Oversight without operations.
You don’t run the buildings — you need to know which ones need a conversation. Insights gives you the record your operators work from, read-only: the whole book ranked by health, every Scorecard with its evidence, finished reports and the discussion around them. And if your team works in Claude, ChatGPT, Gemini or Copilot, they can ask the same record directly.
You backed your operator. You shouldn’t have to wait a quarter to know.
Quarterly decks are curated and stale. The raw plan-to-actual is too dense to read every period. When an asset starts to drift, the signals are already in the data — they’re just easy to miss in a 200-page package.
Insights is the layer in between. Figures are locked at close, commentary is written and approved by the people who run the asset, and the Scorecard is frozen when the report is published. You read that record as it stands — and Insights AI reads it with you.
Four things, every period. None of them curated.
The at-risk book, ranked
Every asset in your access, ranked by health — the same ranking your own AI gets over MCP. Home flags tier drops, your Watchlist and the Top Red Flags, so the first asset you open is the one that moved.
Scorecards with evidence
A 0–100 score across Financial, Leasing, Arrears, Capital, Debt and Valuation, a tier, and each red flag with the figure that tripped it. Lenses for Surveillance, Financial, Occupancy and Arrears.
Finished reports
The published report is the period’s locked record — commentary approved, sums checked, Scorecard frozen at publish. The number you discuss in March is the number that was on record in September.
Discussions, not meetings
Ask a question on the asset and @-mention the person who can answer it. The thread stays with the asset, so the next person to look sees what was asked, by whom, and what came back.
Caught a quarter later, the options narrow.
When a stressed asset is caught in the period it starts to drift, the room to act is wide: open the lender conversation before the covenant trips, renegotiate before the tenant’s lawyers are involved, plan capital without a forced sale. Caught a quarter later, the options narrow. Caught at year-end, you’re writing it down.
- DSCR below target and LTV near threshold — coverage and loan-to-value against the limits that matter.
- Maturity wall and Refi shock — the next maturity, tested against the submarket refinancing rate.
- Arrears escalating, Valuation declining — about 21 named red flags in all, each with its evidence.
You don’t have an asset manager on every property.
You have governance, analysts and a bench of operators you trust. What you don’t have is the headcount to read every operator’s plan against actuals, every period, across hundreds of assets. That’s the job the ranking does.
Insights scores every asset on a model tuned for its type — office, retail, industrial, multifamily and more — and ranks the book each period. Your analysts start with a short list, not a stack.
Your assistant. The same record.
Insights runs an MCP server — 50 tools over the same data your operators report from. Connect Claude, ChatGPT, Gemini, Microsoft Copilot, Cursor or any MCP client, and your analysts can ask the book a question in the assistant they already use.
- Your access, exactly — an agent sees what you can see. A read-only seat stays read-only, whichever AI you connect.
- No deletes, ever — and any agent writing stays off until an administrator switches it on.
- OAuth 2.1 with PKCE — 30-minute access tokens and rotating refresh tokens.
rank_asset_healthassess_asset_healthget_portfolio_healthget_report_commentaryget_debtcompare_periodslist_reportsRead-only by design. Hosted in the customer’s cloud.
Read-only by design
The Investor seat observes, questions and discusses. It doesn’t edit the operator’s plan, the published figures or the commentary. The operator stays in charge of their report; you stay in the loop.
Scoped access
Access is set at portfolio and asset level, with Deny for exceptions and section access on top. An investor scoped to one portfolio sees that portfolio and nothing else.
One database per environment
Each environment’s data lives in its own database, hosted in the customer’s own cloud — Azure, AWS or on-premises. Encryption at rest follows that database’s settings; data is encrypted in transit.
A trail that only grows
Sign-in through Microsoft Entra ID, and an append-only audit log kept for 7 years — with the IP address and user agent behind every entry.
Questions investment teams ask us.
Where do the numbers come from?
From the same record your operator reports from: figures from their own systems, locked module by module at close on the RealTime™ engine, with commentary written and approved in Insights. Nothing is re-keyed for you.
Will the AI ever make a decision on an asset?
No. Insights AI ranks, explains and cites; people decide. An agent can never delete anything, and agent writing stays off until an administrator turns it on.
Can our analysts use their own AI?
Yes. Connect Claude, ChatGPT, Gemini, Microsoft Copilot, Cursor or any MCP client. Each agent gets exactly the access of the person who connected it — nothing more.
Can I see why an asset was flagged?
Yes. “How this score is calculated” shows every threshold beside the asset’s own figure, and each red flag carries its evidence. A published report’s score is frozen, so it doesn’t shift after the fact.
Can I ask the operator a question without a meeting?
Yes. Start a discussion on the asset and @-mention the person who can answer. It stays with the asset for the next person who looks.
Where is the data hosted?
Each environment has its own database in the customer’s own cloud — Azure, AWS or on-premises. The Insights application runs on Microsoft Azure. The security page has the detail.
Where the read comes from.
See your book the way the ranking does.
A 30-minute walkthrough on a sample portfolio: the ranked book, a Scorecard with its evidence, a finished report — and the same question asked from your own AI.