AI knowledge base
A secure assistant that knows how your company works.
Give employees an AI assistant that answers questions using approved company documents — with sources they can check, and without exposing anything they aren’t allowed to see.
Company assistant
Illustration
What do we need before a new subcontractor starts a job?
According to the subcontractor onboarding procedure, three things must be on file first: a signed agreement, a current certificate of insurance, and a W-9. The office manager then creates their vendor record.
Sources
Subcontractor Onboarding Procedure
Why it helps
Stop answering the same questions — and stop guessing at the answers.
In most small businesses, knowledge lives in a few people’s heads and a shared drive nobody can search. A knowledge assistant makes the documents you already have genuinely useful.
- New hires find answers without interrupting senior staff
- Field and front-line staff get the current procedure, not last year's
- Every answer links to the document it came from
- Gaps and outdated documents become obvious — and fixable
- Runs on business cloud AI or on private AI, depending on sensitivity
How it works
From approved documents to trustworthy answers.
Technically this is called retrieval-augmented generation (RAG): the assistant looks up relevant passages in your documents first, then writes an answer based on them.
- 01
Approved documents
Policies, procedures, manuals and product information you choose.
- 02
Searchable index
Documents are prepared for AI search, in the cloud workspace or on private AI.
- 03
Permission check
Each question only draws on documents that employee is allowed to open.
- 04
Answer with sources
A plain-English answer with links back to the documents it used.
What goes in
Start with the documents that answer the most questions.
We begin with a curated set, not your entire file server. Less clutter means better answers and fewer surprises.
Policies
Procedures
Manuals
Internal documentation
Product knowledge
Contracts, where appropriate
Permissions first
The assistant can't show people documents they can't already open.
This is the part many AI projects get wrong. We connect the knowledge base to your existing sign-in and groups, test access with real accounts before launch, and keep sensitive material out entirely where a platform can’t enforce permissions reliably.
Before launch we verify
- Staff only get answers from documents their role can access
- Restricted folders (HR, payroll, owner-only) are excluded or permission-checked
- Removed employees lose access with the rest of their accounts
- Sources are cited on every answer
- Where the data is processed — cloud or private — is documented
Questions
Knowledge base questions
Can AI search our company documents?
Yes. An AI knowledge base lets employees ask questions in plain English and get answers drawn from approved company documents — policies, procedures, manuals, product information — with links back to the source so they can check.
We start with a defined set of documents rather than "everything," clean up duplicates and outdated versions, and decide whether it runs on a business cloud AI service or on private AI depending on how sensitive the material is.
Can different employees have different access?
They should. A well-built knowledge assistant respects the same permissions as your files: if an employee can't open a document, the assistant shouldn't use it to answer their question. We design for this from the start using your existing sign-in groups, and we test it before launch. Where a platform can't enforce document-level permissions reliably, we keep sensitive material out of it.
What is private AI?
Private AI means running an AI model on hardware your business controls — typically a dedicated server in your office or a private environment you own — so that prompts and documents are processed there rather than sent to a public AI provider.
Local models are very capable for many tasks such as summarizing, drafting and answering questions from documents, though the largest cloud models are usually stronger on the hardest problems. We help you decide which work belongs where.
Can we prevent certain information from being sent to cloud AI?
You can reduce the risk significantly, but no tool catches everything. We combine several layers: a data classification your staff can actually remember (our green, yellow and red model), approved tools for each level, restricted connectors, browser and endpoint controls where your licensing supports them, and — for the most sensitive work — a private AI system that doesn't send data to a cloud provider at all.
We'll be clear about what each control does and doesn't do. We won't tell you a system can automatically detect every piece of confidential information, because that isn't realistic.
See what your documents could answer.
Tell us about your team and the questions they ask most. We'll recommend where to start, which documents to include and whether it belongs on cloud or private AI.