How an event planner prices flower arrangements with ChatGPT

Flowers are the largest recurring cost in event planning, and pricing them used to be the slowest part of Sara Rojas's week. She now screenshots an arrangement she likes, sends it to ChatGPT with a colour theme and a budget, and gets back stem counts, seasonal substitutions and a per-bouquet breakdown. She estimates it halves the time, and it removed a class of costing error she had been making repeatedly.

What the job used to involve

Every client arrives with a colour theme, sometimes with specific flowers named, and always with a budget. Her wholesaler's catalogue runs to thousands of entries. Availability shifts by season. Prices shift by holiday. Because she came into the business without a floristry background, each new request meant looking flowers up one by one, then going to the wholesale market and walking the floor.

"It took me hours, and it took me being at the wholesale a lot of time and talking to the salespeople," she says. The walking was not wasted — it is how she learned what pairs well — but it was an expensive way to answer a question about cost.

The workflow now

She finds an arrangement she likes, screenshots it, and sends it to ChatGPT with the colour theme and the number of arrangements needed. What comes back is the part she used to grind out by hand:

• How many flowers the arrangement needs
• Which of them are in season in her area
• What fits the client's budget, and what to substitute when the first choice does not
• A per-bouquet count, including greenery

The substitution step is the one she names as most valuable. When a client wants a flower that is out of budget, the question is not whether it is expensive but what looks close enough for less. That used to be a conversation with a salesperson on the market floor.

The error it caught

Her account of what she was getting wrong before is worth quoting at length, because it is a specific failure mode rather than a general inefficiency.

"I think I hit myself over the head a lot because I was just thinking of it looking visually nice. And then when I saw it, I was like, wait, I didn't really add in like the time that spends and all the other things that you have to input making flower arrangements."

She was costing the flowers and not the work. The per-bouquet breakdown forces the second number into view, because a count of stems and greenery per arrangement multiplied by ten arrangements is also a count of labour.

The part she wanted back

Asked what the tool actually gave her, she does not say money. She says the annoying half of the job disappeared. "That whole part just skips and I can go actually into what I like to do, which is decorating and doing the flower arrangements."

That is the same shape as her inbox result: the model took the deliberation, not the craft. She still chooses the flowers and builds the arrangement. What she stopped doing was looking up names and reconciling a wish list against a budget.

What is still broken underneath

The costing works because she supplies the catalogue knowledge. ChatGPT does not know what her wholesaler has in stock this morning. The wholesaler's own website lists what it can source rather than what is on the floor, tropical stems need a week or two of lead time, and many entries carry no photograph, so she still ends up searching for a name to see what a flower looks like.

Prices move too. Around Valentine's Day everything is marked up, and she used to quote off-peak prices into holiday weeks and absorb the difference. She knows to check now. Knowing to check is not the same as being told, and that gap is a product request in its own right.

Where this comes from

S2E2: The secret AI strategy this small NYC business uses to save thousands — the full interview with Sara Rojas and Luisa Ramirez. Listen or watch: YouTube, Spotify or Apple Podcasts.