How a CRE investor underwrites every deal in its market

Every acquisitions team we've worked with describes the same failure. Deals arrive through brokers, and the team works what arrives. Off-market sourcing, where the actual edge lives, happens when someone has a free afternoon, which is never. The opportunity set is enormous and the pipeline is whatever fit in the inbox that week.

The second failure is subtler and more expensive. When underwriting is done under time pressure across dozens of deals, assumptions stop being visible. A pro forma arrives at an investment committee with a number in the exit cap rate cell, and nobody remembers whether that number came from a comp, a broker's claim, or someone's instinct eight days ago. The model looks identical either way.

We built a coordinated team of agents that runs the acquisition workflow end to end, with each agent owning one stage and handing structured output to the next.

• Sourcing monitors public property records, tax delinquency and lien filings, code violations, ownership and entity data, and refinancing pressure signals, surfacing off-market and lightly marketed properties with the motivation signal that makes each one worth a call.
• Underwriting assesses each sourced property against the firm's actual thesis: asset type, current operations, rent roll and occupancy assumptions, expense loads, insurance exposure, renovation needs, and exit strategy, producing a thesis, the risks, the value-add levers, and a defensible offer range.
• Pro forma builds the model across conservative, base, and upside cases, from purchase price through cash-on-cash, IRR, equity multiple, and sensitivity, with every input traceable.
• Outreach drafts the owner, broker, and lender contact that turns a scored property into a conversation, in the firm's voice and against its suppression rules.
• Orchestration scores every deal on location, asset quality, seller motivation, value-add potential, financing feasibility, and exit potential, then keeps a live prioritized pipeline with a next action on each row.

The client's hardest requirement wasn't accuracy. It was epistemic discipline: never present an assumption as a fact.

So every financial output defaults to conservative unless upside is specifically supported by data. Where a number is missing, the system says what's missing and supplies a labeled placeholder rather than a confident guess. Every deal carries a confidence score and a source record. Click any figure and you get the document it came from. And the arithmetic never runs through a language model: returns, coverage, and sensitivity are computed in tested code, with the model responsible only for extracting inputs and explaining results.

That last one is not a detail. A plausible-looking wrong IRR is worse than no IRR, because it survives review.

The agents never make an offer. They source, score, underwrite, model, and draft, then stop. Investment decisions, offer submissions, and contract execution remain entirely with the firm's people, which is both what the client wanted and what keeps the system on the right side of licensure questions. The system recommends; humans decide and act.

The pipeline is no longer bounded by who had time to search. The team now reviews 45 scored off-market opportunities a week against a thesis they control, with time from first-touch to a committee-ready memo down from 11 days to 2. Every memo arrives with its assumptions labeled and its sources attached.