AI Prompts for Event Planners: Vendor Quotes, Run-of-Show and Day-Of Logistics

It's the Thursday before a 120-guest event. Two caterers have sent quotes. One says 780 per head, the other 720. You paste both into a chatbot and ask which is cheaper. It says the second, confidently. It's wrong, because the second quote leaves out staff and travel, and nobody checked the sum.

AI prompts for event planners work when you give the model the messy inputs, ask it to structure and question them, and keep arithmetic, safety and the final call with you. Below are prompts for vendor quotes, the run-of-show, vendor chasing and a day-of sheet. All of them are untested templates; I haven't run them on a live event. The numbers in the quote example come from a short Python script I ran on 2026-10-09, with invented inputs.

This post is about logistics. For the operations side of a regular business, AI prompts for operations managers covers SOPs and checklists, and AI prompts for project managers covers status and risk in more depth.

How do I compare vendor quotes like for like?

Make the model do the reading and you do the maths. Quotes are written to look good, not to be comparable.

I'm comparing vendor quotes for [SERVICE: catering / AV / decor / photography]
for an event on [DATE] with about [N] guests (range [LOW] to [HIGH]).

For each quote below, extract into a table:
- What is included (quote the wording)
- What is explicitly excluded
- What is NOT MENTIONED but would normally be needed for this service
  (staff, equipment hire, setup and breakdown time, travel, overtime,
  power, permits, cancellation terms, payment schedule)
- Price basis: per head, flat, per hour, per item
- Anything ambiguous that I should ask the vendor to clarify in writing

Do not calculate totals. Do not pick a winner. End with a list of
questions to send to each vendor.

QUOTE A: [PASTE]
QUOTE B: [PASTE]

Then do the sums yourself. Here's the arithmetic for two invented quotes (currency units are made up, guest count is 120), computed by a script, not by a model:

Caterer A: food 93,600 + extras 18,000 = 111,600  ->  930 per head at 120 guests
Caterer B: food 86,400 + extras 30,000 = 116,400  ->  970 per head at 120 guests
100 guests: Caterer A 96,000 | Caterer B 102,000
120 guests: Caterer A 111,600 | Caterer B 116,400
140 guests: Caterer A 127,200 | Caterer B 130,800
Cheaper like-for-like at 120 guests: Caterer A, 111,600; 10% contingency = 11,160

Caterer B quoted 720 per head against A's 780, and still costs more once staff (20,000) and travel (10,000) are added, against A's equipment hire (18,000). The headline number and the real number point in opposite directions. The gap also moves with guest count, which matters if your RSVPs are still coming in.

If you want to reuse that, it's a dozen lines of Python or four rows in a spreadsheet. The point is that a model is good at spotting what a quote leaves out and poor at being your calculator; see spreadsheets vs Python with AI for when each fits.

How do I build a run-of-show with AI?

Give it durations you have confirmed, and ask it to find the problems.

Build a run-of-show for [EVENT TYPE] on [DATE] at [VENUE].

FIXED POINTS (do not move):
- Venue access from [TIME]; must be cleared by [TIME]
- [Keynote / ceremony / performance] at [TIME], duration [X] min
- Catering service window [TIME] to [TIME]

TASKS WITH DURATIONS I HAVE CONFIRMED:
- [Task, duration, who does it, who it depends on]

TASKS WITH DURATIONS I HAVE NOT CONFIRMED (flag these):
- [Task, my guess]

Produce:
1. A table: time, activity, owner, location, dependency, buffer after.
2. Buffers: add 10 to 15 minutes only after handoffs between vendors, and
   say where you added them.
3. A list of clashes: two things needing the same room, person or power
   at once.
4. A list of every assumption you made.
5. A "if this is late by 20 minutes" note for the three most fragile
   points.

Do not invent durations. If a duration is unknown, write [UNKNOWN: ask X].

The [UNKNOWN] rule matters. Without it, the model will assign plausible times to everything, including how long your venue takes to reset a room, and the sheet will look finished while resting on guesses. The 10 to 15 minute buffer in the prompt is a starting point I picked, not an industry standard; adjust it from your own experience of that venue.

How do I chase vendors without writing the same email ten times?

Use one fact base and generate each message from it, so nobody gets stale details.

Fact base (the only source of truth):
- Event: [NAME], [DATE], [VENUE ADDRESS]
- Load-in window: [TIME]; contact on site: [ROLE ONLY, placeholder]
- Guest count currently: [N] (final number due [DATE])
- Changes since the last version: [LIST]

Write a short email to [VENDOR TYPE] that:
1. States what we've agreed (from the lines below) and what is still open.
2. Asks for written confirmation of [SPECIFIC ITEMS] by [DATE].
3. Lists the three details they most often get wrong for this service
   (for example arrival time, power requirements, access route).
4. Is under 130 words and has one clear ask.

What we've agreed with this vendor (paste from your notes): [NOTES]

Send placeholders for personal details and fill them in yourself. If you run several events, a short vendor-specific note in your CRM or sheet beats a long prompt you re-type.

What goes on a day-of contact sheet?

The sheet people read at 7 am with a phone in one hand. Build it from your confirmed data.

From the data below, produce a one-page day-of sheet:
- Timeline (time, what, who) for the first four hours only
- Vendor table: name, role, arrival time, contact (use [PLACEHOLDER]),
  where to meet them, what they need from us on arrival
- Decision list: who decides what if X goes wrong (late vendor, weather,
  AV failure, guest count jump), by name or role
- Three things that cannot slip and the time by which I must know
- A "missing information" list: anything the sheet needs that isn't in the data

Use only the data below. Mark gaps as [MISSING].
[PASTE CONFIRMED RUN-OF-SHOW AND VENDOR DATA]

The decision list is the part most sheets skip. If nobody knows who can approve an extra hour of overtime at 9 pm, you'll find out at 9 pm.

A worked example (illustrative)

An invented scenario, not a real client. A 120-guest wedding reception, load-in at 2 pm, guests arrive at 6 pm, dinner service at 8 pm.

You paste the run-of-show prompt with the fixed points. The model returns a table and, in the clashes list, something like:

• The decor team and the sound check both need the main hall between 3 and 4 pm. Resolve with the venue.
• Caterer arrival is 5 pm, but power for the warming ovens is not confirmed. Ask the caterer.
• Your [UNKNOWN] list: time to flip the hall from ceremony to dinner layout.

That last item is the useful one. The model can't know the answer; the venue manager can. You've turned a vague worry into a question with an owner before the day, which is what the prompt is for.

Where this goes wrong

What not to automate

• Capacity, fire, medical cover, licences, insurance and food safety. Ask the venue, the authority and qualified providers. AI can help you draft the questions, not the answers.
• Contract terms and cancellation clauses. Read them yourself or have someone qualified review them.
• Vendor negotiation and relationship calls. A phone call with a caterer you've used for years does things an email can't.
• Anything with guest data in an unapproved tool. Dietary, accessibility and contact details are personal data; check your client agreement and the tool's terms. What not to paste into ChatGPT is a short checklist.
• Sending the final sheet unread. Walk through the run-of-show out loud with your team once. People will spot the missing hour.

If you're tuning the prompts, giving context explains why real durations and constraints beat clever wording. For a related hospitality angle, see AI prompts for restaurant owners, and the prompt library has more templates to adapt.