- Decision first, one page, numbers in a table
- Never an "FYI" without an ask

## Rules for AI working with this file
- Never state a number, name or date that isn't in the source material. Mark gaps as [CHECK].

Alex runs his whole system from one folder of plain text files that holds his knowledge base, decisions and corrections. He puts every brief, draft and agenda through that folder too, so the system learns from the work he actually does.

These systems usually fail at upkeep. Alex's assistant kept preparing briefing notes on someone who had left the company weeks earlier, because nobody told it.

Update the context file the same day a priority or an owner changes, and review it every Friday.

How do you set rules for what it may do alone?

Write the assistant's rules the way you'd write decision rights for a new hire: what it does alone, what it prepares for you to approve, and what it never touches. Some of this is a setting. The rest is in your instructions and your habits.

Reading, summarizing, finding and drafting can run alone. Sending, scheduling, sharing or assigning work to someone starts with your approval every time. Nothing goes to the board, investors or press in your principal's name without you. Anything about a person's performance, pay or employment stays out entirely, and so do legal matters.

A wrong output looks as finished as a right one. Alex's assistant assigned an action to his boss, with a due date, without flagging it to him first. It credited one colleague's update to another in a meeting summary, and presented a year-old product release as this week's news.

He caught the first in his review queue. He now writes down every failure the same day and checks anything about people against the source before it leaves his laptop. Check anything about people the same way, because people remember when the Chief of Staff got it wrong.

Template: Rules for my AI assistant

Rules for my assistant ([your name], Chief of Staff to [principal])

You may do these on your own

  • Read my calendar, meeting notes, email and the shared folders I've connected
  • Summarize, find, compare and draft
  • Suggest tasks and follow-ups, labeled with their source

Ask me first, every time

  • Sending, replying, forwarding or scheduling anything
  • Sharing, moving or deleting any file
  • Assigning a task or a date to anyone, including me

Never

  • Send anything to the board, investors or press in [principal]'s name
  • Draft anything about a person's performance, pay or employment
  • Use customer, people or legal data, even if you can see it
  • Fill a gap with a guess: say what's missing

Always

  • Mark anything about a named person "check before using"
  • Give the source (meeting, thread, document) for every commitment
  • Use the context file; tell me when it looks out of date
# Rules for my assistant ([your name], Chief of Staff to [principal])

## You may do these on your own
- Read my calendar, meeting notes, email and the shared folders I've connected
- Summarize, find, compare and draft
- Suggest tasks and follow-ups, labeled with their source

## Ask me first, every time
- Sending, replying, forwarding or scheduling anything
- Sharing, moving or deleting any file
- Assigning a task or a date to anyone, including me

## Never
- Send anything to the board, investors or press in [principal]'s name
- Draft anything about a person's performance, pay or employment
- Use customer, people or legal data, even if you can see it
- Fill a gap with a guess: say what's missing

## Always
- Mark anything about a named person "check before using"
- Give the source (meeting, thread, document) for every commitment
- Use the context file; tell me when it looks out of date

How autonomy grows, and where it stops

Two questions set every rule: can the action be undone, and who sees it? A draft or a tag on a task can run alone from day one. Something that reaches another person can earn independence slowly. Anything external, anything about a person, and anything in your principal's name never does, however good the track record.

Take internal scheduling replies, the lowest-stakes message most Chiefs of Staff send. Keep them as drafts for a month and count how many you change. My rule of thumb: if fewer than one in twenty needed an edit, let it send confirmations to colleagues inside the company, and write the new permission into the rules file with the date.

Drafts and summaries run alone from day one. Scheduling replies to colleagues wait for approval, then send alone after a month if fewer than one in twenty needed an edit. Anything external, about a person or in your principal’s name never runs alone.

The judgment call people miss is access. A connector sees what you can see, including the HR folder someone shared with you once and your principal's inbox if you have delegate access. Name those places in the "never" list, and decide with your principal which parts of their inbox the assistant may touch.

What about personal agents?

