25 ChatGPT Prompts for Executive Assistants (Email, Scheduling, Travel, Research)

The job of a good executive assistant is to make decisions for the executive without seeming to. Reply to the email, book the flight, kill the meeting, surface the one number that matters from a 30-page board document. The ChatGPT prompts for executive assistants below were written for that job, not the one HR thinks the job is.

25 prompts grouped by the five jobs an EA actually does every day: managing the inbox, running the calendar, handling travel and logistics, doing prep research, and producing the documents the executive walks into meetings holding. Every prompt has a defined role, a specific task, named constraints, and a structured output. Copy them, swap your specifics into the brackets, and ship the result.

Prompt 4 (the morning inbox digest) was run on free ChatGPT on 2026-05-17 with a real example of 18 morning subject lines baked in. The verbatim response is in the post with editor's notes on what the model nailed and where you still have to override its judgment.

How to use these ChatGPT prompts for executive assistants

Three things to get more out of every prompt below:

Never paste account-identifying customer data or anything personal about a named employee into free-tier ChatGPT. Free-tier chats are not a private workspace. Use a paid tier with data controls if you need to paste real names, real dollar amounts, real medical or legal context. Otherwise anonymize.
Give the model the rules your executive has, even if you think they're obvious. "My exec hates double-booking, won't take meetings before 9am, doesn't do Friday afternoons" turns a generic reply into a usable one. The model can't read your exec's preferences from thin air.
Treat every output as a draft. Even when the model nails the format, the EA judgment call (which person to flag, what to escalate, when to push back) is still yours. The prompt saves the 80% of the work that's structure; the 20% that's decisions stays with you.

Inbox and email

1. Triage long email thread

Prompt: "You are a senior executive assistant. I'll paste a 30+ email thread. Do three things: (a) summarize the latest status in 3 bullets, (b) identify any unanswered question from any participant, (c) draft a one-sentence next-step recommendation for my executive. Skip pleasantries. Treat anything in all caps as urgent. Flag if there's a deadline anywhere in the thread. Output as 3 sections labelled Status, Open Questions, Recommended Next Step."

Use this when an exec asks "what's the latest on the Acme thread?" and you have 45 minutes of reading ahead of you.

2. Reply on behalf of executive (decline)

Prompt: "You are an EA writing on behalf of [executive name]. Politely decline this meeting request: [paste]. Tone: warm but firm. No over-apologizing. Offer one concrete alternative if appropriate (route to me, take async, defer 2 weeks). Output: subject line + email body under 80 words."

The most common EA task. The most common AI failure mode: 200 words of apology where 40 will do.

3. Reply on behalf of executive (delay)

Prompt: "You are an EA. Write a reply that pushes back a deadline by [X days] on this request: [paste]. State the new commitment date. Give one honest reason (not five). No 'we apologize for the inconvenience'. Output: email under 70 words."

Asks for less than you might think, which is the whole point.

4. Daily inbox digest

Prompt: "You are an EA running morning inbox triage. I'll paste 15-20 email subject lines from this morning. Cluster them into 4 buckets: (1) needs exec decision, (2) FYI exec should know but not act on, (3) I can handle, (4) trash. List each subject line under exactly one bucket. If a subject is ambiguous, default to 'I can handle' rather than escalating. Output: 4 named buckets with subject lines listed beneath each."

What this prompt actually returns (tested 2026-05-17)

We ran Prompt 4 on free ChatGPT with this exact bracketed example baked in: 18 subject lines from a realistic morning inbox covering Q3 board materials, an urgent sales signoff, a $1.2M deal pricing approval, a calendar conflict, a Sequoia coffee invite, an EU GDPR audit, vendor invoice approvals over$ 50k, a SaaStr keynote invitation, a SaaS Weekly newsletter, and a family update from mom. Reproduced verbatim:

1) Needs Exec Decision

Re: Q3 Board Materials - draft attached
URGENT - Sales pipeline review needs your signoff
Re: ACME Corp pricing approval ($1.2M deal)
Speaking opportunity - SaaStr 2026 keynote
EU GDPR audit findings - action needed
Vendor invoice review (5 invoices pending approval >$50k)

2) FYI Exec Should Know But Not Act On

Out of office: Sasha Reyes
CFO 1:1 - prep doc enclosed

3) I Can Handle

Quick question on Friday's offsite venue
Lunch tomorrow? - Maya
Calendar conflict on Thursday - need to move 2pm
Holiday party logistics - need final headcount
Re: Re: Re: Re: marketing budget transfer
Auto-renewal: Zoom subscription
Coffee Tuesday? - John from Sequoia

