Give it a poster, get a complete event

Setting up an event is about forty fields. Title, description, category, venue, capacity, door times, ticket tiers, prices, availability windows, refund terms, images, tags. Most of that information already exists — it is on the poster, in the email thread, in the venue booking confirmation — and retyping it into a form is the single most tedious hour in this job.

The AI Event Architect takes whatever you already have and produces a complete draft. A photograph of a flyer is enough. So is a paragraph typed badly into a box at midnight.

What comes back

• Title, long description and short summary, written in the voice of the event rather than in press-release English
• Category and sub-category, matched against the taxonomy the search filters actually use
• A tier structure with prices, quantities and on-sale windows
• Door times, age policy and accessibility notes where the source material implies them
• Suggested tags, which is what most organisers skip and then wonder why discovery is poor

Pricing arrives with its working shown

A number with no reasoning behind it is worse than no number, because you cannot tell whether to argue with it. Every price the architect proposes comes with the comparison set it used: similar events, in a similar city, at a similar capacity, in a similar month, and what they charged. If you disagree, you can see exactly which assumption to change.

It also proposes the tier ladder rather than a single price, because the ladder is where the revenue is. Setting general admission first and adding VIP as an afterthought is the most common pricing mistake in the business.

It creates a draft. It never publishes.

This is a hard boundary, not a setting. No agent on this platform can put an event on sale, take money, or change a price on a live event. The architect writes a draft into your dashboard and stops. You review it, change what is wrong, and press publish yourself.

The reason is simple: a model that misreads a flyer and publishes a £5 ticket for a £50 event has sold real tickets at the wrong price to real people, and those sales are binding. There is no version of that failure that is recoverable, so the capability does not exist.

Every field says where it came from

Each generated field carries its provenance: extracted from your source material, inferred from comparable events, or invented as a placeholder. Placeholders are visually flagged and block publishing until you deal with them. You should never discover after the fact that the model guessed a door time.

What it costs before you run it

A full build is 35–45 ACU and the quote appears before the run starts, not after. Regeneration is scoped — redoing just the pricing costs a fraction of redoing everything, so iterating on one section is cheap.

The images are the part with real risk

Generated imagery is offered, watermarked as generated, and never applied to an artist likeness. If your source material contains a photograph of a performer, it is used as supplied or not at all. Synthesising a picture of a real person who will be on that stage is a legal problem and an ethical one, and no amount of convenience justifies it.