AI for executive assistants works best on prep work: turning a messy inbox into a morning brief, drafting replies in your executive’s voice, comparing calendar and travel options, building meeting packs, and turning rough notes into action lists. You keep the judgment, the relationships and the final check. Start with a tool your company has approved, never paste confidential data into a personal account, and master one workflow at a time.
Below you’ll find 10 workflows with prompts you can copy, a privacy checklist, the skills that matter, and a 30-day training path.
What to hand to AI, what to keep
Most of the job isn’t typing. It’s knowing that “let’s catch up sometime” from one person means never, while from another it means this week. AI can’t know that. It can do the groundwork so you have more time for the part that needs you.
| Hand to AI (first draft) | Keep for yourself (final call) |
|---|---|
| Sorting and summarising the inbox | Deciding what actually reaches your executive |
| Drafting routine replies | Anything sensitive, emotional or political |
| Listing time-zone and travel options | Choosing between them, knowing the person’s habits |
| Structuring briefing packs | Judging what your executive needs to hear before the meeting |
| Cleaning up minutes, extracting actions | Confirming decisions and owners were captured correctly |
| Formatting decks, tightening copy | Message, narrative, what stays out |
| Collecting public background on people and companies | Verifying facts, deciding what’s relevant |
| Summarising expenses and invoices | Approvals, exceptions, anything touching money movement |
| First-pass event checklists and run-of-show | Vendor relationships, guest sensitivities, on-the-day calls |
If you’re wondering whether this is about your job security, that has its own article. Short version: the work is shifting, not vanishing.
One more thing before the workflows. Good prompts decide most of the result. If yours come back vague, our guide on how to write better AI prompts will help. Every example below uses made-up names and data. Swap in your own, inside your company’s approved tool.
10 workflows for executive assistants
Here’s AI for executive assistants in practice: ten jobs from a normal week, each with a prompt to copy and a check you shouldn’t skip.
1. Morning brief from an inbox
You open your executive’s inbox at 7:30 and there are 140 unread messages. A brief turns that into five minutes of reading.
You are an executive assistant’s helper. Below are the emails received overnight in my executive’s inbox (names and details are fictional). Create a morning brief with: (1) Needs a decision today, (2) Needs a reply from me this week, (3) FYI only, (4) Can be ignored or archived. For each item give one line: who, what, deadline. Flag anything that looks urgent or emotional. Do not invent details that aren’t in the emails.
[paste emails]Check you must do: Skim the “ignore” pile yourself. AI is good at spotting urgency words and bad at spotting a quiet message from a board member that matters more than ten loud ones. If you work inside Outlook or Gmail, see our guides on AI email assistants, Copilot in Outlook and Gemini in Gmail for what each tool can do inside the inbox.
2. Replies in your executive’s voice
This is where the “voice and preferences” file (more on that below) pays off.
Draft a reply to the email below as if written by Jordan Hale, CEO. Use the voice notes: short sentences, warm but direct, no exclamation marks, signs off with “Best, Jordan”. The decision is: decline politely, offer a call in March. Keep it under 90 words. Give me two versions: one friendlier, one more formal.
[voice notes] [email]Check you must do: Read it as the recipient would. Does it promise something your executive hasn’t agreed to? Drafts often sound confident about commitments. Also, never send a first draft on a sensitive thread. Use it as a starting point, then rewrite the one sentence that matters.
3. Scheduling across time zones
Some facts that trip people up: India runs on UTC+5:30, Nepal on +5:45, and the Chatham Islands on +12:45. Arizona doesn’t change its clocks. And the US and Europe move clocks on different weekends, so for a couple of weeks a year your usual London–New York gap is off by an hour. A language model can list all this quickly, but it can also get arithmetic wrong.
My executive is in Chicago. Participants are in London, Mumbai and Singapore. Propose three 45-minute slots next week that fall between 7:00 and 19:00 local time for everyone, if possible. Show each slot in all four time zones. If no slot works for all, tell me who has the worst time and by how much. Note any daylight-saving changes that could affect the dates.Check you must do: Verify every slot in your calendar app or a time-zone converter. Never trust the model’s math on offsets. And check your executive’s actual calendar and protected time; AI doesn’t know about the Friday school pickup. For more on this topic, see our piece on the AI calendar assistant.
