Every property workspace has an AI chat panel pinned to the right side. It's an analyst grounded in the property's data and uploaded documents. You ask questions in plain language; it answers from that data and names the documents it drew on.
What it can answer
A non-exhaustive list:
- Income & rents — "what's the in-place vacancy?", "what's the median 1-bed rent?", "show me the lease expiration distribution."
- Expenses — "what was payroll in March in the T-12?", "is the management fee in the OM consistent with our model?"
- Comps — "summarize the three most relevant comps", "what's the cap-rate range across the comps?"
- Market — "what's the supply pipeline within 1 mile?", "is this submarket growing?"
- Risk & opportunity — "what are the three biggest risks?", "what value-add levers does the AI Summary identify?"
- Document-grounded — "what does the OM say about parking ratio?", "find the section in the lease about pet fees."
You can also attach a document to your message: click the paperclip or drag the file into the chat box, and Relm uploads and indexes it before answering.
What it won't do
- Won't fabricate. If a fact isn't grounded in the property's data, your documents, or the web research it runs, the chat says so rather than inventing one.
- Won't editorialize. It won't recommend "buy" or "pass" — that's your call. It can summarize the facts that bear on the decision.
- Won't reach beyond this property. Asking it about an unrelated property in your portfolio won't work in this scope. It can, however, search the web for market context and neighborhood news around this property.
For broader, multi-property analysis, the chat panel scope can be expanded in the future — today, it's per-property.
Research and Edit modes
The switch under the chat box sets the mode:
- Research — read-only and private to you. This is the default on the property overview.
- Edit — the chat can change the property's data, and the conversation is shared with your team. This is the default on section pages. Most changes come back as a proposal card you review and apply; a few actions (such as re-scraping unit data) run right away.
Per-section context
The chat is contextually scoped:
- On the overview, it answers about the property as a whole.
- On a section subpage (e.g. Financial), it's tuned to that section first, with full property context as fallback. If another section covers your question better, it may add a card such as Go to Comps that takes you there.
- Each section keeps its own conversation history, so flipping between sections doesn't clutter unrelated threads.
Citations
When an answer draws on an uploaded document, it names the file (for example, Acme_OM_2026.pdf), and a chip for that document appears under the answer. Click the chip to open the document.
See Citations in chat.
When chat is disabled
Chat is locked, or holds your messages, in a few states:
- Before the first Intel Report. On a property that has never been researched, chat is locked ("Run Intel Report to unlock chat") until an Intel Report finishes.
- During an active research run. Chat stays open, but your messages are queued and answered once research finishes. On a section page whose research has already landed, chat answers right away. While a pro-forma is generating, messages on the Financial page are queued the same way.
- While a teammate is editing. In Edit mode, chat is read-only while someone else has the property open for editing.
- On the demo property. Chat is off.
Conversation history
Each property's chat history is preserved. Click the back arrow at the top of the chat panel to see your conversations — the current section's are listed first, then Other Sections — and open any thread to pick it up again. Research chats are private to you; Edit chats are shared with everyone in your organization.
Privacy
Chat conversations are scoped to your organization, encrypted at rest, and never used to train models. See AI in Relm for the full data-handling rundown.