How you teach Anna your business

Gauti runs a small bakery in Hafnarfjörður. The first morning after he signs up, he sits down with a coffee and tries the chat as if he were a customer: “Do you have gluten-free sourdough?” Anna answers right away, because she has already read his website, and says yes. But she adds that it is available all day, and that is not right. The gluten-free sourdough is baked in a small batch and usually sells out before noon. Gauti corrects her on the spot, approves the correction, and moves on. Two days later a real guest asks the same thing at three in the afternoon, and Anna answers that it is sold out today but comes fresh in the morning. Nobody taught her that answer by hand that day. Gauti did it once, on a Monday, over a cup of coffee.

Like a new hire

How do you teach an AI about a business? Basically the way you teach a new hire. Anna knows nothing about your company when she arrives; she knows how to talk, not what you sell or when you are open. You build her knowledge in a few steps, and none of it becomes an answer until you have approved it.

The ways you build the knowledge

You decide where to start. Her dashboard lays the paths out for you and you take them in whatever order suits you.

• The fastest start is a website scan. You give Anna your site's address, she reads it and sums up what she found, and you review and approve. She re-scans at regular intervals and tells you what has changed, so the knowledge does not trail behind old information.
• You upload documents. Price lists, menus, service descriptions in PDF. Anna reads them, including text inside images, and adds it to what she knows.
• Products and services go in as their own entries she can compare and present to a guest.
• Common questions you record yourself, question and answer, and they are the first thing Anna reaches for when asked, because you confirmed the answer.
• And then the basics themselves: opening hours, location, contacts, and the rules guests ask about most.

None of this is a one-time job that closes afterward. You can add, fix, and revise whenever you like, since Gauti's bakery looks different in December than it does in July.

Training in chat

Documents get you a long way, but the best training is a conversation. In the dashboard you talk to Anna as if you were a guest, question her, and see exactly how she answers before anyone else does. When she gets it right, you let it stand. When she misses, you correct her right there in the chat, tell her what is correct, and the answer updates. This is closer to telling a new hire about the company than to filling in forms, and that is by design.

Some people prefer to let Anna lead. Then she interviews you, asks about the business step by step, and builds the foundation out of your answers.

You approve it, then it counts

Here is the part that matters most, and the part that separates Anna from a chatbot that guesses. Anna does not learn on her own and changes nothing without you. When she proposes an answer, whether it came from a correction, a website scan, or a hard case a team member resolved, the suggestion waits for your approval. You can approve it right there in the chat where the correction happened, or go through the suggestions together in one place when it suits you. Nothing becomes an answer a guest sees until you have signed off on it. You teach her once, you approve it, and after that it holds.

She tells you where she stands

You do not have to guess whether Anna is ready. The dashboard shows you how ready she is, which topics she has good answers for and where the gaps are. She flags the conversations where she had no good answer herself, so you see what still needs teaching instead of finding out by chance. When a matter is passed to a team member and they resolve it, you are offered the option to turn that resolution into an answer Anna knows next time. And she flags knowledge that may be going stale, so an old price or a past event does not sit there and turn into a wrong answer.

Frequently asked questions

Do I need to be technical to teach her? No. The training happens in conversation and by uploading documents you already have. If you can tell a new hire about the company, you can teach Anna.

Does she learn on her own without me knowing? No, and that is the key point. Anna suggests answers but changes nothing until you have approved it. That way you stay in control of every single answer a guest receives.

How long does it take to get her going? The website scan gives her a foundation right away, often on the first day. After that the quality depends on how much you teach her, and a few hours of training early pays off many times over once the questions start piling in.

Can I change what she knows later? Yes, any time. You fix answers, add new ones, and remove what no longer applies, for instance when a price changes or you add a service.

This is why Anna answers the way she does

A generic chatbot off the shelf knows how to talk but does not know you. It answers the same wherever it sits, because it has learned nothing about your bakery, guesthouse, or clinic. Anna is built on the opposite premise, that the answers should come from your own knowledge and that you have approved them. That is why teaching her is not a side task but the core of how she works.

And that is where the part nobody can copy lives. Flawless Icelandic is a good start, but more than one provider offers it. What is only yours is everything Anna has learned about your business, the corrections, the approved answers, the exceptions your staff resolved. The more you teach her, the better she knows your business, and the more valuable she becomes to the company. It starts with a single cup of coffee on a Monday morning.

See what Anna does or try the chat right now, no signup needed. And to dig deeper: you're not getting a chatbot, you're getting Anna and when a team member takes over the chat in real time.

← All articles