Every case has an AI assistant, reached through the Assistant section (and available as a side panel while viewing documents/files). It knows the case: its files, drafted documents, notes, and record.
Threads (multiple conversations per case)
Each conversation is a thread. A case can have many. Open the "New chat" menu in the assistant header to:
- Start a New chat.
- Pick from your recent chats (most recent first).
- Toggle to see archived chats.
- View all chats if there are many.
Threads persist across reloads, tabs, and devices. Archive a thread from its hover menu; unarchive the same way. Document-scoped chats (started while viewing a specific document) stay with that document rather than the case as a whole.
The composer
At the bottom is the message box. Type and press Enter to send (Shift+Enter for a new line). The placeholder adapts to context ("Ask anything about this case" on the case home, "Ask or edit this document" inside a document).
Attaching files for context (@): press @ or click the + button to open a file
picker ("Search files & documents…"). Pick files and/or drafted documents; each appears as a
removable chip above the composer, and the assistant can use them directly.
Dictation (voice to text): click the microphone button (shown when the browser supports it), speak, and your words are transcribed into the composer. Click again to stop. (Requires a supported browser; if the mic button is missing, the browser doesn't support dictation.)
Response modes & context controls
Under the composer, on the right:
Mode dropdown — pick how the assistant balances speed against depth:
- Quick (the default) — fast responses for everyday questions and case lookups.
- Advanced — a more capable model that thinks longer before answering; best for motion drafting and deeper analysis.
- Custom — for power users: choose any available model from a searchable list, plus a Thinking slider (Off / Low / Medium / High / Max, from "Faster" to "Smarter") that controls how long it reasons before answering. While in Custom, the dropdown shows the chosen model's name instead of a mode name.
The choice is remembered per user and applies to your future messages.
Context usage circle — shows how full the conversation's context window is. Click it for a "Context usage" readout (tokens used vs. the model's window). As the window fills, older messages are automatically summarized to make room, so the chat never has to start over.
A reminder line reads: "Albatross can make mistakes. Always have a licensed attorney review its work."
What the assistant can do
- Search the case — search uploaded files and drafted documents; results appear as cards with file names and matching excerpts you can click to open.
- Draft & edit documents — create or revise a document, email, or motion. A line like "Creating the document" → "Created the document" appears and is clickable to open the result in the editor.
- Propose case updates (approval-gated) — the assistant can't silently change the case
record. When it wants to change a field (court, judge, charges, …) or save a fact to case
memory, it shows an approval card:
- Header "Proposed case update" (or "New case memory") with the before → after change.
- "Approve & save" applies it; "Reject" declines. The card then shows the outcome ("Case record updated" / "Saved to case memory" / "Change rejected").
Reading the conversation
Your messages appear on the right; the assistant's on the left. While it works, a status line shows a rotating verb and an elapsed timer, and the send button becomes a Stop button to interrupt. Tool actions ("Searched the case files", "3 results found") appear inline and can be expanded for detail.
Editing a sent message: hover your own message → "Edit". Editing rewrites from that point — everything after it is replaced by the assistant's new response (there's no branching).
Empty state & suggestions
A brand-new chat shows suggested prompts like "Summarize this case", "Find relevant case law", and "Draft a motion to dismiss." Click one to send it.
If there's no active subscription
If the firm's plan is inactive, the assistant shows a "Subscribe to use the assistant" card instead of answering, with a link to start a plan (see Section 23).