Nylivo / Domains / Real Estate

Real Estate

The domain where the source of truth is a document. A lease says what the income is; a rent roll says who's paying it. Get a term wrong and you haven't made a reporting error — you've mispriced an asset or broken a covenant.

Exploring Same Brain Document-grounded Cited or excluded
NOTE

Nothing here is built yet. This is the earliest of the three domains — a considered direction, not a product, and not a commitment to a date. What's real today is the substrate it would run on: the graph Brain, the document connectors, the approval rails, and the outcome ledger, all proven in production by QuantVantage.today. Published here because a platform claim you can't inspect isn't worth much.

Why this domain

The substrate was built for exactly this shape of problem.

Real estate isn't a stretch from finance and retail — it's the case where three of the platform's existing guarantees stop being nice-to-haves.

Fit 01

The truth lives in documents

Leases, estoppels, loan agreements, and inspection reports are the data. Extracting a term without a citation back to the clause it came from is unusable — which is the one thing a provenance-gated Brain refuses to do.

Fit 02

Exposure is a traversal

One tenant across nine properties, or six assets in one submarket, is invisible on any single deal sheet. Portfolio-level concentration is the same multi-hop question look-through risk already answers in the fintech workspace.

Fit 03

Decisions are rare and large

You underwrite an asset once and live with it for a decade. Slow, high-stakes, illiquid decisions are precisely where confirm-first approval and honest outcome grading earn their cost.

Workspace anatomy

Ingest. Graph. Underwrite. Grade.

The same four stages as the fintech and retail workspaces. Only the third one changes — here it's underwriting rather than advising or deciding.

Ingest

Documents first, spreadsheets second

Most real-estate data arrives as a PDF someone typed into a model by hand. The connector layer already reads document repositories and object storage, so the lease itself becomes the source — and the hand-typed number becomes checkable.

  • Leases & amendmentsTerms, escalations, options, and expiries — extracted with a pointer back to the clause, so any figure can be opened and read.
  • Rent rollUnit-level occupancy, in-place rent, concessions, and arrears, with the as-of date preserved.
  • Operating statementsHistorical income and expense by line item — the basis for a defensible NOI rather than an assumed one.
  • Debt termsRate, amortization, maturity, and covenants. A maturity wall is knowable years ahead; it should never be a surprise.
  • Market & compsTransactions, submarket supply pipeline, and absorption — context the asset is priced against.
  • Physical conditionInspection findings and deferred maintenance, because capex is the difference between a yield and a story.
Graph

One tenant, nine buildings

The risk that hurts a portfolio is almost never on the asset you're looking at. It's the tenant you didn't realize you'd underwritten nine times, or the submarket you're six assets deep in. That question is a traversal, not a report.

  • EntitiesProperty, Unit, Lease, Tenant, Owner, Portfolio, Submarket, Comparable, LoanFacility, CapexItem.
  • RelationshipsProperty has Unit · Unit leased to Tenant via Lease · Property in Submarket · Property comparable to Property · LoanFacility secured by Property · Portfolio holds Property.
  • Document provenanceEvery extracted term keeps its source document, page, and clause. A term that can't be cited is flagged for review, never quietly assumed.
  • Semantic layerEmbeddings over lease language, so “which of our leases have unusual co-tenancy clauses?” retrieves the right documents before anything reasons over them.
Underwrite

A defensible range, not a single number

A valuation printed to two decimals is a false promise. The output is a range with its assumptions exposed — each one traceable, each one arguable — plus the exposures that would change the answer.

  • Income & coverageNOI built from the actual rent roll and operating history, with DSCR and debt-yield against real loan terms.
  • Rollover exposureLease expiry laddering and renewal probability — where the income cliff is, and how far away.
  • Tenant concentrationCredit and revenue dependence at asset and portfolio level, with look-through to the shared tenant.
  • Valuation rangeComp-derived bounds with each comparable cited and adjustments itemized, not folded into one figure.
  • Capex & supplyDeferred maintenance drag and submarket pipeline pressure on future rent.
  • Refinance riskMaturity-wall stress at forward rate scenarios — solvable years early, painful only if unseen.
Grade

Slow outcomes still have to be scored

A ten-year hold can't wait a decade for feedback, so grading runs against observable interim outcomes — did the rent print, did the tenant renew, did the expense line land — while the terminal result resolves on its own schedule.

  • Interim barriersProjected versus realized rent, renewal versus vacancy, expense variance — each scored at its own horizon.
  • Terminal outcomeExit or refinance versus the underwritten range, resolved when it actually happens.
  • Closed-only ratesLive deals never count toward accuracy. An in-flight underwrite is not a track record.
  • Assumption attributionWhen an underwrite misses, the ledger shows which assumption was wrong — so the next model corrects the cause, not the symptom.
Direction, not roadmap

The three products this becomes.

Sequenced the same way as fintech — establish trustworthy ground truth first, then act on it. No dates, because there aren't any yet.

Lease & Document Intelligence
Cited extraction across a document set — terms, options, expiries, unusual clauses — answering portfolio-wide questions with a link to the clause. The ground-truth layer everything else depends on.
Property & Portfolio Intelligence
Continuous read on income, rollover, concentration, and submarket pressure across a whole portfolio — the standing view rather than a one-off model.
Deal Underwriting Copilot
Screening and underwriting support that produces a cited range and its exposures. It proposes; an investment committee decides — the same confirm-first control as every other workspace.

Adjacent from here: insurance (documents, exposure, slow outcomes) and construction & capital projects — the same shape again.

Early enough to shape it.

This domain is at the direction stage, which is the useful moment to talk. If you underwrite, own, or lend against real assets and this reads either right or wrong to you, I'd like to know which.