Continuous variance analysis,
not quarterly archaeology.
For the asset manager, the quarter is a dig: actuals land late, you reconcile by hand, and the variance is a quarter old by the time it is explained. Built AI computes variance continuously as actuals arrive, in plain English, traceable to source.
Reconciling actuals against plan against comp, by hand, every quarter
The first two weeks of every quarter disappear into reconciliation, walked by hand across every asset until the team can explain why NOI came in light.
The numbers themselves are not the hard part. The hard part is that they live in pieces. NOI is in the GL. The budget is in last year's model. Occupancy and the rent roll are in Yardi. The lease that explains why a tenant's recovery dropped is a PDF in a data room. The covenant the soft quarter threatens is a clause in a credit agreement nobody has opened since closing. To explain one variance properly you assemble all of those by hand, and across a portfolio you do it again and again until the explanation is stale before it is finished.
So asset management becomes reactive, the opposite of what the seat is for. You find out a tenant went dark when the rent stops, not when the news broke. You find out a covenant is tightening when the lender's certificate is due, not when the trend started bending three months earlier. The information was always in the building. It just was not assembled in time to act on.
The deeper issue is that the knowledge required to read a portfolio well does not scale by hiring. A great asset manager carries the whole position in their head, but that knowledge is trapped in one person's spreadsheets and rebuilt every period. The work is not analysis. It is reconstruction, performed over and over, before analysis can even begin.
Variance computed as it happens, explained in plain English, traceable to source
The reconciliation you ran by hand becomes a continuous computation. Continuous, explained, traceable. How the engine works →
Built AI binds the systems you already run, your accounting, your Argus and underwriting models, your leases and loan documents, into one normalized knowledge graph. Because everything is connected, the reconciliation you ran by hand becomes a continuous computation the moment new actuals arrive.
When the period refreshes, it builds the NOI bridge automatically and names each driver in order of magnitude, every figure linked back to the GL line, lease clause or model assumption behind it.
Crucially, the math is not the model's guess. Every figure the platform reports (NOI, the bridge, debt yield, the covenant tests, the re-leasing scenarios) runs on a deterministic calculation engine, the same way every time, with the work shown cell by cell. The language model reads the portfolio, decides what matters and writes the commentary. The engine does the arithmetic. That separation is what lets you hand a variance pack straight to an investor or a lender without re-deriving it, because the answer to "where did this come from" is a trace, not a story. The full architecture is described on the platform page.
Covenant Watch, before the test date
The agent that changes the seat the most is Covenant Watch. On every GL or Argus refresh, it re-runs the DSCR, debt-yield and LTV tests on every levered asset against the actual covenant language in the credit agreement, not a summary of it but the clause itself. When a test starts tightening, you hear about it while there is still room to act: a forecast that this asset trips its 1.20x DSCR covenant at the next test date if the soft quarter holds, with the cure options laid out and the math behind each one. The breach you used to discover when the lender's certificate came due is now a forecast you see a quarter early.
Lease intelligence works the same way. The platform reads every lease in the portfolio into structured terms (base rent, escalations, options, recoveries, expiries) and keeps a live view of rollover exposure. Which leases roll in the next eighteen months. Which sit above or below market, and what re-leasing at market does to NOI and to the waterfall. Which options are coming up and which tenants are concentration risk. The institutional knowledge that used to live in one person's head now lives in the graph, where the whole team can query it.
What the asset manager actually gets
Not a dashboard to maintain and not another tab to check. Four concrete changes to how the work happens, each one removing a category of manual effort that never created value in the first place.
The variance pack writes itself, then you review it
On every actuals refresh, the NOI bridge is built, the drivers are isolated and the narrative is drafted in your format, with each figure linked to source. You stop assembling the explanation and start reviewing it. The first eighty percent of the work is done before you open the period.
Covenant breaches surface before the test date, not after
Covenant Watch re-tests DSCR, debt yield and LTV on every refresh against the actual covenant language, forecasts which tests are tightening, and proposes cure options with the math behind each. You manage the covenant while there is still time to manage it.
Lease and rollover risk become a live view, not a quarterly rebuild
Every lease is read into structured terms and kept current. Rollover exposure, above- and below-market positions, option windows and tenant concentration are queryable across the portfolio at any time, with re-leasing scenarios run on the same engine your firm already trusts.
