Internal Document
Category: Research — Stage 1 Size: 19.8 KB Lines: 210 Commit: d44c2ec Modified: Apr 25, 2026

Agent9 Showing Model Deep Dive

Research by UNO — April 25, 2026

One-sentence verdict: Option B (self-tour with ID-verified lockbox) is the primary recommendation for Agent9's Q3 launch; it is the only option that preserves the brand promise, is operationally proven at scale (Opendoor runs it today), and does not require Agent9 to build scheduling/payment infrastructure for a human intermediary before launch.


Background

Agent9 is a Q3 2026 Ohio FSBO platform — AI communications layer between buyer and seller, $999 flat seller fee. The showing layer is the single operational gap in the AI-intermediated model. Three options are on the table.


Option A — Seller-Led Showings

How It Works in Practice

Buyer taps "schedule a showing" in Agent9. AI schedules against seller's availability calendar. Seller receives confirmation. Seller meets buyer at property, tours the home, fields questions, handles objections. Agent9 collects feedback via automated post-showing survey sent to both parties.

Agent9 sits entirely in scheduling and follow-up. The showing itself is human-to-human, no platform involvement.

Safety / Liability

Conversion Impact

NAR 2025 Profile of Home Buyers and Sellers: FSBO homes account for only 5% of all sales — an all-time low — and 89% of FSBO attempts eventually involve an agent at some stage. Specific seller-led-showing-to-offer conversion data is not published by Houzeo or Beycome publicly; their platforms primarily facilitate listing and scheduling, not conversion analytics. The REDX dataset of 2.7M leads shows FSBO-to-representation conversion at 31.8%, meaning roughly a third of serious FSBO inquiries convert to something — but this measures lead-to-rep conversion, not showing-to-offer directly. Confidence: Medium. No direct showing-to-offer split by model exists in public data.

General industry pattern from agent-assisted sales: homes where the seller is present during showings are widely reported by buyer's agents to generate fewer honest feedback comments and lower offer rates because buyers self-censor in front of owners. No quantified public dataset found on this specific effect for FSBO.

Cost Per Showing

Compatibility with Agent9 Thesis

Breaks the thesis directly. Agent9's brand promise is "no direct buyer-seller interaction until both parties sign." Seller-led showings put buyer and seller face-to-face at the first touchpoint, before any offer exists. This is the exact moment most loaded with social pressure, premature negotiation, and information asymmetry. Option A is the FSBO-as-usual model Agent9 is supposed to improve upon.

What Agent9 Builds to Support It


Option B — Self-Tour with ID-Verified Lockbox

How It Works in Practice

Buyer requests a showing. Agent9 triggers identity verification: government ID upload + selfie match (Stripe Identity or equivalent). Buyer electronically signs a property access release. Agent9 issues a one-time time-limited lockbox code 30–45 minutes before scheduled window. Buyer tours alone. Code expires at window end. Agent9 sends automated feedback prompt to buyer; sends anonymized summary to seller. Seller never sees buyer identity or contact details until after offer is submitted.

Opendoor runs exactly this model at scale today: identity verification via Stripe Identity, unique access code per showing, 8AM–7:30PM access windows, no human escort. This is a proven operational pattern, not a novel bet.

Safety / Liability

Conversion Impact

Opendoor's self-tour model is their primary showing mechanism and they have processed millions of tours without pulling back the model — the premise in the brief that Opendoor/Zillow "pulled back" on self-tour was not confirmed by research. Opendoor continues operating self-tour as of April 2026 (per their current help center). Zillow's partnership with Opendoor on buyer/seller options is active. No public data found showing Opendoor abandoned self-tour; the premise appears to be unconfirmed. Confidence: High that self-tour remains operational; Low that any "pull-back" data exists to learn from.

Specific showing-to-offer conversion by self-tour vs. seller-led model is not publicly available. The directional logic: self-tour removes the "seller hovering" effect, may increase buyer willingness to return and make offers because the experience feels less pressured.

