Assistant knowledge
Choosing which of your data the assistant can draw on when it answers.
Choosing what your event assistant knows
Who this is for: whoever looks after the event's content — an organiser, a marketing lead, a programme manager. No technical knowledge is assumed. Nothing here can delete data: every switch in this guide changes what the assistant is allowed to read, never what exists.
How long it takes: about five minutes, and you can change your mind at any time.
This is the shareable copy of the guide built into the Command Centre. In the product it lives at Assistant Studio → Config → Knowledge → How this works, where the steps are filled in with your event's real data sources and your assistant's current answer. Use that version if you have access to it — this file exists so the steps can be emailed to somebody who does not.
What this is for
Your event has several kinds of information: the programme of sessions, the exhibitor list, the speaker line-up, your FAQs, the attendee list, and any documents you have uploaded.
The assistant can answer from all of it — or from only the parts you choose.
That choice is worth making deliberately in two situations:
- You have more than one assistant. A registration-desk assistant that only answers about the schedule and the venue is calmer and faster than one that will also volunteer exhibitor recommendations. (Each assistant is a face — one event, several front doors, each with its own name, look and embed link.)
- Some of your data is not ready. A half-loaded exhibitor list produces confident answers about seven of your ninety exhibitors, which reads worse than not answering at all.
If neither applies, leave everything switched on. This guide is entirely optional.
Where it lives
Assistant Studio → Config, then the Knowledge section — it is the second panel, under Identity.
Everything on that screen applies to the one assistant you have selected at the top of the page. If you have several, select the one you mean first: it is the single most common mistake, and the panel above the switches always names what the current selection covers, so you can check at a glance.
Step 1 — Read what it says today
Above the switches is one sentence, in plain words:
This assistant answers from Sessions, Speakers, FAQ and 2 knowledge libraries.
That sentence is the whole point of the screen. If it already describes what you want, close the page — you are done. If it does not, the switches below are how you change it, and the sentence updates as you go, so you can always see what you have just decided.
Step 2 — Turn things on and off
The switches are grouped by where the information comes from:
| Group | What it is | Where you manage it |
|---|---|---|
| Event data | Sessions, exhibitors, speakers, attendees, and your general info / FAQs | Data Hub |
| Knowledge libraries | Documents you have uploaded — a handbook, a policy pack, past articles | Data Hub → Libraries |
| Other data tables | Any extra tables built for this event, e.g. sponsorship levels | Data Hub → Schema |
Each name is your name for it. If you renamed "General Info" to "Event FAQs", that is what you will see here.
Switching something off does not delete anything. The information stays exactly where it is, and every other assistant that has it switched on keeps answering from it.
Step 3 — Attendees also needs a sign-in
The Attendees switch has a note beside it, and it is the one thing on the screen worth reading twice:
Only ever used for a visitor who has signed in. An anonymous visitor gets nothing from here, even with this on.
Turning it on does not make your attendee list public. It means: once a visitor has identified themselves, this assistant may use attendee information to help them — for introductions and who-should-I-meet answers. Somebody who has not signed in gets nothing from it whatever the switch says.
Step 4 — Save
Use the Save changes bar at the bottom of the screen, the same one every other setting on the page uses. Nothing takes effect until you do.
Then ask the assistant a question about the thing you just changed — the live preview beside the editor is a real assistant, not a mock-up. If you switched exhibitors off, ask it about exhibitors: it should tell you that is not something it covers, and offer what it does cover instead.
Questions people ask
I uploaded a document and the assistant ignores it
That is expected, and it is deliberate. Anything new starts switched off. A document you upload could be an internal budget as easily as a visitor handbook, so nothing becomes readable until somebody says so.
Go to Knowledge, find it in the list — it will be there, switched off, never hidden — turn it on and save. If you have several assistants, do it for each one that should have it.
Can two assistants cover different things?
Yes, and that is what the feature is for. Select each one at the top of the page and set its switches separately. The badge above the panels tells you how many sources each one covers, and a "Match primary" button copies your main assistant's answer onto another one if you would rather they agreed.
It says it "can't help with that" for something I know we have
Three things to check, in this order:
- Is the right assistant selected? The switches you edited may belong to a different one.
- Is that source switched on for this assistant? Look at the sentence above the switches, not the switches themselves — it is the shorter read.
- Is the data actually loaded? Data Hub shows how many records each source holds. An empty source that is switched on produces exactly the same "I can't help with that" as a full source that is switched off.
What happens if I switch everything off?
The assistant will still talk — it just will not answer from your event's data. That is a real, supported setup (a greeter that only explains what the event is and hands people on), and the screen says so plainly when you reach it. It is never something you should arrive at by accident.
Where is web search?
It is not on this screen, on purpose. Web search is a tool the assistant can use, not a source of your event's information, so it lives with the other tools: Config → Surfaces → AI pipeline → Configure tools, per assistant. The Knowledge section carries a line pointing there so nobody concludes the assistant cannot search the web.
Does this stop anyone else reading the data?
No — and it is worth being precise, because the difference matters. These switches decide what this assistant will answer from. They are not a permission system for your data, and they are not a substitute for one: your team's access is set by roles, and the public data API has its own settings. Think of it as choosing what one front-of-house assistant is briefed on, not as locking a filing cabinet.
Related: Putting the event assistant on your own web address
