Private AI for your documents
Every department’s files and systems, one chat, nothing sent out.
Short answer. A chat assistant that answers staff questions from your own documents and business systems, using an AI model that runs inside your infrastructure. Each person only gets answers from the files and records their role is allowed to see.
What is included
Built for finance teams, law firms and any company that cannot send client or internal data to an outside AI service.
Runs on your infrastructure
Open-weight models on your own servers or private cloud. No outside AI service sees a question or a page.
Thousands of documents, every department
PDFs, Word, Excel, email and scanned files from shared drives, SharePoint or your document system, kept in sync.
Connected to your other systems
Live answers from your CRM, accounting, practice management or ticketing systems, through MCP servers, custom code connectors and APIs.
Role-based access
Permissions follow the groups you already use, in documents and connected systems alike, so HR stays with HR and matter files stay with the matter team.
Answers with sources
Every answer links to the document and page it came from, and says so when nothing matches.
One chat for every employee
A web chat behind your single sign-on, with Microsoft Teams or Slack if you want it there.
An audit trail
Who asked what, which documents were used and when, logged for your compliance team.
How it runs
Map
Departments, documents, systems, roles and what must never be shared.
Pilot
One department, its files and one system, accuracy measured.
Roll out
More departments and connectors, permissions tested role by role.
Run
Sync, connector upkeep, model updates and monthly accuracy checks.
Questions we get
What is a private AI knowledge base?
A chat assistant that answers staff questions from your own documents and business systems, using an AI model that runs inside your infrastructure. Each person only gets answers from the files and records their role is allowed to see.
Does any of our data leave our network?
No. The model, the search index and the documents all run inside your infrastructure, and the system makes no calls to outside AI services. We document every network connection so your IT team can check it.
How does it stop people seeing files they should not?
Access is checked before the model sees anything. Search only returns passages from documents the person’s role can open, using the same groups as your file shares and sign-on.
Can it use data from our other systems?
Yes. We connect it to the systems your teams already use, such as your CRM, accounting, practice management or ticketing tools, through MCP servers, custom code connectors or their APIs. Each connector runs inside your network and only reads what the person asking is allowed to see.
What hardware does it need?
That depends on the number of users and the size of the model. We size it in the mapping step. It can run on a GPU server you own or in a private cloud account hosted in Canada.
How accurate are the answers?
Before rollout we test it against real questions from each department. Every answer shows its sources, and accuracy is checked monthly.
Related services
Often scoped together with this one.
AI assistants and chatbots
Answers for customers and staff, from your documents.
APIs and integrations
Your CRM, payments and calendars, connected.
Cloud hosting and migration
A move to modern hosting without downtime.
Tell us what you’re building.
- A call about the goalThe budget, the deadline and what success looks like.
- A written scopeWhat we build, in what order, and the quote.
- Work in sprintsA demo at the end of each one.