From website to answers, in three steps
The first step is the import: you enter the site address and 4chaty crawls the public pages, extracts the useful text and splits it into coherent fragments. The second step is completion: you upload the documents that are not published — the rate card, the contract terms, the returns procedure — and write the FAQs you get most often. The third step is verification: you ask the bot the real questions your team fields every day and correct whatever it gets wrong by adding an explicit FAQ. After that short cycle, coverage usually reaches the point where most conversations close without an agent.
Why answers cite their source
An answer without a source cannot be verified, neither by the customer nor by you. 4chaty attaches to each answer the fragment and page the information was drawn from. For the visitor, that means a link to the official page with the full detail. For you, it means you can immediately diagnose why the bot answered wrongly: almost always the problem is not the model but an old page left in the index, carrying a price or a deadline that no longer applies. Delete or exclude the page and the answer corrects itself, with no retraining.
Knowledge gaps: what the bot does not know yet
4chaty records the questions for which it found no relevant fragment and groups them by theme. That list is the most valuable report in the product: it shows exactly what customers are looking for and not finding on your site. Often the gaps are not solved in chat alone — they point to pages missing from the site entirely: a delivery page, a warranty page, cancellation terms. You add the FAQ and the bot answers from the very next conversation.
What belongs in the knowledge base and what does not
The knowledge base is for stable information: services, rates, procedures, terms, opening hours. It is not the place for data that changes minute by minute, such as exact stock or the status of an individual order — those are handled by escalating to an agent. Never load customers’ personal data into the knowledge base: information there can surface in answers to other visitors, which would be an unauthorised disclosure under the GDPR.
