There are two versions of this problem and they deserve different answers. Internal enterprise search is about employees finding what the organisation already knows. A website answer agent is about visitors getting a straight answer without hunting through your site. Both are retrieval problems; only one is public-facing, and the security models differ completely.
Internal Knowledge Search
Your organisation’s knowledge is spread across SharePoint, shared drives, ticketing systems, wikis, code repositories and email. People do not fail to find things because search is bad — they fail because the knowledge is in six systems with six search boxes.
- Natural-language search across systems — one question, many sources.
- Permission-aware retrieval — results respect what each person is already allowed to see. This is non-negotiable, and it is the part cheap implementations skip.
- Cited answers — every response points to the document behind it, so it can be verified.
- Honest gaps — saying “this is not covered” instead of inventing an answer.
- Unanswered-question reporting — a map of where your documentation is actually thin.
The retrieval architecture behind this is covered on our RAG and enterprise search page.
Website Answer and Response Agents
A public answer agent is not “a chatbot” in the 2017 sense. Done properly it does semantic search over your real content, recommends the relevant service, cites its sources, captures qualified enquiries into your CRM, offers scheduling, escalates what it cannot answer, and reports the questions it failed on.
It is also a public attack surface. Anything a visitor can type is untrusted input, so a public agent gets tighter tool scoping and stricter output handling than an internal one — the reasoning is set out in our sister practice’s article on prompt injection.
The Common Requirement
Both versions live or die on grounding. An answer that cannot be traced to a source is not usable in a business context, whether the reader is an employee or a prospect.
Tell us what people are failing to find — internally or on your site — and we will scope the right approach.
Related services
- RAG and Enterprise Search — The retrieval engineering behind it: RAG and enterprise search.