4) Trash

Webinar invite: "AI in customer success 2026"
Newsletter: SaaS Weekly - top stories May 17
Family update from mom

Editor's notes. Six reads on what the model got right and where you'd override. (1) Structure landed cleanly: 4 named buckets, all 18 items placed, no duplicates. The model also followed the "default to 'I can handle' for ambiguous items" rule honestly (no over-escalation). (2) The biggest miss is dropping "Family update from mom" into Trash. That's exec-personal correspondence; it should never be inside a triage system at all, let alone classified as trash. Override before showing the bucket to the exec or, better, scrub personal email out of the digest entirely. (3) "Coffee Tuesday? - John from Sequoia" went to "I can handle." That's a judgment call most EAs would push the other way: a VC coffee on short notice is usually a yes/no/route decision only the exec can make. Promote to "Needs Exec Decision" for most fundraising-aware execs. (4) The SaaStr keynote correctly went to "Needs Exec Decision," but the model did not flag the date-deadline that "speaking opportunity" almost always carries. Note the urgency when you walk the bucket to your exec. (5) "Re: Re: Re: Re: marketing budget transfer" landed in "I can handle." Four levels of re: usually means someone is stuck or someone is angry. Worth opening the thread before you confirm you'll handle it. (6) "Auto-renewal: Zoom subscription" went to "I can handle," which is appropriate, but only if you actually run the quick seat-count sanity check that auto-renewals are supposed to trigger. Net: the model gives you a clean first pass that saves 5 minutes of manual sorting; you'll override 2-3 items based on context the model can't see (especially the personal-mail leak and the VC-coffee call).

Prompt: "You are a research analyst. I'll paste the intros from 5 industry newsletters I read this morning. Give my executive a 5-bullet brief covering only items relevant to [our industry / company size / function]. Max 15 words per bullet. No marketing fluff. Cite which newsletter each came from. Output: bulleted brief."

Replaces the "is anything in these 5 newsletters worth my exec's 90 seconds" decision.

Calendar and scheduling

6. Reschedule with multiple stakeholders

Prompt: "You are an EA scheduler. I need to reschedule a meeting with [N attendees] from [old time] to one of these alternatives: [list]. Tone: neutral, no over-explanation of why we're moving. Give attendees a clear path to confirm or counter-propose. Output: a single email with 3 proposed times and 1 line for how to respond."

The hardest part is the over-explanation. This prompt strips it.

7. Calendar audit for deep work

Prompt: "You are an EA. I'll paste my executive's calendar for next week. Identify three things: (a) 3-hour focus blocks I can protect from new bookings, (b) meetings that look stackable or could be moved without consequence, (c) any recurring meetings that don't have a clear owner this week. Output: 3 sections with specific time blocks and a 1-line reason each."

Weekly Friday-afternoon job that pays for itself in Monday-through-Thursday deep work for the exec.

8. Prep doc for tomorrow's meetings

Prompt: "You are an EA. I'll paste tomorrow's 4 meetings with title, attendees, and 1-line brief. For each, give: 1-sentence purpose, 1 person my exec should know about (background line, only facts I provide), 1 question my exec should be ready for. Max 50 words per meeting. Do not invent biographical facts about attendees; if I haven't given you their background, say 'need from me'. Output: structured prep doc."

The "do not invent facts about people" line is load-bearing. Without it, the model fabricates bios.

9. Reschedule reply when someone else cancels

Prompt: "You are an EA. Someone canceled and needs to reschedule. The original meeting topic was [topic]. Reply with 3 alternative times and a 1-line reminder of why the meeting matters. Max 50 words. No 'Sure, no problem'. No calendar booking links unless I include them. Output: short email."

Use this when you don't want the reply to read like a form letter.

10. Common availability across multiple calendars

Prompt: "You are an EA. Given these 5 attendees' availability (paste), find 3 time slots that work for all 5 within the next 2 weeks. Prefer 30-min over 60-min. Prefer 10am-3pm [timezone]. Flag if no slot exists. Output: numbered list of times with timezone, plus a fallback if you can't find 3."

Better than the average scheduling tool because it respects time-of-day preferences.

