What an AI assistant on your own content is
An AI assistant on your own content answers questions using only your material: website pages, FAQs, policies, procedures and past email replies. Each answer links to its source, and questions it can't answer go to a person. We build them for UK businesses, as an internal tool for staff or a chat window on the website.
Two kinds we build
For staff
A private assistant in the browser or in Teams. It answers from procedures, policies, product sheets and how the office has handled things before. New starters ask it before they ask a colleague.
For customers
A chat window on your website. It answers from your public pages and FAQs, takes the details when a question needs a person, and passes the conversation to your inbox or CRM.
What goes into a good one
The model is the easy part. Most of the work is in the content and the edges.
- Your content gathered and cleaned: old pages removed, duplicates merged, policies checked for the current version.
- Past email answers stripped of signatures, images and quoted chains, and personal details removed, before anything is indexed.
- A link to the source under every answer, so the reader can check it.
- A clear "I don't know" when the content doesn't cover it, followed by a handover to a person.
- A log of every question and answer, so you can see what people ask and where the content has gaps.
- Limits on what it will discuss, so a customer assistant stays on your products and services.
A better FAQ page might do
An assistant only helps if the answers already exist somewhere and people can't find them.
When an assistant is worth building
- The answers live across many documents, and staff ask each other instead of looking.
- Customers ask questions in their own words, which a search box doesn't match.
- The same questions reach the inbox every day, and the answers are already written down.
- You need the answer and its source logged for compliance.
Fix the FAQ page or the intranet search
If there are twenty common questions, a well-written FAQ page answers them faster and costs nothing to run. If staff can't find documents, a tidy SharePoint with search may be enough. We'll suggest that first.
Where it runs, and who sees what
An internal assistant can run privately inside your own cloud account, with access set by the same groups you use for files. Someone in sales won't get answers from HR documents they can't open. Our secure AI platform page covers how that is set up.
Customer assistants only see public content. We start with a Workflow Audit to count how much time the repeat questions take, and quote each phase at a fixed price.
Questions about AI assistants
Will the chatbot make things up?
It is built to avoid it. It answers only from your content, shows the source under each answer, and says it doesn't know when nothing matches. We test it against a set of real questions before launch and log every answer afterwards, so a bad one is found and fixed.
Where does our data go?
It can stay inside your own cloud account. The documents, the search index and the model can all run there, and nothing is used to train a public model. Customer-facing assistants only ever see content you have already made public.
Can it hand over to a real person?
Yes. When it can't answer, or the person asks, it collects their details and the conversation so far and passes them to your inbox, CRM or a staff member in Teams.
How does it stay up to date?
It reads from your content, so updating a page or policy updates the answers. We set it to re-read your sources on a schedule, and the question log shows which topics need new content.
Tell us what people keep asking
On a free call, describe the questions and where the answers live now. If a tidier FAQ page would do the job, we'll tell you.