Questions, answered

A good one says so and fetches a person. A NanoPorter assistant answers only from the documents the business has written, and it is not permitted to fill a gap from general knowledge, so the usual way a chatbot embarrasses a business is closed off. When it cannot answer, it says so and offers to take your details.

It should not, and ours does not. Each business sets what its assistant must never discuss, and for a dental practice, a clinic or anything regulated that list is written before the assistant goes live. Watch the dental example on the main page: it will not advise, and it gives the emergency route instead.

It stops, says it is handing over, and captures the details the business needs to ring back. The enquiry arrives with the whole conversation attached, so whoever calls already knows what was asked and does not start cold.

Only if nobody sets the voice. The three assistants on the main page answer the same six questions from the same engine and sound nothing like each other, because each one is given its own tone, its own rules, and its own documents. The difference you are reading is the point of the demonstration.

It answers from your documents and nothing else. It is not allowed to fill gaps from general knowledge, so the usual way a chatbot embarrasses a business is closed off. If it cannot find the answer it says so and offers to pass on your details to a real person rather than guessing.

You also tell us what it must never discuss, and that list is yours to change.

In the unlikely event that it does get something wrong, tell us and we can fix it. The most common cause is two documents disagreeing with each other, which is worth knowing about anyway.

Yes. It answers from your documents, in your own folder, and you change them whenever you like; the assistant picks up the change overnight. The list of things it must never discuss is yours too.

Yes. It is one line in your website, and removing the line removes the assistant. The agreement itself ends with 30 days’ written notice.

Most of the time is yours: five short documents, from our templates. Once they are in, we build the assistant and send you a private link to try. When you say yes, it is one line on your site, and changes to your documents go live the next morning.

No. It is one line pasted into the site you already have, by you or whoever looks after your website, and nothing else on the site changes.

No, deliberately. It hands you the enquiry, and sends the customer to your own booking page — links to your booking and directions pages are included at every tier. It never takes a payment.

Only if nobody sets the voice. The three assistants on the main page answer the same six questions from the same engine and sound nothing like each other, because each one is given its own tone, its own rules, and its own documents.

Because there is a setup. Other tools hand you an empty box. We hand you a working one, and somebody watches the first week.

Most website chatbots are scripted. Somebody writes a list of questions and the replies to them, and the widget matches what your visitor typed against that list. Ask it anything the list does not cover and it apologises, or it offers you the list again.

Yours reads your documents instead. You write what is true about your business, in your own words, and it answers from that: your prices, your opening times, what you do about dogs, the thing you would never think to script. It also knows when the answer is not in there, and says so rather than guessing, which is the part a scripted widget cannot do at all.

The rest is the service around it. We build it, we watch the first week, every enquiry reaches you with the whole conversation attached, and every week you get a summary of what people asked.

You can, for yourself, and plenty of people do. But ChatGPT is a separate website and app. It is not embedded on your website, ready to answer your customers’ questions in your voice, from your business documents. NanoPorter does a different job. It sits on your website, so your customers can use it. It answers only from the documents you give it about your business and from your calendar, and when the answer is not in there, it says so. It takes the visitor’s details and emails you the whole conversation. And it is built, monitored and kept running for you, rather than being something you set up and look after yourself.

Yes, a plain-English one, and it runs month to month. You can end it at any time with 30 days’ written notice; if we ever ended it, we would give you 90. Before your assistant goes anywhere near your website you also sign a data processing agreement with us, and when you leave, your data is returned or deleted under it.

A full conversation, however many messages it takes — not single messages. Each tier’s figure is a fair-use guide to what it is designed for; if you regularly go over yours, we will talk about the tier that fits.

GDPR compliance belongs to your business, not to a piece of software, so we will not pretend to hand it to you. What we do is sign a data processing agreement before anything goes live, keep your data in the UK or EU, delete chat transcripts after 90 days, and make sure the AI never trains on it. The agreement covers the rest in plain English, including the parts that stay your responsibility.

Into a database in a UK or EU region, encrypted, with encrypted backups in the EU. To answer a question it is sent to the AI provider that writes the reply, which processes it in the United States under the standard legal safeguards, and does not train on it. When somebody leaves their details, an email comes to you with the conversation attached.

That is the whole list, and the full version with every company named comes with your agreement. If it ever changes you are told first.

You do not have to add a notice, because the assistant does not pretend to be a person. It is always called your business name, Virtual Assistant. It never gets a human first name, it never claims to be a member of staff, and if somebody asks whether they are talking to a human it says no.

There is a privacy link inside the chat window for anyone who wants the detail.