Travel and logistics

11. Flight and hotel decision matrix

Prompt: "You are an EA. My executive is traveling to [city] for [purpose] on [dates]. Given these flight options [paste] and hotel options [paste], build a decision matrix scoring on (a) total trip cost, (b) total travel time including layovers, (c) hotel-to-venue distance, (d) any red eye risk. Output: markdown table with all options + my recommendation in 2 sentences."

Saves the worst part of travel booking: the spreadsheet that nobody writes down.

12. Travel itinerary document

Prompt: "You are an EA. Build a travel doc for [exec name]'s trip to [city] on [dates]. Include flights with confirmation numbers, hotel address + phone, ground transport options at the airport, dinner reservation, key meeting locations with arrival buffer estimates. Constraints: 1 page, scannable on phone, key numbers in bold, no marketing language. Output: structured itinerary."

The one document the exec actually opens at the airport.

13. International travel checklist

Prompt: "You are an EA. My exec is traveling from [home country] to [destination country] for a [business / multi-stop / conference] trip. Generate a checklist of documents and preparations they need at 30 days out, 7 days out, 1 day out, and at departure. Do not invent visa or vaccination rules; flag anywhere I need to verify with the embassy or a travel service. Output: 4-stage checklist with caveat note at the top."

Anti-hallucination instruction is mandatory here; visa rules are exactly where the model will make things up.

14. Local recommendations for a client dinner

Prompt: "You are an EA who knows business hospitality. My exec is meeting [client name / company] in [city] for dinner. Suggest 3 restaurants suited to a business dinner: [cuisine preference, dietary needs, price tier]. Must be reservation-friendly. Must not be too loud for conversation. Prefer something near [neighborhood] if possible. Output: 3 named options with a 1-line read on each + a caveat that I will verify availability myself."

The caveat is the point. The model will recommend restaurants that closed last year.

15. Conference prep doc

Prompt: "You are an EA. My exec is attending [conference name] on [dates]. Given the agenda [paste], identify (a) 3 sessions that match [our company's current priorities], (b) 5 people from the speaker list my exec should try to meet, (c) one or two meetups or dinners that fit the networking goal of [goal]. Output: prep doc with 3 sections."

Turns a 60-page conference agenda into the 1 page the exec will actually read.

Research and background

16. Background research on a person

Prompt: "You are a research assistant. I'm prepping my exec for a first meeting with [name], [title] at [company]. Give me a 1-page background brief covering: their role and focus, any public posts or articles they've written that you can name, shared connections that we would realistically know about, 2 topics they would likely want to discuss. Only cite sources you can actually name. Flag anything you cannot verify."

The "only cite sources you can name" line is what turns this from a hallucination factory into a usable brief.

17. Background research on a company

Prompt: "You are a research assistant. I'm prepping my exec for a meeting with [company name]. Give me a 1-page company brief covering: what they sell and to whom, recent funding or news in the last 6 months, 2 named competitors, their typical buyer profile, one open question my exec should be ready to ask. Name your sources. Flag any unverified claims. Output: 1-page brief."

Same anti-hallucination posture, applied at the company level.

18. Compare 3 vendor proposals

Prompt: "You are a procurement analyst. I'll paste 3 vendor proposals for [service]. Build a side-by-side comparison on (a) total price, (b) deliverables, (c) timeline, (d) terms that look risky to me, (e) what's missing from each proposal. Output: markdown table + a 1-paragraph recommendation with reasoning."

The "what's missing" column is the most useful one.

19. Process a board document into an exec summary

Prompt: "You are a research assistant. I'll paste a 20-page board document. Give my exec (a) the 5 most important decisions or asks in the document, (b) any number my exec needs to memorize before the meeting, (c) anything that contradicts something the leadership team said last quarter. Do not summarize the whole document; pick only the items the exec needs to hold. Output: 3 labelled sections."

A general "summarize this" prompt produces a generic summary nobody reads. This one filters.

20. Press and news monitoring brief

Prompt: "You are a research assistant. I'll paste 10 news article headlines from today. Tell me which 3 my exec needs to see (and why in one line), and which 7 they can skip. Prioritize anything mentioning [our company], [our top 3 competitors], or [our regulators]. Output: 'Read these 3' + 'Skip these 7' with one-line reasons each."

The skip section matters as much as the read section.

Documents and exec support

21. Leadership meeting agenda

Prompt: "You are an EA helping the exec prepare an agenda for a 60-minute leadership meeting covering these topics [list]. For each topic give: timebox in minutes, owner, 1-sentence decision needed. Protect 10 minutes at the end for 'any other business'. Output: structured agenda with running time markers."