Cost Per Showing

Compatibility with Agent9 Thesis

Fully preserves the thesis. Buyer and seller never interact before an offer exists. All communication flows through Agent9 AI layer. The lockbox is the physical implementation of the "AI-mediated access" promise. This is the brand-aligned option.

What Agent9 Builds to Support It


Option C — Third-Party Showing Agent (Per-Showing)

How It Works in Practice

Buyer requests a showing. Agent9 dispatches to a network of licensed showing agents (via Showami API or Agent9-built network). Showing agent is assigned, meets buyer at property, conducts tour. Agent9 facilitates payment from seller and/or buyer to showing agent. Showing agent submits feedback through Agent9 system after tour. Seller never present.

Safety / Liability

Conversion Impact

No specific per-showing conversion data found for showing-agent-facilitated FSBO tours. Logical expectation: professional agent provides better buyer guidance than unsupervised self-tour, which may improve conversion. But the showing agent has no economic stake in the sale, so motivation to advocate is limited.

Cost Per Showing

Compatibility with Agent9 Thesis

Partially preserves the thesis — buyer and seller do not interact directly. But the showing agent is a human intermediary outside Agent9's control. Their behavior, what they say about the property, what advice they give the buyer, is unmanaged. Agent9's brand promise of "AI handles all communications" gets a human exception for the most important physical touchpoint.

What Agent9 Builds to Support It


Trade-Off Matrix

Dimension Option A: Seller-Led Option B: Self-Tour / Lockbox Option C: Showing Agent
How it works Seller schedules + escorts buyer; Agent9 handles scheduling + feedback only Buyer ID-verified, signs release, gets one-time code, tours alone Licensed agent dispatched, meets buyer, tours on seller's behalf
Seller safety Seller alone with stranger, no documented audit trail ID-verified + signed release + timestamp log Professional on site; strongest oversight
Buyer safety With seller; conflict risk Alone; injury/hazard exposure Escorted by professional
Agent9 liability Low (no physical involvement) Medium (platform facilitates access; needs E&O + release language) High (payment facilitator, potential unlicensed-broker question)
Ohio licensing issue None (owner exemption) None (owner granting permissive entry) Yes — showing agent must be licensed; Agent9 must verify
Conversion impact Unknown; "seller hover" effect likely depresses offers Neutral to positive (less pressure); Opendoor uses at scale Unknown; agent quality variable, no economic stake
Cost per showing (seller) $0 cash, 30–90 min time $100–$250 hardware (one-time) + ~$15 ID verification per 10 tours $50–$150/showing ongoing
Brand promise preserved No — breaks buyer/seller separation Yes — AI-mediated access, no human contact before offer Partial — human intermediary outside AI layer
Q3 2026 build complexity Low — scheduling + survey only Medium — ID verification, lockbox API, release flow High — agent network, license verification, 1099 stack

Primary Recommendation: Option B

Recommendation: Launch with Option B (self-tour, ID-verified lockbox) as the default showing model.

Reason: It is the only option that simultaneously (a) preserves the brand promise, (b) has an existing at-scale operational proof point (Opendoor), (c) is buildable by Q3 2026 with a focused sprint, and (d) does not expose Agent9 to the unlicensed-broker legal risk that Option C carries.

Honest failure mode: If Ohio sellers are uncomfortable putting their home on a fully unescorted self-tour model — either because they don't trust ID verification or because their property has condition issues they'd rather manage in person — Option B adoption rate will underperform. Early seller onboarding calls should surface this resistance. If more than 25% of sellers in the first 30 listings request an escort option, add Option C as an opt-in add-on (not the default).

Option A should not be the default because it breaks the core product thesis at the most important touchpoint. It can exist as a fallback for sellers who explicitly request it, but it should not be promoted.


Option C Compensation Model (If Pursued)

If Agent9 adds Option C as an optional premium layer:


Confidence Levels and Gaps


Sources