4. Travel itineraries and options
Create a clean itinerary from these fictional bookings: [flight details, hotel, meetings]. Format: day by day, local times, addresses, who is meeting whom, and travel time between stops with a 20-minute buffer. Then list three risks (tight connections, meetings too close to landing, no meal window) and suggest an alternative for each.Check you must do: Confirm every flight number, date and address against the original confirmations. Travel documents are where a small slip becomes an expensive one. Also check preferences: seat, hotel chain, dietary needs. Those live in your voice and preferences file, not in the model.
5. Meeting briefing packs
Build a one-page briefing for my executive’s meeting with [fictional company]. Include: purpose of the meeting, attendees and roles, last three interactions (from the notes below), what they probably want, what we want, three questions my executive should ask, and one risk. Keep it scannable, under 350 words.
[notes, previous emails]Check you must do: Make sure “what they probably want” is based on the notes, not model guesswork. Mark anything you can’t source. A good pack separates facts from assumptions.
6. Minutes and action items
Here is a transcript of a leadership meeting (fictional). Produce: (1) a five-line summary, (2) decisions made, (3) action items in a table with owner, task and deadline, (4) open questions. If an owner or deadline wasn’t stated, write “not stated”. Do not guess.Check you must do: Compare against your own notes. Models smooth over disagreements and sometimes assign a task to the wrong person who simply spoke last. “Not stated” is your friend; chase those gaps in person.
7. Board or leadership deck prep
I’m preparing a 10-slide leadership update from the rough notes below (fictional). Propose a slide-by-slide outline with one headline per slide that states the takeaway, not just the topic. Suggest where a chart would help. Point out where the story is weak or missing data.
[notes]Check you must do: Every number, date and name must be checked against the source. Decks go to the board, and a wrong figure on slide 6 is the one people remember. Also decide what shouldn’t be in the deck. That’s your judgment, not the model’s.
8. Research on a person or company before a meeting
Using only publicly available information, summarise [fictional company] in 200 words: what it does, recent public announcements, leadership names and roles, and anything my executive should avoid mentioning. For each claim, cite the source. If you can’t find a source, say so.Check you must do: Open the sources. AI search tools can cite a page that doesn’t say what the summary claims. For people, stay with professional, public information. If something feels like prying, it probably is.
9. Expense and invoice summaries
Below is a fictional list of expenses for Q3. Group by category, total each, and flag anything unusual: duplicates, round-number amounts, items without a description, or amounts far above the category average. Output a table, then a three-line summary for my executive.
[expense data]Check you must do: Recalculate totals in a spreadsheet. Treat AI flags as hints, not findings. And don’t paste real financial data unless your company’s approved tool and policy allow it.
10. Event planning
Plan a 2-hour leadership dinner for 12 people in Chicago (fictional). Give me: a planning timeline working backwards from the date, a vendor checklist, a draft run-of-show with times, seating principles, and five things that commonly go wrong with this kind of event and how to prevent them.Check you must do: Call the venue. Check dietary restrictions, accessibility, and who must not sit together. AI gives you structure fast; it doesn’t know that two guests are in a dispute.
Build your executive’s “voice and preferences” file
Most mediocre AI output comes from missing context. A single reusable file fixes that. Keep it somewhere secure, and keep it free of passwords, financial details and anything your company’s policy doesn’t allow in an AI tool.
Copy this template:
Name and role: [Jordan Hale, CEO]
Voice: [short sentences / warm but direct / no exclamation marks / uses first names]
Sign-offs: [Best, Jordan]
Words they like: […] Words they hate: […]
Example emails (2–3, anonymised): [paste]
Meeting preferences: [no meetings before 9:00, 25-minute default, video off for calls under 15 minutes]
Travel preferences: [aisle seat, no early flights, hotel within 10 minutes of venue]
Food and dietary notes: […]
People who always get through: [roles, not necessarily names]
Topics to escalate to a human immediately: [legal, press, family, board matters]
Never do without asking: [accept invitations, commit to money, share documents]
Use it at the start of a conversation: “Here’s my executive’s voice and preferences file. Use it for everything in this chat.” Update it every month or so. It takes about 20 minutes to build and gets better every time you correct a draft.
Confidentiality rules for EAs: a checklist
You see more sensitive information than almost anyone in the building: salary talks, deals, health, hiring, board drafts. That’s why the rules matter more for you than for most. For the bigger picture, read whether AI tools are safe to use at work.
