Opinion

What Building Management Shows About Moving AI Agents From Chat to Action

Ana Sofia Herazo
Add Unite.AI to your preferred sources on Google
Key PointsYardi grounds its Chat IQ agent in live Voyager records so answers match staff data.
View more
  • Yardi grounds its Chat IQ agent in live Voyager records so answers match staff data.
  • SkyPSI uses AI agents to reconcile hundreds of conflicting sources into a usable building database.
  • AppFolio's Realm-X now executes multi-step property workflows inside its own accounting and compliance guardrails.

For most professionals, AI is still a chat window. They use it to draft a message, then decide what to do with it. The harder step is giving an agent access to the systems a business runs on and letting it finish the task itself. Building management, an industry of leases, work orders, invoices and physical maintenance, is one of the places where that step is now being tested.

Property managers already use AI on their own. More than half now do, according to a survey by the Institute of Real Estate Management (IREM). In 2023, it was about one in five. Most of them use it every day, often to draft routine messages.

Companies are further behind. Only 8% of commercial real estate executives say their organizations have integrated AI, according to Deloitte. The rest are still researching or piloting it. Even so, almost half already have AI agents running in some live workflows.

In the past two months, several companies in the sector have shipped agents meant to close that gap. Each one goes after a different part of the problem.

Grounding Agents in the System of Record

Property teams get questions all day, by phone, email and text. A generic chatbot can handle the easy ones. It doesn’t know which unit has an open repair. Many requests also carry real stakes. About half of the issues residents report are emergencies, according to a study by HappyCo, a maintenance software company.

Yardi, the real estate software company, built Chat IQ, the conversational agent in its Virtuoso Enterprise platform, around that problem. In its announcement for the Canadian Apartment Investment Conference in Toronto in September, Yardi said it rebuilt Chat IQ on an agent-based architecture that grounds its answers in live data from Voyager, Yardi’s core property management system, so the agent works from the same records as the staff.

It also flags how urgent each conversation is. That tells staff when to step in. The HappyCo data shows why that layer matters. Most residents were fine using AI for general questions, and about half for routine maintenance, but more than eight in ten said they want to reach a person first when something urgent breaks.

Peter Altobelli, president of Yardi Canada, described what now separates these products in the same announcement. “AI is no longer a differentiator on its own,” he said.

Building the Data Layer for Physical Work

Outside the building, the work is changing too. Cleaning the windows of a mid-rise building has usually meant crews on ropes or scaffolding. Drones can now do part of it from the ground.

But a drone doesn’t find the jobs. A small crew still has to figure out which buildings need cleaning and what to charge. Big contractors have sales teams for that. Small crews mostly guess.

Dallas-based SkyPSI, which trains and certifies independent operators to clean commercial buildings with drones, is using agents to build that missing data layer. Its free app, Operator+, released on Oct. 27, maps commercial buildings across the U.S., with property type, square footage, an estimated price for exterior cleaning and, where available, the decision-makers in each management office. The company says it covers 17.1 million buildings, with an average job worth $13,500.

Behind each record are many public and commercial sources that do not agree with one another. “We use hundreds of data sources from Google to different data providers,” founder Vic Pellicano told Unite.AI. “I have a team of AI agents, honestly, that go through and sort this all out.” On top of that layer, an operator can generate a proposal, turn an accepted one into an invoice synced with QuickBooks and ask an assistant by text which jobs are on the schedule. About 1,000 people registered during a three-week beta.

The other side of the transaction is still learning about the technology. Property managers often pay for window cleaning with a single yearly check and ask whether a drone works as well as a person, whether it is safe and who certified the pilot, Pellicano said. SkyPSI answers with a certification program in which new operators complete a real job alongside a company trainer in Dallas.

Letting Agents Run Multi-Step Work

Answering a question is one step. Much of a building’s back office is a chain of them, spread across rent collection, receivables, maintenance, accounting and reporting, and a single request, such as finding which units are behind on rent and also have an open repair, can touch several of those modules before anything gets done.

AppFolio, another property management software company, has been building toward agents that run that chain inside its platform. In June, it connected its Realm-X AI suite to Anthropic’s Claude through an agent-to-agent connector: a property manager can start an operational job in Claude, and AppFolio executes it inside its own platform, following its accounting rules and compliance guardrails. Travis Bryant, Anthropic’s head of Americas mid-market, described the goal as “a secure, proactive system of action.”

In September, AppFolio introduced a new version of Realm-X that can carry out tasks with several steps. It can also build a workflow from a single prompt. A new voice agent now answers incoming calls and passes the harder ones to staff.

Usage is moving with it. AppFolio says AI actions completed through Realm-X grew sevenfold in a year. In earlier research with IREM, the company found that the people it called “AI advocates” managed about a third more units per employee. “We believe the most powerful technology in this industry should be available to every customer,” said Kyle Triplett, AppFolio’s chief product officer. The new version is included in all of the company’s plans.

Where AI Has Barely Arrived

Several parts of running a building are getting far less of that attention. Construction professionals surveyed by the Royal Institution of Chartered Surveyors (RICS) ranked safety management and sustainability last when asked where AI matters today, and put safety and low-carbon construction near the bottom for the next five years as well. RICS noted that these are the areas where obligations are rising.

Emergencies are another open space. HappyCo found that fewer than a third of residents are comfortable reporting one to an AI.

Underneath both is data. Commercial property professionals told RICS that poor or missing data is their biggest barrier to AI, followed by connecting it to the systems they already use. Deloitte found something similar: most firms have mapped their data, but fewer than half have put stronger process and security controls in place.

That leaves a lot of room, and the spending plans point the same way. More than nine in ten executives in Deloitte’s survey plan to spend more on data and technology next year, and nearly half of the commercial property professionals RICS surveyed plan to invest more in AI. “This is the moment for the profession to lead, not follow,” said Maureen Ehrenberg, RICS’s president-elect.

Ana Sofia Herazo is a Reporter with Espacio Media Incubator covering AI, technology and business. Her work has appeared in International Business Times, Contxto, Fast Company Mexico, 150Sec, and more.