Products & Tools

Property Management Puts AI Agents to Work on Real Tasks

Yardi, SkyPSI and AppFolio are giving property-management AI agents live data and multi-step authority, while poor data and safety concerns still slow adoption.

What Building Management Shows About Moving AI Agents From Chat to Action
What Building Management Shows About Moving AI Agents From Chat to ActionVale_Photography / pixabay
By James Calloway6 min read

Why it matters

  • Only 8% of commercial real estate executives said their organizations had integrated AI, according to Deloitte.
  • More than half of property managers now use AI, versus about one in five in 2023, according to IREM.
  • SkyPSI says Operator+ covers 17.1 million U.S. buildings, with an average job worth $13,500.
  • AppFolio says AI actions completed through Realm-X grew sevenfold in one year.
  • More than nine in ten Deloitte survey executives plan to spend more on data and technology next year.

By James Calloway

What does AI actually automate in property management?

Only 8% of commercial real estate executives said their organizations had integrated AI, according to Deloitte, even as almost half reported agents running in some live workflows. Property management offers a concrete test because it coordinates leases, work orders, invoices and physical maintenance. AI agents are software that can act inside business systems, rather than only generate text.

More than half of property managers now use AI, according to an Institute of Real Estate Management (IREM) survey, compared with about one in five in 2023. Most use it every day, often to draft routine messages, which measures adoption more clearly than autonomous execution.

Deloitte says the rest are researching or piloting AI. During the two-month release window covered by Unite.AI, Yardi, SkyPSI and AppFolio shipped agents aimed at operational work. Neither figure specifies how much authority those agents receive or how often they fail. The measured shift is from AI drafting messages to AI attempting work, but most deployments remain narrow.

Can an agent answer from live property records?

Property teams receive questions by phone, email and text. Unite.AI notes that generic chatbots cannot identify which unit has an open repair, while HappyCo reports that about half of resident-reported issues are emergencies.

At the Canadian Apartment Investment Conference in Toronto in September, Yardi said it rebuilt Chat IQ, an agent inside its Virtuoso Enterprise platform. It uses an agent-based architecture, a design that lets the system plan and take actions, and grounds (connects) responses to live Voyager data.

Voyager is Yardi’s core property-management system, so staff and Chat IQ work from the same records. Chat IQ also flags each conversation’s urgency, which tells staff when to intervene. According to HappyCo, most residents accept AI for general questions, and about half accept it for routine maintenance. More than eight in ten want to reach a person first when something urgent breaks.

Yardi Canada president Peter Altobelli said: “AI is no longer a differentiator on its own.” The announcement provides no error rate or conversation volume. Live records improve answers, but urgent maintenance still needs a person.

How can agents turn scattered data into paid work?

Cleaning the windows of a mid-rise building usually requires crews on ropes or scaffolding. Drones can now handle part of that work from the ground. SkyPSI frames this as a way to reduce the labor required for exterior cleaning.

But a drone does not find jobs. A small crew still has to identify buildings and set prices. Pellicano said large contractors use sales teams, while small crews mostly guess.

Dallas-based SkyPSI trains and certifies independent operators to clean commercial buildings with drones. Its free Operator+ app, released on Oct. 27, maps commercial buildings across the U.S.

Each record includes property type, square footage, an estimated exterior-cleaning price and, where available, decision-makers in each management office. SkyPSI says the database covers 17.1 million buildings, with an average job worth $13,500. The announcement provides no independent audit of those estimates.

Behind each record are many public and commercial sources that do not agree. Pellicano told Unite.AI: “We use hundreds of data sources from Google to different data providers.” He added: “I have a team of AI agents, honestly, that go through and sort this all out.”

Operator+ can also generate a proposal. An accepted proposal can become an invoice synced with QuickBooks. An assistant can answer scheduling questions by text.

About 1,000 people registered during a three-week beta. Those registrations do not establish completed paid work.

Property managers often pay for window cleaning with one yearly check. They also ask whether drones work as well as people, whether they are safe and who certified the pilot, Pellicano said.

SkyPSI answers through a certification program in which new operators complete a real job alongside a company trainer in Dallas. The announcement does not disclose completed-job or pricing-error rates. SkyPSI turns scattered data into potential jobs, but its scale remains a company claim.

Can agents execute multi-step property workflows?

A building’s back office links rent collection, receivables, maintenance, accounting and reporting. A request to identify units behind on rent that also have an open repair can cross several modules before staff finish the job.

In June, AppFolio connected its Realm-X AI suite to Anthropic’s Claude through an agent-to-agent connector, a link that lets one AI system hand work to another. A property manager can start an operational job in Claude, and AppFolio executes it inside its own platform.

That execution follows AppFolio’s accounting rules and compliance guardrails, constraints on what the agent may do. Anthropic’s head of Americas mid-market Travis Bryant described the goal as “a secure, proactive system of action.”

In September, AppFolio introduced a new version of Realm-X that can complete multi-step tasks, meaning sequences of dependent actions. It can also build a workflow from one prompt, while a new voice agent answers incoming calls and passes harder calls to staff.

AppFolio says AI actions completed through Realm-X grew sevenfold in a year. In earlier research with IREM, the company found that employees it labeled “AI advocates” managed about a third more units per employee. The company disclosed neither a baseline for the sevenfold figure nor evidence that the unit-per-employee difference came from AI.

AppFolio chief product officer Kyle Triplett said: “We believe the most powerful technology in this industry should be available to every customer.” AppFolio includes the new version in every company plan. AppFolio is testing broader action, but its metrics do not show reliability or causation.

Where does AI still fall short?

Construction professionals surveyed by the Royal Institution of Chartered Surveyors (RICS) ranked safety management and sustainability last when asking where AI matters today. They also placed safety and low-carbon construction near the bottom for the next five years, even as RICS said obligations are rising in those areas.

HappyCo found that fewer than one-third of residents are comfortable reporting an emergency to AI. Commercial property professionals told RICS that poor or missing data is their biggest barrier to AI. System connection ranks second.

Deloitte found that most firms have mapped their data, but fewer than half have added stronger process and security controls. Mapping data does not guarantee that records are complete, current or protected. Those findings leave agent deployments dependent on records that may be incomplete and rules that may remain weak. The hardest areas remain safety, emergency reporting and data control.

What will spending decide?

More than nine in ten executives in Deloitte’s survey plan to spend more on data and technology next year. Nearly half of commercial property professionals in the RICS survey plan to increase AI investment. RICS president-elect Maureen Ehrenberg said: “This is the moment for the profession to lead, not follow.”

The spending plans could accelerate agents that write, reconcile and execute routine property work. Neither the plans nor the adoption figures show how many deployments will pass accounting, safety and maintenance checks. The next test is whether that spending produces reliable workflows companies can check, rather than more pilots.

Source: unite.ai

Share this article:

More from James Calloway

James Calloway

Show full bio

News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.

722 articles

Related articles

  1. The EU's Next Chapter in AI Is Agentic, New Analysis Claims
  2. Australia Leads AI Agent Adoption at 46%, but Managers Pay the Price
  3. Behind the AI Agent Boom, Humans Are Still Making the Phone Calls
  4. Australian firm opens the books on five months of AI agent controls
  5. Zillow Rentals expands proactive AI agents after four-week pilot

« Previous articleNext article »