- Use only approved tools. If IT hasn’t approved it, don’t use it for work. If there’s no policy, ask for one rather than guessing.
- Keep business and personal accounts apart. Business plans and consumer plans often treat data differently, including whether chats may be used to improve models. Read the provider’s official data-handling page and follow your company’s policy.
- Anonymise by default. Replace names, companies and figures with placeholders when you can: “Person A”, “Client X”.
- Never paste credentials or access links. Passwords, banking details, private calendar links, and document links with open access stay out.
- Treat anything unreleased as off-limits unless cleared. M&A, layoffs, earnings, legal matters, unannounced products.
- Know where outputs go. Check whether chat history is saved, who in your organisation can see it, and how to delete it.
- Tell your executive what you use. No surprises. If they’d be uncomfortable hearing about it, rethink it.
- When unsure, don’t paste. You can always do it by hand. You can’t take a leak back.
The AI skills that make an EA more valuable
The people who get the most out of AI aren’t the ones who know the most tools. They’re the ones who are clear about what they need. AI skills for executive assistants come down to a short list:
Writing a clear brief. Say who the output is for, what it’s for, the format, the length, and what to avoid. This is the skill that improves everything else.
Spotting a confident mistake. Models state wrong things in a calm voice. Your instinct for “that doesn’t sound right” is the most valuable thing you bring.
Knowing what’s safe to share. Judgment about privacy turns you from a risk into the person the company trusts to use these tools.
Building reusable systems. A voice file, a prompt library, a standard meeting-pack format. You’re setting up a workflow, not just doing a task.
Teaching others. Once you’ve got a few working prompts, other assistants will ask for them. That visibility matters.
If your path is wider than executive support, you might also look at our guides on AI for administrative assistants and AI courses for virtual assistants.
30-day training path
You don’t need a big block of free time. Twenty minutes a day is enough if you pick a real task from your actual week. Here’s a plan.
| Week | Focus | Practice task | Options |
|---|---|---|---|
| 1 | Foundations and safety | Read your company’s AI policy. Learn basic prompt structure. Run workflow 1 (morning brief) on fake data. | Free: official help pages from your approved tool (Microsoft Copilot, Gemini, ChatGPT, Claude); our prompt guide. Paid: a structured beginner course. |
| 2 | Writing and communication | Build your voice file. Run workflows 2 and 6. Compare drafts with what you’d really write. | Free: trial and error with your executive’s past emails (anonymised). Paid: a course with writing exercises. |
| 3 | Planning and logistics | Run workflows 3, 4 and 10. Practise double-checking every time-sensitive detail. | Free: calendar and travel features in your existing tools. Paid: a course on productivity workflows. |
| 4 | Prep, research and presentation | Run workflows 5, 7, 8 and 9. Pitch one workflow to your executive (see below). | Free: peer swaps with other EAs. Paid: a certificate programme for a clear finish line. |
Check any provider’s current price, certificate and plan limits on their own page before you commit. Features can vary by plan, so confirm what your company’s licence includes.
If you want a guided route, AI training for executive assistants doesn’t have to mean a long course. Coursiv’s AI courses are built around short daily lessons, which suit a demanding calendar. The 28-Day AI Certificate Program gives you a month of daily practice and a certificate of completion at the end. Coursiv doesn’t have a course made only for EAs, but the skills transfer directly to the workflows above.
How to pitch AI use to your executive
Don’t open with “I want to use AI.” Open with a problem they already feel. Here’s a script to adapt:
“You told me last week that mornings are a mess. I’d like to try something small for two weeks: I use our approved AI tool to prepare a one-page summary of your inbox before you start. I’ll check every line, nothing leaves our systems, and I won’t send anything on your behalf without your say-so. If it saves you time, we keep it. If not, we drop it. Okay?”
Why this works: it’s one task, it has an end date, it names the risk, and it keeps control with them. After two weeks, show them the result. Then propose the next workflow.
Ready to get ahead of this?
The EAs who train on AI now become the person leaders can’t work without. If you’re looking for an AI course for executive assistants that fits around a packed day, start with Coursiv’s 28-Day AI Certificate Program. Short daily lessons, real-world practice, and a certificate of completion. No promises about jobs or pay, just skills you can use on Monday morning.