AI that does the setup typing
Build an event from a poster. Turn a picture of your seating chart into a working seat map. Import a menu from a photograph. What it will not do is decide anything that costs you money.
BoxOfficeTech uses AI for the setup work, not for decisions. It drafts event descriptions, categories, tags and performers; reads an event poster and a printed menu; turns a photograph of a seating chart into an editable seat map with real sections, rows and seat numbers; and proposes portal branding from your existing website. It never generates prices, dates, venues or capacities — those come from you, and everything it produces is reviewed before it goes live.
What it does
Each of these is a button in the admin, and each is verifiable in a trial.
A seat map from a picture
Upload your venue’s chart. Individual seats are found in the image, grouped into sections and rows, numbered, and oriented to the stage — then you adjust rather than build.
An event from a poster
Photograph the poster. What is printed on it lands in the event form; what is not legible stays empty.
Descriptions, drafted
Give it the event name and it drafts the description, category, tags and performers. Shorten, lengthen or polish what you have with one click.
Search listings, written
Page titles and meta descriptions generated for each event, so your listings read properly in search results without you writing them.
A menu from a photo
Photograph a printed menu or upload a PDF and tick through the items it found. Unpriced and flagged, ready for you to price.
Section labels read off the chart
The printed section names, the stage position and shape, and the price-tier legend are read from the image and applied to the map.
Your brand, from your website
Point it at your existing site and it proposes a palette, fonts and logo for your portal from what is actually on the page.
Venue details, constrained
Venue descriptions are written from known facts about the venue — a real source always wins over generated prose.
What it will not do
The limits are the design, not a disclaimer.
It never sets a price. Not on a ticket, not on a seat category, not on a menu item imported from a photograph that had the prices printed right there. Imported items arrive at zero and flagged. A wrong price is the one mistake that moves real money before anyone notices, so it is the one field we will not guess at.
It never invents a fact about your event. Dates, venues, capacities and performers are read from something you gave it or left empty. A poster scan that cannot make out the date returns no date — it does not produce a plausible one.
Nothing publishes itself. Every one of these features fills in a form you then look at. There is no path where a model’s output reaches a buyer without a person having seen it first.
Nothing runs in the background. Each feature is a button. You will never find that something was rewritten, re-categorised or re-priced while you were not looking.
And some things we deliberately do not call AI
Because calling them AI would make the rest of this page less believable.
Plenty of software labels ordinary logic as AI. We had that problem ourselves and fixed it: the dashboard used to badge its readings as AI insights when they were threshold rules, so we removed the badge and now state plainly that no model is involved. The rules were useful; the label was not true.
Three things in this platform sound like AI and are not. Best available seats is geometry — it ranks seats by distance to the stage and keeps a group together. Portal search is fuzzy text matching with some date and price parsing, which is why it is fast and free. Dashboard readings are threshold rules you can check against the numbers above them.
Seat detection in a chart image sits in an interesting middle: it is real computer vision, running on your own device, with no model call and no cost. We describe it as what it is rather than folding it into a broader claim.
Specifications
| Capability | Behaviour |
|---|---|
| Event description, category, tags, performers | Generated — you edit before publishing |
| Event poster reading | Reads printed text; illegible fields left empty |
| Seat detection from a chart image | On the device — no model call, no cost |
| Section labels, stage, price legend | Read from the image by a vision model |
| Menu import | Photo or PDF; items reviewed before creation |
| Ticket prices | Never generated, suggested or filled |
| Dates, venues, capacities | Never generated |
| Menu item prices | Never read across — items arrive unpriced and flagged |
| How features run | On demand, per button press — nothing runs in the background |
| Models | Google Gemini, with Anthropic Claude as fallback |
| Training on your data | None by us |
| Additional cost | None — included in the platform |
Questions we get about the AI features
It can write the parts that are writing. Give it an event name and it drafts a description, suggests a category, tags and performers — then you edit it. It does not invent dates, venues, prices or capacities, because those are facts about your event that only you have.
Bring a seating chart and a poster
The fastest way to judge this is to watch your own venue’s chart become a seat map.