Write what it should say, yourself
You know how your business talks better than we do. So the instructions are yours to write, in plain language, and yours to change at four in the afternoon without raising a ticket with anybody.
What it does
- Creates a new assistant from the dashboard — as many as your plan includes — with a starting manner chosen from four kinds.
- Takes its instructions in plain language: who it is, what it may promise, what it must never do, what to do when it does not know.
- Separates saving from publishing, so nothing you type reaches a customer until you say so.
- Keeps every version, with Restore to bring an older published one back and Unpublish to stop it being used at all.
- Runs a setup check on every change and tells you in plain words what is wrong, not a stack trace.
- Tests live in your browser through the real pipeline, so what you hear is what a customer would get.
- Keeps the platform’s own safety rules underneath whatever you write.
How it works
- Create the assistant. It starts as a draft, switched off on every channel — nothing you create can reach a customer by accident.
- Write its instructions and save a draft. Saving changes nothing a customer hears.
- Publish. The setup check re-runs, and if it passes the assistant goes live by itself.
- Switch on the channels you want it to answer on, one at a time.
- Change it whenever you like. The old version is still there if you want it back.
Where it stops
Every product page in this market lists capabilities. This is the other half — the things it does not do, so you find out here rather than in month two.
- Instructions are behaviour, not facts. Prices belong in your catalogue and information belongs in your knowledge base — an instruction block full of prices goes stale the week you change them.
- Your plan sets how many assistants you can have, drafts included, and raising it is not self-serve today.
- The platform’s guardrails stay on underneath whatever you write. You cannot instruct your way past them, and that is not configurable.
- A fill-in slot we do not recognise blocks the publish. The save warns you and names the offending token rather than letting it reach a customer as literal braces.
- You describe behaviour; you do not draw it. There is no visual canvas where the conversation is boxes and arrows, and there will not be one — if a decision tree is what you want to build, we are the wrong product.
Common questions
Do I need Kooday to change what my assistant says?
No. Write it, save it, publish it — it is live from that moment. The only things that still come through us are raising your assistant limit and setting up a new fill-in token.
What if I publish something that makes it worse?
History lists every version and Restore brings an older published one back. Unpublish stops it using your instructions at all while you think. Nothing you do here is one-way.
Can I make it say anything I want?
Within your own business, yes — how it talks, what it promises, what it refuses. What you cannot do is remove the platform’s own safety rules, which stay underneath every assistant regardless of what its instructions say.
Worth reading first
Related
Would this work for your business?
Message us and say what your week actually looks like. You will be talking to our own assistant — the thing this page describes, doing the job on our side of the conversation.
Ask us on WhatsApp