AI assistants
We configure assistants around approved information. They can classify enquiries, draft replies and hand conversations to a person. We test their limits with sample questions before launch.
What it can take
It sorts what comes in and drafts a reply in the way you already write. The questions that belong to it are named before anything is switched on.
What stays with you
You still read anything that should not go out on its own. A question outside the ones you named comes back to you, with the person attached.
It does not run the business.
IN PRACTICE
We design when the assistant should stop, too.
Explaining opening hours is different from giving a bespoke quote. We separate questions answerable from approved information from uncertain or sensitive requests that need a person.
Get the starting point clear.
Prepare recurring questions, approved answers and topics the assistant must not decide. We agree the channel and the limits of its access to your tools.
Will every answer be correct?
No. AI responses can be wrong. We limit the scope, test it and use human review where needed. Examples from live use can inform further improvements.
Tell us the questions, and which ones must always come back to you. The fee is agreed with that limit, not after it.
Tell us the questions