It can include one, but the point is the collection behind it. A question box that answers from approved documents and shows the source is useful; a chatbot that answers from anywhere is a liability. We build the first kind, and where a simple search over well-organized documents does the job, we recommend that instead.

Where does our information live?

In an account you own, with access rules you set. The scope names the storage and any provider involved, what each one can see and how it is switched off. Restricted material can be kept out of the system entirely or in a collection with its own access list.

How accurate are the answers?

As accurate as the collection and the checking. Before launch the system is run against real questions from your team and its answers compared with the documents; anything it cannot answer from the collection it must say so. Accuracy is re-checked whenever the material changes significantly.

What do we have to prepare?

A first collection and a list of the questions people actually ask. The review of the material is part of the project, so the documents do not have to be tidy first; they do have to have an owner who can say which version is current.

Can this connect to our other systems?

Yes, where those systems expose their data and where the access rules allow it. A common pattern is a knowledge system that answers procedure questions and, through workflow automation, files the request the procedure describes. Each connection is scoped and tested separately.

Find the answer.
Know where it came from.

Make approved documents and team guidance easier to search, with source references and appropriate access.

Discuss your project See all Services

03 · Knowledge systems

Example projects

Searchable team guidance

Bring a defined collection of policies, instructions, and useful documents into one navigable experience.

Source-backed answers

Help people ask useful questions and check the supporting material instead of relying on an unexplained answer.

Research briefs

Organize recurring research into a consistent brief, with visible sources and room for human judgment.

A useful first project

Start with one trusted collection.

Choose a manageable set of information and the questions your team needs it to answer.

What we agree before building

Which sources are current, approved, and in scope
Who can access each collection and who maintains it
How answers show their sources and handle uncertainty
A set of real questions to check usefulness and accuracy

Duplicate, outdated, or restricted material needs attention before it enters the system. Information ownership and review remain part of the project.

Who this is for

A team whose answers live in too many places.

Every business accumulates guidance: procedures, policies, pricing rules, supplier notes, the way a particular customer likes things done. It sits in shared drives, old emails and a few people's heads, and new staff learn it by asking. A knowledge system is for a team that wants one place to ask, with answers that show where they came from.

It is also for a business that has started using AI assistants and noticed the problem: a fluent answer with no source is not something you can act on. The systems we build keep the source in view, so a person can check the document behind the answer in one step.

What is included

One trusted collection first, with its owners named.

We start small on purpose, with one collection that matters and the questions it has to answer.

A review of the material: which documents are current and approved, which are duplicates or out of date, and which are restricted. Nothing enters the system until that is settled.
Access rules per collection: who can read it, who maintains it, and how a change is reviewed before it becomes the answer.
Search and question answering that shows its sources beside every answer and says plainly when the collection does not cover a question.
A set of real questions from the team, used to check usefulness and accuracy before launch and again after every significant change to the material.
A maintenance routine: how new documents are added, how old ones are retired, and who is responsible.

How pricing works

Scoped by collection, not by ambition.

A knowledge system is quoted for a defined collection and a defined set of questions, after the material review. That review is the first piece of work, because the state of the documents decides most of the effort: a well-kept folder of current procedures and a decade of unsorted files are different projects.

Any service the system depends on, such as hosting, a search index or a language-model provider, is named separately with its cost. Adding a second collection later is its own small scope; the access rules and the question set travel with it.

In practice

Systems that show their working.

The platform built for PC NET TECHS carries its own knowledge layer for the people who run it: a Build Notes page that maps every part of the system file by file, a command console whose commands are listed and explained rather than guessed, and a strategy view that quotes the business's published facts directly from the website's own content, so the two cannot drift apart. Our Social Media Engine works the same way: it writes only from a business's published facts, and a check fails the build if a price, an hour or a claim it cannot support appears.

The GCI CLI prototype on this site shows the interaction in miniature: a fixed, visible set of commands over a browser-local workspace, so you can judge for yourself how a system that explains itself feels to use.

Questions people ask

Before you write to us.

Is this a chatbot?: It can include one, but the point is the collection behind it. A question box that answers from approved documents and shows the source is useful; a chatbot that answers from anywhere is a liability. We build the first kind, and where a simple search over well-organized documents does the job, we recommend that instead.
Where does our information live?: In an account you own, with access rules you set. The scope names the storage and any provider involved, what each one can see and how it is switched off. Restricted material can be kept out of the system entirely or in a collection with its own access list.
How accurate are the answers?: As accurate as the collection and the checking. Before launch the system is run against real questions from your team and its answers compared with the documents; anything it cannot answer from the collection it must say so. Accuracy is re-checked whenever the material changes significantly.
What do we have to prepare?: A first collection and a list of the questions people actually ask. The review of the material is part of the project, so the documents do not have to be tidy first; they do have to have an owner who can say which version is current.
Can this connect to our other systems?: Yes, where those systems expose their data and where the access rules allow it. A common pattern is a knowledge system that answers procedure questions and, through workflow automation, files the request the procedure describes. Each connection is scoped and tested separately.

Start with your next idea

Ambition, meet
possibility.

Tell us what you want to create or improve. We’ll help you define a useful starting point.

Start a project