Most AI assistants have a fundamental problem: they know nothing about your work. Every conversation starts from scratch, with no memory of what was decided last week, what the last quarter's numbers showed, or what was said at a meeting three weeks ago. To be genuinely useful, an assistant needs to actually know your organisation — and to know it, it must be able to remember it.
This project explores exactly that. A local AI platform that runs on your own hardware and gathers your information into a single searchable memory, so the assistant can answer real questions about your work.
"What did we agree on at the lunch meeting three weeks ago?"
"Analyse the sales of Nike Speedrunner shoes in Q3, after the Gothenburg race."
These are not generic AI responses. They are built on your actual material: your meetings, your documents, and your data.
All formats, one memory
The knowledge in a company is rarely neatly organised. It lives in meeting recordings, email threads, PowerPoint presentations, PDF files, spreadsheets, and handwritten notes. The platform is built to ingest information regardless of form — audio, text, presentations, or documents — and merge it into a unified memory the assistant can analyse. Nothing needs to be reformatted or manually entered in advance.
Connects to your tools
The integration pattern is designed to connect to the tools an organisation already uses — for example:
- Communication and meetings: Email, Slack, Teams, and Zoom, so that conversations and meeting recordings become part of the memory.
- Documents and storage: Google Drive, SharePoint, PDF files, and local folders.
On your terms
Crucially, all of this happens locally. Your data stays on your own computer. Integrations with external services are isolated and only read what you allow. This makes it possible to get the benefits of an AI that genuinely knows your organisation without handing over information to an external cloud — an increasingly important concern as data protection and governance requirements tighten.
What the project explores
The platform is composed of modules that together form a complete system:
- A memory architecture that lets the assistant remember things over time.
- Secure integrations with tools you use day to day.
- A proactive agent that surfaces what matters before you even ask.
- Security layers that protect against the risks that arise when an AI is given access to real business data.