The total U.S. rental housing stock runs roughly 24M apartment units (Census / ACS-derived industry estimate, not a Fillment figure). Our addressable subset is the professionally-managed slice — 200+ unit portfolios, REITs, and boutique firms using a modern PMS — which is on the order of half the stock, ~12M units. At our $1.50/unit ARPU that is a ~$200M/year SAM at full SOM penetration, and our working assumption is low single-digit share over the next 3–5 years. Every number above is an industry estimate; current traction metrics are in the deck.

Two asymmetric moats. First, the inbound-message loop — we sit where Yardi and RealPage do not, because they sit downstream of it. The first-party PM relationships and the real inquiry + tour data flowing through nightly are ours. Second, their buyers pay for control, not automation, so a fully automated reply product is a surface Yardi will not ship. Migration cost reinforces the lock-in: switching a PMS is a multi-quarter project, swapping an inquiry-response layer is a single weekend. Regional call-centers (Real Synergy, Ellis, and similar) compete on labour arbitrage — an AI cost curve breaks that.

The $1.50/unit ARPUs hold up against the $149/month floor under portfolios of ~100 units as a linear price; below that the floor kicks in. Variable COGS is LLM inference for the inquiry-response loop plus outbound SMS — order-of-magnitude estimate is low-tens of cents per handled inquiry at current model pricing, not a real P&L. Our working contribution-margin band is ~70–80% gross, expected to improve as prompt-cache hit rates climb. These are forward-looking targets, not filed actuals; current actuals are shared on request under NDA.

In scope: the federal Fair Housing Act, state fair-housing acts, TCPA, the Florida Telephone Solicitation Act, the Oklahoma mini-TCPA, Washington's CEMA, and CCPA / CPRA. FCRA exposure is bounded because applicant and credit data stays in the PMS — Fillment stops at inquiry and tour booking, so we never sit on a consumer report. Guardrails shipped in the product: no auto-decisioning on protected-class inputs (the LLM is never given race, family status, disability, or source-of-income); every outbound message routes through a manager-review queue (no fully-automated send); SMS cadence caps and quiet-hours enforced at the integration layer; opt-out honoured within seconds.

These are public reference points, not a forecast exit multiple. In PropTech, the closest public-market peer is AppFolio (historical ~10x ARR trading multiple range). RealPage has a track record of acquiring category-leaders including Kigo and Modern Message; Yardi's private M&A pattern runs longer. In vertical AI SaaS, Harvey AI (legal) and Tennr (healthcare operations) have reported meaningful funding traction as reference points; LayerPath sits closer to a comparable-curve deal. None of this is a forecast for Fillment — it is the category context an LP will ask for.