AI Owner-Propensity Scoring: Which CRE Owners Are Likely to Sell

Every off-market deal starts with the same question: of the hundreds of owners in your target market, which ones are actually likely to sell in the next 6 to 12 months? You cannot call all of them, so the real job is ranking. AI owner-propensity scoring takes a list of owners or parcels and orders it by likelihood-to-transact using observable, public signals: how long each owner has held, when their loan matures, whether their portfolio is churning, how old the holding entity is, and whether there are deferred-maintenance or tax-delinquency flags. The output is a ranked shortlist with the reasoning spelled out per owner, so your first calls go to the names most worth your time.

In CRE terms, this is a call-order model for off-market sourcing: rank which owners deserve verification, enrichment, and outreach first. It slots into the front of a sourcing pipeline, so if you have not built one yet, start with the free CRE Off-Market Sourcing Agent pack and the Claude Skills for Real Estate templates, then layer propensity scoring on top to decide call order. One honest framing up front: AI is the analyst that ranks the list. You still make the calls and the judgment.

The signals that actually move the needle

Propensity scoring is only as good as the signals you feed it, and the useful ones are observable from public or commercially available records, not guesses about intent. Five carry most of the weight. Hold period versus typical hold: an owner sitting well past the normal hold for that asset class and market is statistically closer to a decision than one who just closed. Loan maturity and the refinance window: a balloon coming due into a tighter rate environment is one of the strongest catalysts, because the owner has to act. Portfolio churn: an owner who has been trimming or recently sold nearby assets is signaling a posture. Owner life-stage and entity age: a long-aged individual-held entity or an estate situation changes the calculus. Deferred-maintenance and tax-delinquency flags: visible distress or a tax lien correlates with willingness to move.

None of these is a guarantee, and that is the point. Each is a correlation, not a prediction. Scoring them together just tells you where the conditions for a sale are most present, so you spend your outreach on owners where the odds are tilted rather than working the list alphabetically. For the wider context on how this fits modern sourcing, see AI deal sourcing for real estate investors.

The scoring prompt

The workflow is a paste-and-rank: you give the model the owner or parcel list (an assessor export, a CRM pull, or a CSV of LLCs and addresses) plus whatever signal columns you have, and you tell it how to weigh them. A workable prompt structure looks like this: "You are an acquisitions analyst. Here is a list of property owners with the fields I have for each (hold period, loan maturity date if known, recent sales nearby, entity age, any maintenance or tax flags). Score each owner 0 to 100 on likelihood to sell in the next 6 to 12 months, weighting loan maturity and over-typical hold period most heavily. For each, give the single most important reason for the score and flag where a key field is missing so the score is low-confidence."

Two rules keep the output honest. First, force the model to cite the signal behind every score, so you can sanity-check its logic instead of trusting a black-box number. Second, make it flag thin data: an owner scored on two fields is not the same as one scored on five, and you want that distinction visible. The model is doing the same triage your AI Deal Screener does for inbound deals, just pointed at owners instead of offering memoranda.

Building the ranked shortlist

The output you want is not a wall of scores, it is a working call list. Have the model return the owners sorted high to low, grouped into tiers (call this week, call this month, monitor), each row carrying the score, the one-line reason, and a confidence note. A good shortlist reads like a brief: "Owner held 14 years against an 8-year typical hold, loan matures Q1, scored 86, high confidence" tells you exactly why this name is at the top and what to lead with on the phone. Owners with strong signals but missing data land in a "verify first" bucket rather than getting a falsely high or low score.

From there the shortlist feeds the rest of the pipeline: skip-trace and verify the top tier, enrich contacts, then start dialing. If you are still assembling your sourcing stack, the best off-market deal sourcing tools for CRE rundown covers where data and contact enrichment come from, and the AI deal-sourcing solution wires the whole chain into one workflow.

The first-touch outreach line

For the top names, the same scoring pass can draft a first-touch line, and the value is that it is grounded in the specific signal that ranked the owner, not a generic "are you interested in selling." An owner flagged on a maturing loan gets a different opener than one flagged on a long hold or an estate situation. Ask the model for a short, specific, low-pressure line per top owner: reference the property, lead with credibility, and make the ask soft. You are not sending these blind. You read each one, adjust for what you actually know about the owner, and decide whether a call, an email, or a letter is the right channel. When an owner does say yes, the next step is paper, and you can move straight into drafting terms with a CRE letter of intent drafted with AI.

Where it stops and what stays human

Be clear-eyed about what this is. Propensity scoring prioritizes who to contact from observable signals. It does not predict the future, it does not read an owner's mind, and a high score is a reason to call, not evidence a deal exists. It also does not replace the verification layer: before you act on a top name you still skip-trace to confirm the real owner behind the entity, confirm contact details, and validate the signals against current records, because stored data goes stale. And it never replaces the relationship work that actually closes off-market deals, the repeated, patient human contact that earns a seller's trust over months. The model ranks the list and explains its reasoning. The investor makes the calls, builds the relationship, and owns every judgment call. The same human-in-the-loop line holds across the workflow, including when you turn a property tour into a writeup with a site-visit voice memo into an investment summary.

Where to start

Start with the list you already have: a county assessor export or your existing CRM of owners in one target submarket. Install the free CRE Off-Market Sourcing Agent pack to build the pipeline, then add a scoring pass that ranks owners on hold period and loan maturity first, since those are the highest-signal fields and the easiest to source. Keep the model's reasoning visible, verify the top tier before you dial, and treat the score as a call-order tool, not a verdict. When ad-hoc prompts stop scaling and you want owner scoring running against live data feeds with your CRM, our investor AI advisory maps the build with the fastest payback for your acquisitions team.