Overview
AI Memory is the control layer over what the AI is allowed to see when it answers anything in your organisation. Instead of one giant retrieval index over everything, you pin specific documents, registers and folders into named memories — "Cabling standards", "Client X reference projects", "Internal templates".
Each project chooses which memories are active. When the AI answers, retrieval is restricted to the active memories, which is what makes the answers feel grounded instead of generic.
It is the difference between asking a junior who read your standards last week and one who has heard of them.
When to use it
- You have organisation-wide standards or templates you want the AI to apply consistently.
- Different project types should draw on different reference material.
- You want hard control over what the AI is and is not allowed to consult.
How it works
1. Create a memory
Give it a name and a description. The description is what the AI sees as a label — make it specific.
2. Pin sources
Add documents from the Knowledge Base, registers, or whole folders. Pinned items are indexed for retrieval.
3. Activate per project
In the project settings, tick the memories that should be available. Inactive memories are invisible to the AI on that project.
4. Audit
Every AI answer shows which memory and which document a citation came from. Use this to tune the memory contents.
Tips
- Smaller, well-curated memories beat one big "everything" memory. Retrieval quality drops as the haystack grows.
- Re-curate quarterly — outdated reference docs poison answers in subtle ways.
- If an answer is wrong, check the citation first; nine times out of ten the memory is the fix, not the prompt.