The portfolio answers questions in minutes, across every source
Ask which assets breach a covenant if rates move fifty basis points, or which leases above market roll inside eighteen months, and the answer is a single traversal of the graph, not a two-day reconciliation across five systems. The standing context is the graph, so there is nothing to brief.
| The work | Today | With Built AI |
|---|---|---|
| Variance analysis | Built by hand, two weeks after close, stale on arrival | Continuous, computed as actuals land, narrative drafted |
| The NOI bridge | Re-keyed into a spreadsheet each period | Built automatically, every driver traced to source |
| Covenant testing | Discovered when the lender's certificate is due | Forecast before the test date with cure options |
| Lease rollover | Reconstructed from PDFs when someone asks | A live view across the portfolio, always current |
| Portfolio questions | Two days reconciling five systems by hand | Answered in minutes by traversing the graph |
| Where risk is found | Late, reactively, after the quarter turns | Early, the day the trend bends |
The case for an AI-native operating model, in numbers
The strongest argument is not a headcount cut, it is capacity. The market is loading more onto each asset manager every year, and the firms that absorb that without scaling cost in lock-step are the ones that win. The figures below are drawn from public industry data.
Independent industry data shows asset managers are already being asked to carry materially larger portfolios. Built AI automates exactly the manual production layer, the modeling, data assembly, and reporting, that otherwise caps how much each asset manager can handle. That is what turns rising portfolio load from a cost problem into operating leverage.
With 59% of firms reporting rising expenses against only 46% reporting revenue growth, most firms saw profitability decline. The McKinsey value is concentrated in repetitive, human-intensive back-office work such as financial reporting and reconciliation, exactly the production layer an AI-native operating system absorbs.
The illustrative team and cost model
The capacity shift also changes the cost of running the business, and the effect compounds as the book grows. The model below sizes the asset-manager layer directly from the NAREIM productivity benchmark, a portfolio needs more managers at 13 properties each than at 16, with the manual production layer behind them being the part an AI-native system automates.
The saving grows as the book grows
The team-cost saving grows as a firm adds mandates, so a larger book compounds the operating-model advantage. The bars show the annual team-cost saving at four representative book sizes.
Sources. Capacity figures (16 vs 13 properties, $40m vs $28m NOI per AM, 42% salary-to-revenue): NAREIM & Ferguson Partners, 2026 Global Management Survey (79 firms). AI value ($430–550bn): McKinsey Global Institute, 2025. Salary medians: Glassdoor / Comparably (2025–26), fully loaded at 1.3×. Spreadsheet error rate (88% of 150+ formula sheets contain at least one error): R. Panko, University of Hawaii. Figures are illustrative and frame the order of magnitude of the operating-model shift, not a quotation, and would be refined against your actual portfolio and team.
It sits on top of Yardi, Argus and your model, not instead of them
An orchestration layer that reads your systems of record and writes back only on approval. Read-then-write-on-approval. See the integration model →
No institutional firm wants to rip out the systems it trusts to run its assets, and Built AI never asks you to. The platform is an orchestration layer that reads from your systems of record, reasons across them, and writes back only when a person approves. Your property accounting stays the system of record for the books. Argus stays the valuation authority. Your underwriting model keeps its logic. The platform makes them talk to each other and to the documents.
Yardi · MRI · Entrata
Rent rolls, GL actuals, charge codes and fund accounting flow into the graph as live, normalized entities, read-only by default, so the books stay the books. The variance you review is computed against the same actuals your accounting team just closed.
Argus
Argus stays the valuation truth. The platform reads its cashflow projections into the graph, runs re-leasing and stress scenarios against its assumptions, and reconciles its output to the model and the GL. Argus is invoked, never overwritten.
Leases & loan agreements
Every lease and credit agreement is extracted into the graph with a field-level audit trail. Recoveries, escalations, options and covenant language become queryable and computable alongside the books, so a clause on page 14 is part of the answer.
Connections default to read. The platform pulls from your systems, normalizes what it finds and reasons over it, and at that point it has changed nothing in your environment. The first time anything leaves read-only is the moment you click approve. You get the speed of an automated system with the control of a manual one, and there is no migration and no re-platforming to get there. The full integration model lives on the platform page.
Carlos Olea, CFO, Howard Hughes Holdings (NYSE: HHH)
See continuous variance on your own portfolio.
Bring a fund, an asset and its actuals, and watch the platform build the NOI bridge, forecast the covenants and surface the rollover risk, every number traceable to source. No migration, nothing leaves your tenant, and nothing acts without your sign-off.