Knowledge assistant

Ten thousand internal documents, and the answer to a routine question is spread across six of them. Staff ask an AI assistant in plain language and get a written answer with its sources attached. The whole system runs inside your perimeter.

How it works

You giveYour documents, in whatever shape they are in today: onboarding manuals, process guidelines, product specifications, compliance policies, technical runbooks, and the ticket and chat archives around them.

The system doesThe reading. It parses, chunks, and embeds the whole corpus into a private index, works out which passages actually bear on the question, and writes a direct answer from them.

You stay in controlOver what the assistant can see and say. Access follows the permissions your staff already have, every answer carries the documents behind it so anyone can check the reasoning, and new documents enter the index without a curation step.

You getA plain-language answer with its sources attached, in seconds, from a system running entirely on infrastructure you control.

Inside the system

06 components

Corpus ingestion

Every document is parsed, chunked, and embedded into a private vector index: onboarding manuals, process guidelines, product specifications, compliance policies, technical runbooks.

Closed deployment

The whole system runs on your infrastructure. No document, no question, and no answer reaches an external service, which is what makes it usable in places where external AI tools are prohibited outright.

Plain-language queries

People ask the way they would ask a colleague. No query syntax, no folder navigation, and no need to know which department owns the document that holds the answer.

Grounded answer synthesis

The assistant retrieves the passages that bear on the question and composes a direct, structured answer from them, instead of returning ten results and leaving the reading to whoever asked.

Source transparency

Every answer lists the documents behind it. Someone who needs certainty opens the source; someone who needs the answer takes the answer and moves on.

Incremental indexing

New and changed documents enter the index automatically, so the knowledge base tracks what the company actually does now.

Who it's for

Good fit

Companies where external AI tools are prohibited by policy

Support teams answering from internal documentation

Onboarding-heavy organisations with regular intakes

Regulated industries with compliance and audit documentation

Engineering teams with runbooks and specifications

Any company with a decade of accumulated process documents

Not a fit

Small document sets where people already know where things are

What you can configure

What the assistant can seeSources go in by collection, and access follows the permissions your staff already have, so nobody reads through the assistant what they could not open directly.

How much it may inferAnswer strictly from the documents, or reason across them. You set where that line sits, per collection.

Where people askA web app, your intranet, a chat client, or inside the tool your team already spends the day in.

Where it runsYour own hardware, your private cloud, or a managed deployment, depending on what your policy allows.

What we need from you

Your documentsThe corpus as it exists today, in whatever formats it accumulated in.

A place to run itOne machine or a private cloud tenancy. The system is built to stay inside your perimeter.

Access rulesWho is allowed to see what, so the assistant inherits your permissions instead of inventing its own.

One reviewerSomeone who can say whether an answer is correct and complete, a few hours a week during the build.

Example use cases

Answering staff questions from 10,000 internal documents

Staff at an online education company lost up to 15% of their day searching 10,000 internal documents, with external AI tools banned. A closed AI assistant on the company's own servers now answers in plain language with sources, and support tickets fell 40%.

Often deployed with

Live

ChatLogix

Turn community chatter into a searchable knowledge base.

ReadsDaily community messages, threads, and questions.

ReturnsA two-minute digest and a self-growing knowledge base.

2 minto catch up on a day

See the solution →

Live

In-editor document assistant

An AI assistant inside Word, on models you host yourself.

ReadsA long document open in Word, on your own servers.

ReturnsSummaries, checks, and tracked-change edits in the side panel.

~15 minto check a document

See the solution →

Your team already wrote the answer down. Let people find it.

Book a demo →