Agentic AI and the Future of Real Estate: Evolution and Implementation

This white paper defines agentic AI, explores how it’s evolving and shares strategies for effective implementation.

Agentic AI, already a driving force for many leading real estate companies, promises to transform the sector in the next few years. Moving far beyond simple automation, AI agents will execute complex tasks proactively without human intervention. They will handle maintenance requests and schedule repairs; screen residential applicants; juggle routine communications from tenants; and alert management to the need for preventive maintenance by estimating when equipment will fail.

Defining the Technology

One rule-of-thumb definition of agentic AI: a solution that extends beyond simply passing information to a language learning model (LLM) and gaining a response. Agentic AI can make autonomous decisions and act on behalf of users, frequently showing the ability to adapt to changing environments of data. The solution could manifest as an individual agent conducting research that connects to corporate data or take the form of multiple agents working collaboratively, with each one handling one component of a complicated workflow.

Agentic AI is capable of scanning a workflow, identifying repeatable and routine tasks and making decisions within predetermined guidelines.

By taking over some high-friction, time-consuming chores, agentic AI crucially frees staff to take on more elevated and higher-touch tasks.

Agentic AI can handle routine tasks like lease renewals, predictive maintenance, and tenant communications without human intervention. It operates independently, learning and fine-tuning decisions and insights on the basis of historical information, market conditions and real-time insights.

A helpful comparison is to think of agentic workflow as a kind of virtual department consisting of multiple team members, which work together for a specific output. Agent #1 creates a task the platform can carry out on its own. The job of agent #2 is to obtain all the data for a certain category — say, expenses or revenue. That agent, in turn, hands off the data to agent #3, which serves as an analyst and organizes the data into a readable format. And a big shift is on the way. By 2027, tasks that are currently done on computers will be done by computers, Miles Brundage, an AI policy researcher and former head of research at OpenAI, has predicted. Those who view AI as a kind of soupedup Google don’t yet recognize that it will be capable of arriving at decisions and will do so effectively.

Expanding Power

Agentic AI is proceeding through a natural evolution that is producing stronger, better LLMs as their intelligence expands. The base model is becoming more powerful all the time. When ChatGPT was unveiled, it was on a par with a high school student asked to perform a few simple tasks. Interacting with GPT4 could be likened to an interaction with a college student. With the recent release of GPT5, the model from a knowledge standpoint is now about on par with a Ph.D. level student with a high IQ, according to David Franklin, industry principal for AI at Yardi.

To determine net operating income, an individual can pull data and promptly come up with a figure. Now, agentic AI can be given the task. Various agents could take on accounts receivable, accruals and other tasks, while a master agent orchestrates the processes to close the books.

Another advancement is agentic AI’s ability to use generative AI to create content. Companies that once relied on off-the-shelf training can now create their own customized training using the language model. For instance, Aspire builds out training for Yardi software, but also enables clients to build their own training. An example might be creating a lesson around workplace safety. The language model that has been trained on human knowledge can search the web and find what’s already been written on the topic.

Agentic AI can already respond to prospect inquiries, review payables, summarize correspondence, communicate with residents about lease renewals and respond to work orders, reviewing recent correspondence with a prospect or customer and summarizing what’s important.

Another way to express agentic AI’s current capability is by monitoring chat technology, which currently allows companies to handle about 92 percent of prospect and resident inquiries, providing a simulated interaction with a team member.

Chatbots can handle routine, repetitive tasks such as answering questions about availability of units, renters insurance and whether 15-month renewals are an option. Yardi is also helping clients train their employees in ways that make them more productive. For example, Yardi Virtuoso can serve as an in-app coach that answers questions about performing a variety of tasks.

More advanced functions are now possible, as well. One example: leveraging known outcomes from the past to predict future outcomes in areas such as customer default, customer expansion, space utilization or a project’s chance for success. Bringing data together in an understandable format through agentic AI can enable the company to intervene in a new way.

Agentic AI's Future

Experts predict that agentic AI will enable clients to use a prompting interface to issue commands and enlist the technology to dive into data and arrive at solutions swiftly. The directives run the gamut from comparing budgets to performance, identifying overspending at properties and preparing a letter to request a meeting with vendors.

Tone matching will allow clients to both respond to inbound requests and handle outbound functions, such as reaching out to residents nearing renewal or in arrears on rent payments.

