Designing a concierge-level operating model for premium residential assets

Aurora Residences owns and operates a portfolio of premium rental properties in the U.S. Sunbelt and along the Mediterranean coast the kind of assets where the tenant experience is part of the investment thesis. Residents expect concierge-level responsiveness; the operator's margins depend on running the portfolio lean. Those two facts were in tension long before we got the call.

The previous operating model was a rotating front desk supplemented by a ticketing system. At 2 a.m. when a resident had a burst pipe, the on-call line rang a voicemail box that a property manager checked in the morning. Tenant NPS was slipping. Renewal rates were slipping. Replacing a rotation of night-shift concierges across fifteen properties was not economically viable.

A multi-agent concierge

We built a three-agent system that fronts every tenant-facing communication channel inbound phone, SMS, and resident app with a voice and written interface the tenant experiences as a single named concierge.

• A Concierge Agent handles all first contact. It has full context on the tenant (lease, payment history, past tickets, communication preferences) and can resolve the majority of requests package coordination, amenity bookings, access codes, guest registrations without escalation.
• A Maintenance Agent takes over when a request requires physical work. It triages severity, dispatches to the correct in-network vendor, schedules against tenant availability, and tracks the job through completion and tenant confirmation.
• A Renewals Agent engages tenants 120 days before lease expiry with personalized renewal terms, handles negotiation within operator-set bounds, and routes exceptions to the asset manager.

The handoff that matters

The hardest design problem was not what the agents do it was when and how they hand off to a human. A tenant complaining about noise at 2 a.m. is almost always resolvable with an agent; a tenant reporting a water leak is not. We built a severity classifier that operates on live transcript, and an on-call rotation of human property managers who are paged only for events the agent rates above a contextual threshold. The agent stays on the line, briefs the human on everything it has already gathered, and transfers with full context. The tenant does not repeat themselves.

Results after twelve months

Tenant NPS is up 34 points across the portfolio. Average first-response time on tenant inquiries dropped from 42 minutes to under two minutes, measured 24/7. Renewal rates improved 11 points. The number of human property-manager FTEs required per 100 units dropped from 1.4 to 0.6, without reducing the service level that residents see because the service level actually went up.

“Our residents don't know or care that it's an agent. They know that when they call, someone who knows them picks up, and their problem gets solved.”

Chief Operating Officer, Aurora Residences