How it works

AskConcierge turns what you have already written about your event — a programme, a site, a PDF, a run-of-show deck — into a concierge your attendees can ask questions. You share a link or a QR code; they open it in any browser, with no app and no account.

It answers only from your content. It does not know things about your event that you have not told it, and when it does not know something it says so and asks you — once.

Everything the concierge knows comes from sources you add in the Knowledge tab. There are three ways in, and you can mix them freely:

• A website — give it a URL and it reads the page, or follows links across the site as far as this concierge’s pass allows.
• A file — PDF, Word (.docx), PowerPoint (.pptx), Markdown, or a photograph of a printed sign or timetable, which is read and transcribed.
• Pasted text — anything you can copy, including notes that exist nowhere else.

Older .doc and .ppt files are a different format from .docx and .pptx. Open them and use “Save As” first.

Sessions, speakers and venue details are extracted into structured records, which is what lets the concierge answer with a schedule you can add to a calendar rather than a paragraph. Everything else stays as text and is searched when a question needs it.

Before anyone arrives, the Gaps tab checks your content against the questions attendees ask at almost every event — arrival times, parking, accessibility, payment, lost property and the rest — and lists the ones your content cannot answer yet. The check reads your content only: it spends none of your questions and records nothing in your analytics. When you create a concierge, onboarding walks you through a few of these checks so you can see the result before you share anything.

• Answer it — type the answer, or hand it the page or file that contains it, and the question is checked again against what you added.
• It’s already there — say where. The claim is verified against your sources before the question is settled, and if a page could not be read, you are told which one.
• Dismiss — some questions are not yours to answer. A dismissed question stays dismissed on every later run.

An answer typed here expires when your event ends — next year’s answer should come from next year’s content. Answers you write in the help desk do not expire unless you give them an expiry date.

Most answers are a sentence. Some are better as something you can act on, so the concierge chooses a component: a schedule with add-to-calendar on every row, a speaker card, a map and directions, or copyable Wi-Fi credentials. Every one of those arrives with a sentence alongside it, never a wall of text repeating what the component already shows.

It replies in the language the question was asked in, even when your content is in another — and it never translates the names of stages, rooms or sessions, because an attendee sent to a room that does not exist under that name is worse off than one reading English.

When the concierge cannot answer something, it tells the attendee plainly and records the question in your Help desk tab. You answer it once. From then on, anyone asking the same thing — in any language, in different words — gets your answer directly, before the model is even called.

Your answer outranks your content. If the site says 14:00 and you answer 15:00, attendees are told 15:00.

An answer stays until you change or remove it. You can also give any answer an expiry date — useful for something true only this edition — after which the concierge falls back to your content.

For what is true right now and wrong in an hour: a stage change, a delayed shuttle, a set starting. Write it in the Announcements tab with a start and end time, and it appears in a bar at the top of every attendee’s screen for that window.

The concierge is told about live announcements too, so its answers agree with the bar. An attendee who reads “shuttles are delayed” and then asks when the shuttle leaves is not handed the unchanged timetable.

Publish, and the link is live. The Publish & share tab has the link, a QR code you can download as PNG or SVG for print, and a way to email invitations with both attached.

A public event needs no sign-in at all. A private event asks for an email address, checks it against your allowlist and sends a magic link — attendees see no other difference.

Traffic per day, what attendees asked about grouped into categories and their specific sub-topics, unanswered questions, and a CSV export. You can also ask your analytics a question in plain language — “which day was busiest”, “what could it not answer”.

Two more things come out of the same numbers. A print-ready post-event report — one page: what people needed that nobody had written down, what they asked about, and how often they had to ask twice; your browser’s print-to-PDF is the export. And during the event, if many people ask about the same thing in a short window and your content has no answer, the concierge emails your contact address while there is still time to say something.

Attendees are never identified to you. Nothing in the product — including the analytics assistant and the API — can tell you who asked what, and every number is derived from question shapes rather than transcripts, which are deleted on schedule.

Two things, and they are easy to confuse. A pass belongs to one concierge and sets what it can do — how many questions it answers, how many pages it reads, whether it carries your branding. A plan belongs to your workspace and buys credits, seats and support. Publishing a concierge spends credits at its pass; drafts are unlimited and cost nothing.

Every concierge starts on the smallest pass and grows into a larger one on its own, when what it actually uses asks for more. You are told when that happens, and only the difference is charged. A concierge that is already live is never paused for it — an attendee mid-event is never turned away because a balance ran out.

A published concierge stays live for twelve months and can be renewed. Preparing next year’s event while this year’s is running costs nothing.