Case Study · Founder, Orchestrator & Payments Agent · 2026

Property Pilots

A multi-agent AI system that gives a single landlord the operating capacity of a full property-management company.

TL;DR

Small, independent landlords own roughly 60–70% of America's ~50 million rental units, yet they're stuck choosing between paying a property manager ~10% of rent or drowning in manual coordination. Property Pilots replicates an entire PM team as specialized AI agents — payments, onboarding, listings, leasing, compliance — coordinated by a single orchestrator. I lead the team and own the Payments Agent. We stood up a working multi-agent prototype on Claude in under 30 minutes after a heavier framework stalled.

01 Context

There are about 50 million rental units in the United States, and an estimated 60–70% of them — roughly 30–35 million units — are held by small, independent owner-operators rather than large institutions. It's a massive, fragmented, underserved majority.

These landlords rarely have the time or scale to run their properties like a business. Military homeowners feel it most acutely: a deployment or PCS move turns a house into a remote asset that has to be managed from another state or country. The market has a framework for this problem — the Frei & Morris lens of "solve for trust" and treating new capability as a workforce decision — which is exactly the frame Property Pilots is built on.

02 Problem

Traditional property managers charge around 10% of monthly rent for work that is largely coordination — screening prospects, chasing paperwork, fielding maintenance calls, managing listings and vendor emails. For a small landlord on thin margins, that fee is painful; doing it all yourself is a second job.

The forced choice: high cost, or high time. Property Pilots exists to remove it. How might we give landlords the full capabilities of a property-management team — without sacrificing their time, margins, or control over their most valuable asset?

03 Approach & Process

Instead of building one general-purpose "property bot," we modeled the product on a real property-management org chart. Each human expert on the team defines the rules for the AI agent that mirrors their role — so the system is structured like a team, with clear responsibilities and hand-offs, not a single black box.

FunctionHuman expertAI agent
CoordinationCalvin Riley — team leadRex — Orchestrator; routes tasks and holds the workflow logic
Payments & financial opsCalvin Riley — paymentsPayments Agent — rent collection, processing, financial reporting, repair-payment coordination
Onboarding & productGigih Nugraha — resident PMOnboarding Agent — site surveys, owner contracts, property data intake
Listings & marketingAngela Xu — marketingListing + Marketing Agent — Zillow/Airbnb posting, lead capture, tenant messaging
Leasing & complianceJoshua Cole — legalTenant + Compliance Agent — screening, lease generation, Fair Housing guardrails

04 AI Integration

A central Orchestrator ("Rex") receives a task, decides which specialized agent owns it, and routes accordingly — keeping the workflow coherent across payments, onboarding, listings, and compliance. The operating principle is deliberate: the human defines the rules; the AI executes at scale.

I own the Payments Agent — the piece closest to my day job in payments. It automates rent collection and payment processing, produces real-time financial reporting the landlord can actually read, and coordinates payment for facility repairs. The human still owns judgment: approvals, exceptions, and any decision that affects a tenant's housing. AI handles the volume; the person keeps the wheel.

05 Iteration

The most useful lesson came from a tooling dead end. We first tried to stand up a custom local agent framework, then pivoted — and the pivot is the whole point.

DecisionOptions weighedCall & rationale
Prototype platformOpenCLAW (custom local agent framework) vs. Claude ProjectsClaude Projects. ~1.5 days on OpenCLAW on an old Mac Mini produced nothing usable in the timeline; Claude configured a working multi-agent property manager in under 30 minutes — no API, no infrastructure.
ArchitectureOne general agent vs. orchestrator + specialistsOrchestrator + specialists. Mirroring the PM org chart made roles explainable, auditable, and easy for each human expert to own.
Pricing model~10% of rent vs. flat subscriptionFlat monthly subscription. Aligns cost with software value delivered, not with the size of someone's rent check.

06 Evaluation

Handing an AI system partial control of someone's most valuable asset — and decisions that affect where people live — raises real risks. We designed guardrails in from the start rather than bolting them on.

RiskMitigation
Bias in tenant screeningFair Housing compliance baked into screening logic; a human can always review and challenge any decision
Sensitive data exposureEncryption, strict privacy compliance, and minimal data collection — only what's necessary
Over-automation of a major assetHuman-in-the-loop on approvals and exceptions; AI executes, the owner decides
Loss of trustEvery AI action is explainable and auditable — what happened, why, and what it cost — with real-time financial reporting

07 Aspirational Impact & KPIs

Targets Property Pilots is built to own at scale:

~10% → flat
Replace percentage-of-rent PM fees with a predictable monthly subscription
30–35M
Underserved owner-operator rental units in the addressable market
5 agents
Specialized roles coordinated under one orchestrator

08 Key Learnings & Stewardship

If I scaled this: keep the human as the decision-maker on anything that touches a person's housing, keep every agent's action auditable, and let the orchestrator — not a person's inbox — absorb the coordination load.