Personal agents run on their own computer in the cloud, sign in to your accounts and act for you over hours or days. Their terms and data handling vary widely, so run these checks first:

Personal agents: what to check before using one for work

QuestionWhy it matters for a Chief of Staff
Does it log in as you, with stored credentials?It can act anywhere you can, including in your principal's shared inbox or calendar
Can you separate read and write access, by service?You want reading on and sending off until you trust it
Does it ask before sending, booking or buying?A send it makes without asking goes out under your name
Is your data used for training, and can you switch that off?Company information may leave the company
What do the terms say about licenses, retention and deletion?Some terms grant broad rights over what you connect
Is it on your company's approved list?If not, keep it to personal errands

One agent forwarding the wrong thread from your principal's inbox can cost you the role. I wouldn't connect a work account to any of these until IT has reviewed it, however good the demo looks. If you already use one, check which accounts it's signed into and remove any work account your company hasn't approved.

How do you build a chief of staff agent in Claude or ChatGPT?

In Claude, a chief of staff agent is a project, connectors, a skill and a scheduled task. In ChatGPT, it's a project, apps and a scheduled task. Apps are ChatGPT's name for connectors, and a plugin bundles apps with reusable instructions. In either, keep the order: context and rules before any connection, calendar before email, a week by hand before you automate.

Schedule only jobs that read and draft. When a step needs your approval, the run may wait for you or finish without that step. If you'd rather work from files on your own computer, as Alex does, the desktop app lets Claude read and write a local folder. Scheduled tasks that need that folder run only on your computer. ChatGPT's counterpart is Work, an agent that works across your files, plugins and connected apps, and asks you to approve important actions.

Build your AI chief of staff

Claude

  1. On day one, write a chief of staff skill for your written work, such as the weekly update and meeting briefs. It holds only your procedure, so you can write it before IT approves a plan. Our Claude guide has one to copy.
  2. Create a project for your principal. Add the context file and your rules as project instructions and files, so every chat inside it starts from the same page.
  3. Ask your admin to enable the Google Workspace or Microsoft 365 connectors. Before you connect anything, set every action that sends, shares or deletes to ask for your approval first, and switch Claude to Manual. Then turn on the calendar, and add notes and email a week apart.
  4. Run the daily brief prompt by hand for a week inside the project.
  5. Once the daily brief runs cleanly, add it to your chief of staff skill.
  6. Schedule it from Cowork. Keep the job to reading and drafting. Read every brief for the first two weeks before you act on it.

ChatGPT

  1. Create a project for your principal. Add the context file and your rules as project instructions and files.
  2. Ask your admin which apps for your calendar, email and documents you may use, and with which actions. Before you connect anything, set every action that sends, shares or deletes to ask for your approval first. Then connect the calendar, and add notes and email a week apart.
  3. Run the daily brief prompt by hand for a week inside the project.
  4. Save the working version as a plugin, which combines reusable instructions with connected apps, or keep it in the project's instructions.
  5. Set it up as a scheduled task that only reads and drafts, because actions that need approval pause it until you review them. Read every brief for the first two weeks before you act on it.

In Microsoft 365 Copilot, an agent built in Agent Builder holds the instructions and knowledge, and scheduled tasks run a saved prompt for you. Both depend on the Microsoft licenses your company holds. For a ready-made chief of staff skill and help choosing a Claude plan, see our guide to Claude for Chiefs of Staff.

How do you connect your calendar, meeting notes and email?

Connect in order of risk: calendar first, then meeting notes, then email, then shared documents. Each connection gives the assistant more to work with and more ways to embarrass you, so add one, use it for a week, and then add the next. After each one, ask a question only that source can answer, and check the answer yourself. For the calendar, run a one-paragraph brief on tomorrow's meetings, check it against what you know, and fix the context file before you connect anything else.

A connector, or an MCP (Model Context Protocol) connection, lets the assistant read from, and sometimes act in, another system using your login and your permissions. Always allow reading, because a read can't change anything. Set every action that sends, shares or deletes to ask for your approval first. On most plans, Claude starts on Auto, where it decides on those actions itself, so check this before you connect anything. Switch it to Manual, so Claude asks before it acts. ChatGPT calls its connectors apps, and you set them the same way where the app allows it: reading allowed, and your approval first for anything that sends, shares or deletes.

