Every figure cited. Every decision yours.
Insights AI reads every asset’s period down to the GL line — drafting the commentary, ranking the book by health and suggesting the next action, with a source behind every number it writes. Nothing is accepted, sent or decided without a person. And through the Insights MCP server, the AI your team already uses — Claude, ChatGPT, Copilot — can work from the same record, with exactly your access.
The analyst work, done in the period it matters.
Insights AI works inside the modules, on the period’s locked figures. It suggests; your people decide.
Ask Insights
Plain-English questions about any asset or portfolio, answered down to the GL line, framed for the person asking — and a link straight to the screen that shows it.
Drafts the commentary
During Run close, through Insights AI Assist, or with Redraft commentary — in Standard, Deep or Hybrid mode, in the voice your firm chooses.
Cites every figure
Every number it quotes links to its source — “Show where these figures came from” — and is checked against the record before you can accept the draft.
Ranks the book
Every asset ranked by Scorecard health, with the red flags behind each move. Home shows the tier drops, the Watchlist and the top red flags.
Suggests the action
Reads the period — vacancies, expiries, projects, arrears, flags — and filed documents, then proposes actions and commentary. Accept, edit or dismiss.
Writes the Home Brief
Each morning’s brief: what moved, what’s due, what needs you — before the first meeting of the day.
Trust the draft because you can check it.
A draft is only useful if you can sign it. Every figure Insights AI writes is tied to the locked number it came from — the GL line, the lease, the debt record — and checked before Accept is available. If the figures change after a refresh, the section says so.
- Sources shown inline and in “Show where these figures came from”.
- Drafts work on the period’s locked figures, never a guess.
- Accepted text is marked as written by Insights AI, on the record.
It writes the way your reports read.
Administrators choose the voice Insights AI writes in, so the same facts read right for the relationship: a manager reporting to an owner, an owner reviewing its manager, or an owner-operator writing for itself. Try it.
- Personas frame every answer — an investor and a property manager get the same facts, told for their job.
- Runs on Anthropic’s Claude models; an environment can bring its own Anthropic key.
We ended September with occupancy at 82.0%. Meridian Grocers has been dark since May and has signalled its intent to surrender Suites 100–400. We hold a backfill LOI at 8% above the current face rate and will bring terms to the October review.
The manager reports occupancy of 82.0%, with Meridian Grocers dark since May and signalling surrender of Suites 100–400. The backfill LOI at 8% above face is not yet signed; we have asked for terms and a lender update before the October review.
Occupancy is 82.0%. Meridian Grocers has been dark since May and wants to surrender Suites 100–400; we are negotiating terms with a backfill LOI in hand at 8% above face. Terms go to the October review.
Bring your own AI. Keep your own rules.
Your team already has an assistant. Connect it to Insights over the Model Context Protocol and it works from the same record your people do — 50 tools across every module, with exactly the signed-in person’s access. Claude, ChatGPT, Gemini, Microsoft Copilot, Cursor and other MCP clients connect the same way.
- Exactly your access. An agent sees the sections, modules and processes its user can — nothing more.
- Writes off by default. Seven write tools exist; an administrator switches them on. There is no delete — an agent can only flag an action for a person to remove.
- Standard sign-in. OAuth 2.1 with PKCE and dynamic client registration, 30-minute access tokens, rotating refresh tokens.
- On the record. Every agent write lands in the append-only audit log.
Portfolio & health
rank_asset_healthassess_asset_healthget_portfolio_healthcompare_periodsget_portfolio_reviewlist_assetsFinancials & leasing
get_financial_summaryget_gl_variancesget_occupancy_varianceget_expiring_leasesget_new_leasesget_base_rentCapital, arrears, debt
get_capex_varianceget_capex_projectsget_arrears_flaggedget_debtget_appraisal_historyget_inspectionsWrites · admin-enabled
create_actionupdate_actionupdate_action_noteflag_action_for_deletionpropose_review_itemwrite_commentarywrite_commentary_batchHuman decisions, human authority — always.
The AI surfaces risk and explains its reasoning. People step in early, while there’s still room to act — restructure the debt before the covenant trips, renegotiate the lease before the lawyers are in the room.
Suggest, never decide
Drafts wait for Accept. Suggestions wait for a person. Nothing publishes itself.
Show the source
Every figure cited, every citation checked. If it can’t be traced, it doesn’t ship.
Your access, no more
Insights AI and every connected agent inherit the user’s permissions exactly.
Your key, if you want
Insights AI runs on Anthropic’s Claude. An environment can use its own Anthropic key.
Ask your portfolio a question.
A 30-minute walkthrough: Ask Insights on a sample portfolio, a cited commentary draft, the ranked book — and your own assistant connected over MCP.