The useful knowledge inside a company rarely lives in one knowledge base. It is spread across tasks, Docs, meeting notes, database rows, support tickets, email threads, forms, files, and the people who work with them.
That creates an awkward gap for AI. A model may be able to write a fluent answer, but it still needs to know where the team's current information lives, which projects the person asking can access, and which source supports each claim.
Company Brain is Aamu.app's answer to that problem. It gives Launch AI and @ai a permission-aware way to retrieve information from the Aamu workspace, answer a question, and link back to the objects where the answer was found.
From a support knowledge base to workspace knowledge
Aamu already has Team Brain for maintained Helpdesk knowledge. Team Brain works well when a team wants to choose canonical support sources such as Docs, FAQ snippets, web pages, sitemaps, or selected resolved tickets. It answers a focused question: what approved knowledge should support use?
Company Brain has a broader job. It makes the operational information already stored in Aamu retrievable for internal questions. Instead of requiring every useful fact to be copied into a separate knowledge source, Company Brain indexes the top-level workspace objects that the team uses every day.
The two layers complement each other:
Team Brain contains selected, maintained knowledge for workflows such as Helpdesk answers and customer-facing drafts.
Company Brain helps people ask broader questions across the work taking place in Aamu.
What Company Brain can find
The first Company Brain version indexes the main Aamu workspace objects:
projects and team members,
tasks and their discussion context,
Docs and Doc content,
meetings, AI notes, and speaker-attributed transcripts,
forms and form structure,
file metadata,
databases, tables, and rows,
Helpdesk tickets,
the signed-in user's own email, and
newsletters and newsletter issues.
Git repositories and Penpot designs are deliberately outside the first version. Their data models need more specialized indexing than normal workspace objects. File contents also need their own handling: the initial version indexes file metadata, not arbitrary binary files stored in object storage.
Ask from Launch AI or mention @ai
Company Brain is not a separate chatbot. It is part of the same AI path people already use in Aamu.
A person can open the Launch AI dialog and ask a broad workspace question:
What did we decide about the onboarding change, and is there already a task for it?
The same retrieval is available when someone mentions @ai in a task, Doc, meeting, ticket, or another supported discussion. The current item and discussion provide immediate context, while Company Brain can retrieve related information from elsewhere in the accessible workspace.
This shared path matters. Launch AI and @ai should not behave like two assistants with different memories. They should use the same permissions, the same indexed sources, and the same grounding rules.
Answers include links back to Aamu
A useful company answer should not end with “the AI says so.” Company Brain results keep the source type, title, project, and Aamu URL alongside the indexed text.
When AI uses those results for a factual answer, it is instructed to cite the supporting sources. Launch AI stores those source references with the conversation and renders them as clickable links.
That makes an answer a starting point for work rather than a dead end. A teammate can open the meeting where a decision was made, inspect the task that tracks it, read the original Doc, or check the relevant database row.
Permissions are part of retrieval
Indexing company data is only useful if retrieval respects the same boundaries as the workspace.
Every Company Brain document carries its team, project, source type, and visibility metadata. Before a query reaches the index, Aamu resolves the projects available to the signed-in user. The retrieval query then includes only:
company-level information for the current team,
project information from projects the user belongs to, and
user-specific information owned by that user.
Email is the clearest example. Email content can be useful context, but one team member's mailbox must not become another team member's AI search result. Company Brain therefore marks email chunks as user-owned and requires both the owner and project checks during retrieval.
The access check happens before source text is added to the model context. The model is not asked to decide which information the user should be allowed to see.
Meetings become searchable knowledge
Live meetings are an important Company Brain source because decisions often appear in conversation before they reach a Doc or task.
When the AI meeting assistant is enabled, it can join the LiveKit room as a participant, transcribe each participant's audio with speaker attribution, and produce meeting notes. Company Brain can index the meeting title, description, transcript, summary, decisions, and action items.
A later question can then connect the conversation to the rest of the workspace:
Who agreed to contact the customer after Tuesday's meeting?
The answer can point back to the meeting page instead of treating the transcript as anonymous text.
Incremental indexing keeps the cost manageable
Company Brain does not need to recreate every embedding on every run. Each indexed document has a content hash. Unchanged documents are reused, changed documents are split into fresh chunks, and removed objects are deleted from the index after a successful run.
An admin starts the first Company Brain indexing run from the team's AI settings. Once enabled, Aamu checks the index periodically and refreshes only the material that changed.
What Company Brain changes
The practical shift is small but important. People no longer need to remember whether the answer was hidden in a task comment, meeting transcript, Doc, ticket, or database row before they can start searching.
They can ask the question in the language of the work:
What did we promise this customer?
Which tasks mention the migration deadline?
What decisions came out of the latest product meeting?
Where is the current onboarding process documented?
Do we already have a database row for this company?
Company Brain does not replace navigation, search, or maintained documentation. It connects them. AI can help find and explain the relevant information, while the linked Aamu objects remain the source of truth.