What each connection gives you, and what to watch

ConnectionWhat it gives the assistantMain riskWho approves
Calendar (read)Who you're meeting, when, and the agendaLow; private event titlesYou, within company policy
Meeting notesWhat was said and agreed in each meetingRecording without consent; sensitive meetings capturedIT, legal for recording rules, and the people in the room
Email (read and draft)Threads, asks and commitmentsSends in your name; people and customer dataIT
Shared documentsPlans, decks, memos and their historyOver-sharing restricted filesDocument owners and IT
Chat (Slack, Teams)Discussion and decisions between meetingsPrivate channels; casual remarks treated as decisionsIT and channel owners

If your principal has an executive assistant, talk to them before any assistant touches the principal's calendar or shared inbox. They own that ground, and a brief that contradicts their scheduling costs you more than it saves.

Meeting notes need the most care, because recording changes the room. Aria Finger, then Chief of Staff to Reid Hoffman, told our podcast she records all of her meetings with a notetaker and asks an assistant about anything she didn't catch. Some notetakers run quietly on your laptop, which is why you have to tell people. Our meetings guide has the consent rules and the list of meetings never to record.

Email comes third because it's where mistakes leave the building. Let the assistant pull asks and commitments out of threads and draft replies for you to edit, and keep it on read and draft for at least the first month.

Which recurring jobs should you build first?

Build the recurring jobs you already do by hand, one at a time, starting with a daily brief and a meeting prep pack. Each one should save time you can name, every week, before you add the next.

A recurring job is a saved set of instructions the assistant runs on demand or on a schedule: a skill in Claude, or a plugin or saved prompt in a ChatGPT project.

Alex added his jobs one at a time over his first 30 days:

  • Days 1 to 5: one folder, a connected calendar and meeting recorder, and one real piece of work done through the assistant.
  • Days 6 to 14: a 6:30 weekday brief under 200 words, with one line and one thing to prepare per meeting, plus anything he promised in the past 72 hours. After a week, one new rule based on what it kept flagging.
  • Days 15 to 21: a Friday afternoon scan of the week's work for steps he did by hand more than once, which suggests one new standing instruction or scheduled task.
  • Days 22 to 30: a 1pm check comparing the morning's plan with what got done, then a review that cuts any job that felt like noise.

He calls the Friday scan "the task that builds the system." Run it every Friday for a month, he says, and you'll have four new rules or scheduled tasks, each one something you were doing by hand.

He moved faster than this guide suggests. He connected everything his company allowed and scheduled the brief on day 6. If your principal or their EA will see the output, keep the slower order.

Start with the daily brief. It runs five times a week, so you get five rounds of corrections before Friday, you can grade it in a minute, and when it's wrong only you see it. Run these two prompts by hand for a week before you schedule anything.

Prompt: Daily brief for a Chief of Staff

You are helping me, a Chief of Staff to [principal, title], start the day. Use only my calendar, meeting notes from the last 3 working days, my email from the last 24 hours and my context file. Do not guess.

Write a brief of no more than 250 words:
1. Today's meetings, in order. For each: why it matters to this quarter's priorities, what was decided last time, and one question I should ask.
2. Decisions waiting on [principal], with how long each has waited.
3. Commitments made to or by me in the last 72 hours that are due or at risk, with the source.
4. Anything urgent in email that needs me before noon.

Flag anything about a named person as "check before using". If a source is missing, say which one.

The meeting prep pack is the second job, for the meetings where a decision is at stake. The pack is for you. If your principal needs something before the meeting, send a meeting brief.

Prompt: Meeting prep pack

Prepare me for this meeting: [title, date, attendees].

Use my context file, the last two sets of notes from meetings with these attendees, and any documents linked in the invite.