"You’re not just doing things better and faster. You’re launching new jobs, new businesses, new products — new business models." — Justin Segal, Boxer Properties

“Our vision of the future is that agents will be inside the organization structure adjacent to the humans doing some of the repetitive tasks,” Yardi Vice President Richard Malpica said. He suggests that these chores could include reviewing open work orders, creating new work orders, ordering materials and handling end-of-month closing items.

Agentic AI is undergoing an evolution that some compare to that of the telephone, which was originally a relatively simple apparatus tethered to a single place and is now a sophisticated mobile device capable of interfacing with a huge array of platforms.

“It will actually change the nature of what our company is, impacting the type of work people do, the way resources are allocated and really powering a true transformation — not just improving efficiency but transitioning the nature of the business as a whole,” said Segal.

Agentic AI enables Boxer Property to conduct due diligence for 100 deals at once, 10 times as many as before. As the technology’s capability improves, the company may be able to process as many as 1,000 deals simultaneously. At that volume, Boxer may even be able to add a new business line that screens deals for outside firms. “You’re not just doing things better and faster,” Segal noted. “You’re launching new jobs, new businesses, new products — new business models.”

"We have to make jobs more compelling for people to want to do them." — Cynthia Fisher, Kettler

Franklin foresees a day when the abundance of labor provided by agentic AI will liberate firms to accept projects that are prohibitively costly today. Tasks that require sifting through data will help produce better decisions, whether regarding asset prices or the cost of delivering services.

That improved information will empower end users to peer further into the future. Freeing staff from maintenance requests and other time-consuming duties will cut the cost of providing services.

AI’s capacity to write codes has significantly cut the cost of developing software. “Being able to be nimble and change will be very important,” Franklin said. “The people at risk in the workforce are those unwilling to adopt AI and to use it to become more effective.”

For those in service industries, agentic AI will be a game-changer, executives predict. It won’t mean the elimination of employees but will instead facilitate placing those human resources in positions where they can add the greatest value. KETTLER’s Fisher likens the change to the robotics evolution in the automobile industry, which led to greater capacity and consistency.

“We have to make the jobs more compelling for people to want to do them,” she said. “We’re taking out those tasks that won’t make people inspired or excited about what they do each day.” The result will be improved job quality and property performance. Using data to better manage properties without adding staff will bring consistency.

In about five years, agentic AI will reach a “truly transformational” phase, predicts Wyatt Mayham, CEO & co-founder of Northwest AI Consulting. The AI-powered asset manager won’t just react to a work order but will take action, using IoT sensor data to predict an HVAC breakdown and schedule preventive maintenance. That will help free the human asset manager to focus on high-level strategy, including AI management, capital projects and acquisition decisions.

Transitional Challenges and Strategies

Agentic AI has the potential to make employees happier in their workplaces by removing them from heavily administrative, paper-oriented environments. Yet as is almost always the case with new paradigms, fear of change poses a potential obstacle. That reality calls for organizations to reset their teams’ responsibilities to the way work is now performed and help them grasp their recalibrated roles, as well as what’s coming.

Even if AI were perfect, it would take several years for employees to grow accustomed to the new concept and for their operations to fully integrate the new technology. “We have this tension that’s occurring because the rate of change is so fast that it’s too fast for people to absorb,” Franklin said. “By the time they figure out a way to use it, it has changed.” He advises organizations to avoid chasing every new thing and think through what the change means and the best way to adopt and adapt to it.

End users need a clear and responsible road map featuring solid, structured data. The second commodity they require is a platform that’s rigorous, secure and user-friendly. Finally, they need expertise, from within the company but also from outside resources, including the right vendors, partners and other team members.

Unless organizations’ approach to agentic AI is organized, coordinated and orchestrated, “it’s going to get messy very quickly,” Segal advised. It’s essential to consider safety, security, business risks, regulations, audits and implementing AI in a responsible, manageable way, he added. “To the extent they’ve tried things and had proofs of concept, these things are difficult to manage and keep track of and they can often do important things without oversight within the organization,” Segal said. “And as we get more powerful tools, those risks become greater.”

To elevate their odds of success, organizations should give their people freedom to explore agentic AI, Franklin recommends. While employees may be afraid to leak data by interacting with ChatGPT, companies should err on the side of providing them access to tools and help them discover the value they can create with agentic AI.

“That’s how an organization can succeed,” Franklin said. “You want to get your reps in and get better at using it. Next thing you know, you’re a power user.”