The decision-needed column forces the agenda to be useful, not performative.

22. Pre-read for a meeting

Prompt: "You are an EA. Convert these bullets [paste] into a 1-page pre-read for my exec's meeting with [audience]. Lead with the decision needed (not the background). Put context in the back half. Output: 1-page memo with a clear hierarchy: Decision Needed, Background, Options, Recommendation."

Most pre-reads bury the lede. This forces it to the top.

23. Expense report descriptions

Prompt: "You are an EA. I'll paste 8 raw expense line items. For each, write a clean business-purpose description for our expense system. Under 15 words per line. Factual, no embellishing. Flag any line that looks borderline (could be questioned in audit) so I can ask the exec before submitting. Output: table with original, cleaned, and flag column."

The flag column is what makes this safe to actually use.

24. Warm note for a milestone

Prompt: "You are an EA writing on behalf of [exec name]. Write a warm note for [person]'s [birthday / work anniversary / new role announcement]. Tone: warm, not over-familiar. Reference one specific thing this person did or that the two of them know each other for. Max 40 words. No 'many happy returns', no 'wishing you continued success'. Output: 2 short variants for the exec to pick."

Avoids the corporate-greeting-card tone that real recipients can smell.

25. Out-of-office handoff document

Prompt: "You are an EA. Build a 1-page handoff document for my backup covering [date range]. Include (a) my exec's standing meetings during the window and which can move, (b) 3 stakeholders the backup is most likely to hear from and how to handle each, (c) any pending tasks with deadlines that fall inside the window. Output: 3-section doc with clear escalation guidance."

The version of this you'll thank yourself for writing the morning you actually take vacation.

What ChatGPT should not do for your role

Five things to keep out of the workflow no matter how convenient.

It should not write replies that misrepresent the executive's calendar, position, or commitments. If the model invents an availability or a soft yes, the exec is on the hook for it.

It should not store or process anything that names a specific employee in a sensitive context (performance issues, comp, terminations, health, family situations). Free-tier ChatGPT is not a private workspace.

It should not generate background research about a person and present it as fact without source citations. Hallucinated bios make exec meetings start on the wrong foot.

It should not draft sensitive negotiation language, board-confidential summaries, or anything covered by an NDA on the free tier. Use a paid plan with data controls or do it yourself.

It should not be the final read on anything with money, legal exposure, or regulatory impact. Vendor contracts, expense audits, GDPR replies, sponsor agreements: assistant-helps-EA, EA-checks-with-Legal, Legal-signs-off.

FAQ

Can I really paste my executive's calendar into ChatGPT?

Not on the free tier with real names and titles. Anonymize first ("Mon 9am: CFO 1:1; Mon 2pm: customer call with top-5 account"). On a paid tier with explicit data controls and no-training settings, it's a separate decision your IT or legal team should weigh in on. When in doubt, anonymize before pasting.

How do I get the model to stop being so apologetic in reply drafts?

Put it in the prompt explicitly: "No 'apologies for the inconvenience', no 'thank you so much for your patience', no 'please don't hesitate to'." Negative constraints work better than positive ones with this kind of polite-script drift.

Will these prompts work on Claude or Gemini instead of ChatGPT?

The role-task-constraints-output structure is portable. Claude tends to be slightly more conservative and is a better fit for sensitive comms (declines, deferrals, difficult notes). Gemini is sometimes more verbose. If you have access to more than one, run the same prompt on two and pick the better reply.

Can the model find me real flight prices and hotel rates?

Free-tier ChatGPT cannot reliably look up live prices. Paste the options you have from your travel platform and let the model build the decision matrix; don't ask it to search for and price flights on its own.

How do I keep my exec's voice when the model drafts on their behalf?

Paste 3-5 short samples of how the exec actually writes (cleaned of confidential info) and add "match the tone and rhythm of these samples" to the prompt. The first draft will be closer; the second draft after you red-line is what ships.

Three more posts in this cluster that overlap with the EA workflow above:

The PRSO framework for writing better ChatGPT prompts explains the structure behind every prompt in this list.
ChatGPT prompts for HR and recruiting covers people-ops adjacent jobs (JDs, onboarding, candidate comms) that EAs often partner with HR on.
How to use ChatGPT for project management covers status reports, stakeholder comms, and meeting pre-reads that EAs run alongside PMs.

Bookmark this page, pick one prompt for tomorrow morning, and override the model's judgment where your context says you should. That's the