Give me:
- The purpose of the meeting in one sentence, and the decision or output we need by the end.
- What each attendee is likely to care about, based only on what they've said in our notes.
- Open commitments between us and them, with dates.
- Three questions I should ask, and the one risk I should raise.
- A one-line summary I can send [principal] afterwards, with blanks for the outcome.

Keep it under 200 words. Mark anything you inferred rather than read.

A week of tuning, step by step

In this illustration of a first week, each day's error becomes a line in the instructions.

  1. Monday. The brief gives every one of eleven meetings a full paragraph and runs past 400 words. You skim it and miss the one that mattered. New line: "Only meetings where a decision or commitment is at stake get a paragraph. Everything else gets one line."
  2. Tuesday. It lists the pricing decision as still open. Your principal settled it by text at 9pm, which no connector saw. You add the decision to the context file, plus a line: "If the context file records a decision more recently than the notes, the file wins."
  3. Wednesday. It credits an action to the wrong VP, the same mix-up Alex found in his own summaries. New line: "Name an owner only when the notes name one. Otherwise write 'owner unclear'."
  4. Thursday. It buries a late-night email from the board chair under routine threads. New line: "List messages from [names] first, whatever time they arrived."
  5. Friday. You fold the four lines into the saved instructions and run Monday's inputs again to check the fixes hold.

Those four rules came from your principal's habits and your company's gaps, which is why a borrowed prompt rarely works as well as one you've corrected for a week.

After the first week, run Alex's Friday scan each Friday afternoon and add only the one thing it suggests.

Prompt: Friday scan for the next job

Review the work in [folder or project] from the past 7 days: tasks, briefs, drafts and meeting notes.

For each piece of work, note:
- Where I worked around the system (used another tool, or did by hand something you could have done)
- Any step I repeated more than once
- Any decision I made that isn't written down as a standing instruction

Then suggest ONE new standing instruction or scheduled task. Make it specific enough to run as written: what it reads, when it runs and what it produces. Don't suggest more than one. If nothing qualifies, say so.

The third job: your weekly 1:1 with your principal

Once the brief and the prep pack hold up, make your weekly 1:1 with your principal the next job. Before you write your weekly update, have the assistant mark each action from last week's follow-up as closed, in progress or at risk, so you know what's drifting before the 1:1.

Where people go wrong, and how to tell

Most people schedule too early. My test: if you changed more than a line in any of the last three briefs, keep running it by hand, because a brief you rewrite every morning has taught you to skim. Don't schedule work you do less than once a week. Keep it as a saved prompt.

The last three briefs, marked up. When each needed no more than a one-line fix, schedule the job. When more than a line changed in any one of them, keep running it by hand.

Every correction adds a line, and every line adds output. If you've stopped reading past the first section two days running, cut the brief back to the items you acted on that week. For a CTO, the brief wants incidents, roadmap slips and hiring. For a head of sales, it wants deals, forecast changes and escalations. In a remote company, ask for the chat connection earlier, or the brief will miss most of what was decided.

Alex is honest about a cost: once the routine work disappeared, his days filled with back-to-back demanding work and less recovery time, and he doesn't have a fix yet. Protect a block each day that the system doesn't fill. Your principal may also start forwarding your briefs, so an error in one now travels.

How do you know it's working, and what does it give back?

It's working when you can point to hours returned, decisions made sooner and fewer dropped commitments, measured against a baseline you wrote down before you started. Measure for a month, then decide what to keep. The first two measures below are enough to start.

A one-month scorecard for your setup

MeasureHow to take the baselineWhat counts as a result
Morning catch-upTime it on three normal daysMinutes saved, minus the minutes you spend checking the brief
1:1 prepTime your prep for the next three 1:1sPrep time down, and no question you wished you'd asked
Commitments chasedCount last week's commitments you had to chase or found lateFewer found late
Decision waitList open decisions for your principal and how many days each has waitedFewer days waiting on each decision

Alex's results came after weeks of tuning, so don't judge your own setup on week one. The hours count when they move into work only you can do, such as preparing the decisions that matter; if the time saved went to clearing more email, the setup made you faster without making you more useful. Write down each Friday where the saved hours went, and count the checking honestly. Those before-and-after numbers are the evidence to bring to your principal.

