AI email parsing — from the request to the quote lines
"Hi, I'd like 500 business cards 9x5 in two colours on 350gsm board, plus 1,000 A5 flyers, four colour double sided, matt laminated." You retype that email into quote lines ten times a day. AI Analyze does it in two seconds — and, more to the point, it tells you what the customer didn't tell you.
Where to press
Two routes, same result:
- On the Email page, open the message and hit "New Quote from this email". The quote is created and the AI starts reading immediately — "AI is reading the email and preparing the items…".
- Inside an existing quote, in the AI side panel, hit "AI Analyze" on a linked email or on one you find by search (
from:customer@...).
The result lands as line items on the quote, not as text for you to rework.
It isn't only about printing
A print shop doesn't only sell printed paper, and the AI knows it. It first works out what kind of job is being asked for, then keeps only the fields that make sense:
- Stamps — quantity and size. It won't ask what paper a rubber stamp is on.
- Signage, banners, vinyl — here "paper" means material.
- Services (installation, delivery, artwork) — no print specs at all, just the description.
- Unknown — it writes it down the way the customer said it, without interrogating it about paper.
That's the difference between an assistant and a questionnaire.
The hierarchy that doesn't allow lies
This is the heart of it. Every field on an item belongs to one of three categories:
1. Confirmed — the customer said it. It goes in as fact.
2. Suggestion from history — they didn't say it, but this same customer has ordered the same type of job before. The AI fills in the old value and marks it "suggestion from history", with the quote and date it came from (e.g. QT-2026-0288 · 12/05/2026). It's an assumption, and it looks like one.
3. Missing — not stated and not in history. Marked "Pending from customer".
⚠️ What it never does: fill in a "default". No four-colour because that's usual, no 300gsm because it's reasonable. A silent default in a quote is a wrong price you discover on delivery day.
The button that buys back the hour
When details are missing, hit "Email customer for missing details". PressCal drafts the message with the questions that need asking — only for the fields that make sense for that product — and puts it in the Reply: "Draft added to Reply — review and send".
You read it, edit it if you like, send it. Nothing goes out without you.
The customer is matched automatically
The sender's address is matched to a contact or company on its own. That's also where the history for suggestions comes from — so the tidier your contacts, the smarter the AI looks.
What language it writes in
Free text (description, notes, summary) is written in the language of your organisation's country, not the language of the email. A Greek shop receiving an English email gets a Greek description, because the description is read by your own people.
What to realistically expect
It reads normal customer emails well, even when they don't use correct terminology ("sort of little cards, stiff paper"). It struggles where you would struggle: vague quantities ("a few"), specs buried in an attached PDF instead of the body, messy threads with five changes of mind.
Always look at the result before you send a price. The AI sets up the scales; you're still the one weighing.