AI Analysis of a 48-Unit Multifamily Property

This case study walks through a real-shape multifamily acquisition opportunity surfaced from an Eastdil-style broker package, run end-to-end through the LargeKite multi-agent pipeline. The deal numbers are representative of mid-market multifamily activity in the Southeast in mid-2026 and were chosen because the broker pro forma sits in a familiar middle ground — not egregiously aggressive, not unusually conservative. This is the kind of deal that gets approved or passed on assumption discipline rather than headline metrics.

The deal at a glance

A 48-unit garden-style multifamily built in 1985, located in the West Tampa submarket, asking $12.4M ($258K/door). 91% occupied on the cover; trailing-twelve actual occupancy 89.4%. Three-building configuration on 2.4 acres with a unit mix of 18 × 1BR, 22 × 2BR/1BA, and 8 × 2BR/2BA averaging 905 sqft. Broker pro forma NOI of $725,400 implies a 5.85% cap.

The seller is the original owner since 1998. Last interior refresh was 2008. The property has been managed by a regional generalist PMC for the past nine years.

Step 1: OM extraction

The OM PDF was uploaded to the analyzer. In 47 seconds, the structured extraction returned:

| Field | Value | |---|---| | Purchase Price | $12,400,000 | | NOI (broker pro forma) | $725,400 | | Cap Rate (broker) | 5.85% | | Occupancy (cover) | 91% | | Average Rent | $1,410 | | Total Expenses | $408,300 | | Unit Types | 3 |

The extraction agent also surfaced four red flags without prompting:

Trailing-12 vacancy of 9.1% materially above pro forma 5.0%. 34% of leases expire within 90 days of close — significant rollover risk. Three units have been vacant 120+ days at listed asking rents. Property tax pro forma does not reflect post-sale reassessment uplift in Hillsborough County.

Each of these would have been findable by a careful analyst reviewing the document. The point is not that AI sees things humans don't — it's that AI surfaces them every time, in under a minute, in a consistent format.

Step 2: Specialist agent fan-out

Four specialists ran in parallel:

Underwriting recast the cap rate from 5.85% to 4.95% by adjusting three line items the broker had stripped: post-sale property tax basis (~$22K/year incremental), full-load management fee at 3.5% (~$14K incremental), and replacement reserves at $300/door/year (~$14K incremental). It also flagged the rent push assumption (7.1% year one) as aggressive given submarket supply pressure.

Market Research identified the West Tampa submarket as a 5-year out-performer (rent CAGR 5.2% vs metro 4.6%) currently absorbing 1,840 units of new construction through Q4 2026 — about 4.1% of existing stock, which it flagged as a yellow zone. Demographic and employment fundamentals were graded healthy.

Risk Analysis scored seven categories, three of which came back HIGH: rent assumptions (aggressive given supply), vacancy (9.1% trailing vs 5.0% pro forma combined with concentrated rollover), and financing sensitivity (DSCR 1.18x year one). Demographics and economic weakness graded LOW. Overall risk score: 62/100.

Operations identified the current PMC as a regional generalist without strong renovation-execution credentials and recommended transitioning to a multifamily specialist at close. It scoped capex at $9,200/door for interior turns plus $57,600 for non-discretionary HVAC replacements plus ~$140K for common-area refresh.

Step 3: IC synthesis

The IC agent consumed all four specialist outputs and produced:

Classic value-add play with honest upside if underwritten honestly. The 1985 vintage in a strong submarket at a defensible basis ($258K/door) supports a renovation-and-rent-push strategy that should deliver a 16-19% IRR over a 5-year hold. The deal works if — and only if — we recast the broker pro forma to reflect TTM actual vacancy, post-sale tax basis, and a realistic 18-month stabilization timeline. At the asking price of $12.4M, the math is tight; a $400-500K price reduction would create the margin of safety we want.

Recommendation: Conditional. Counter at $11.9M with a 60-day diligence period.

What the agents got right

Two things stood out as genuine value-add from the agentic workflow versus a junior analyst doing the same work:

1. The tax reassessment catch. Hillsborough County reassesses on sale, and the pro forma did not reflect the uplift. A junior analyst on their 12th deal of the quarter would probably catch this; on their 25th, maybe not. The agent catches it every time.

1. The lease ladder analysis. The 34% rollover within 90 days was buried in the rent roll, not the OM narrative. The risk agent surfaced it as a high-severity flag with the specific number rather than a generic "rollover risk" callout.

What the agents got wrong (or couldn't see)

Three things the analyst still needs to do after the agents finish:

1. Inspect the actual PMC transition feasibility. The agent recommended transitioning at close but doesn't know which multifamily specialists are currently taking on assets in the West Tampa market. The analyst has to make calls.

1. Verify the submarket supply pipeline. The 1,840-unit figure is based on the model's best understanding of public data. Real CoStar or Yardi data may show concessions and lease-up velocity that change the underwriting. The analyst should pull primary data before committing.

1. Source-of-funds and sponsor-fit. The recommended counter-bid ($11.9M) assumes the buyer has 25% equity and access to Fannie DUS at 65% LTV. The agent doesn't know the buyer's actual stack.

The counter-factual

If this deal were underwritten on broker pro forma alone, the headline cap rate of 5.85% would have made it look attractive enough to pursue. The recast to 4.95% — combined with the rollover concentration and supply pressure — is what moves it from "pursue at ask" to "conditional with a counter." That's a material difference in outcomes: bidding at ask on a deal that's actually a 4.95% trade is how funds end up with a 7-year hold returning 8% IRR rather than the modeled 17%.

The point of the structured analysis is not that the AI is smarter than a good analyst. It's that the AI catches the same things every time, surfaces them with consistent severity, and produces them in 90 seconds rather than 8 hours. The judgment — whether to counter at $11.9M, whether to walk if the seller won't move, whether the buyer's capital stack supports this even at $11.9M — remains with the human.

What you can do with this

If you're evaluating a similar value-add multifamily, run your numbers through the Deal Analyzer. The pipeline is the same as the one used here; only the inputs change. For OMs, upload the PDF and the extraction agent will return the structured fields.

If you want the long version of the underwriting framework behind the recast, the Lessons from Analyzing 100 Multifamily Listings research note covers the broker-assumption-stack pattern in depth.