If it works, expect your principal to ask you to roll it out to the leadership team, often around month two. Don't hand over a copy: your context file holds notes about people and your principal's priorities, and each leader needs their own file, connections and rules. Start with one leader whose week looks like yours, and treat it as the first step of a company program, which our AI guide covers.

Alex puts his setup at about $50 a day at full use, on a usage-billed enterprise plan, and says a top-tier individual subscription could run all of it. Across 250 working days, $50 a day comes to about $12,500 a year. He weighs that against the work it does, which he says would otherwise need another person.

Write down how long your morning catch-up and 1:1 prep take today, so you can show what the setup returns in a month.

Put a 30-minute review in your calendar four weeks from today.

At the end of the month, decide job by job: keep a job that saved time you can name, rewrite one that needed heavy edits, drop one you stopped reading.

Then choose the next job from your own week: the step you did by hand most often that month, which a Friday scan like Alex's will find for you.

Swap context files, rules and prompts with other Chiefs of Staff building the same setup. Membership includes peer groups, events and the Power Skills program.

See membership

Take this guide with you

The full guide as a PDF: every section, template and chart, ready to print or share with your team.

Your PDF is ready.

Download the PDF
Oops! Something went wrong while submitting the form.

Free. We may email you new guides; unsubscribe anytime.

Frequently asked questions

Do I need to code to build an AI chief of staff?

No. Connectors, projects, skills and scheduled tasks in Claude and ChatGPT need no code. A useful setup runs on a one-page context file, two or three approved connections, a few well-tested instructions and written rules for what the assistant may do alone.

Can AI be your chief of staff?

Partly. An assistant can take over the catching up: briefs, meeting prep, follow-up drafts and status tracking. It can't hold the relationships, read the room or make the judgment calls at the center of what a Chief of Staff does, so it works best as the Chief of Staff's own assistant, or as a founder's until they hire one. Our AI guide explains why it won't replace the role.

Is it safe to connect my company email to an AI assistant?

Yes, on a plan your company has approved, with sending set to need your approval and a written list of data you keep out. Check whether the vendor trains on connected data and how long it keeps it. On a personal account outside company controls, the risk sits with you, whatever the tool.

How long does it take to set up?

A first version takes about a week alongside your normal work once IT has approved it: about 30 minutes for the context file and a few days of running the daily brief by hand. Approval itself often takes weeks. A mature setup like Alex Craig's took a few hours a week for about two months, adding one recurring job at a time.

What is the best AI chief of staff tool?

The one your company has approved for company data. For most Chiefs of Staff that's Claude or ChatGPT on a business plan, and both work the same way here: a project, connections, rules and scheduled jobs. Buying a dedicated app makes sense only in the cases our AI tools guide lists.

How do I set up an AI assistant for my CEO?

Use the steps above, with the CEO's priorities in the context file and their EA's ground written into the rules. You run it. If the CEO wants their own, set it up as a separate system with their own context file, connections they approve and rules they agree to. Never hand over a copy of yours, because it holds your notes on people.

Sources and methodology

Learn the job with people who do it

Membership gets you the Power Skills program, member events and a private community of working Chiefs of Staff.

Land your first Chief of Staff role

Launchpad helps you prepare, position yourself and interview for the role.

Starting out in the role?

The Fundamentals email course walks you through scoping the mandate, setting the cadence and earning trust.

Lead AI adoption from the Chief of Staff seat

The Chief of AI Fellowship is for operators turning AI into how the company runs.

Hiring a Chief of Staff?

Post the role on the Chief of Staff job board, or browse open roles.

Get your resume ready for the role

Get feedback on your Chief of Staff resume before you apply.

Let the right roles find you

Join the talent network and hear about Chief of Staff roles first.

The Chief of Staff newsletter

What Chiefs of Staff are reading, hiring for and building.

Get the full report

Every chart and cut from our latest salary research, sent to your inbox.

Using AI with sensitive data?

Learn the legal side of using AI at work.

More series