Agentic AI—the idea that artificial intelligence can operate semi-autonomously, take initiative, and complete complex workflows without step-by-step human prompting—is moving from theory to practice in real estate. On this week’s inaugural Insights by Blueprint Subscriber Call, four industry leaders explored the growing role of Agentic AI:
- Stephanie Versin, SVP Sales, Marketing, and Customer Experience, Brookfield;
- Anne Baum, VP Marketing, Towne Properties;
- Harshit Shah, CTO, Divcore Capital
- Saharsh Chordia, VP, Cortland
Through the hour-long discussion, panelists described how they are already using agentic systems in their organization and what comes next. Their conversation revealed early proof points, organizational challenges, and the likely trajectory of AI adoption across leasing, operations, and investment management.
This week’s report will outline those learnings and explore where agentic AI is going from here. Subscribers also have access to the full recording linked at the bottom of this report.
1. Early ROI: Leasing, Renewals, and Collections
Leasing automation is where most firms have achieved the first measurable returns.
Brookfield’s Stephanie Versin described how “virtual listing agents, scheduling tours, answering phones, voice AI” now handle top-of-funnel interactions around the clock. Brookfield uses Elise AI, which integrates voice, web chat, and self-guided-tour functions: “You have everything on the same platform and it does integrate with your legacy system, which is the most important for us.”
Towne Properties’ Anne Baum reported similar gains: “We’re testing Elise AI for leasing, collections, and renewals.” Her metrics focus on shorter renewal cycles and higher on-time payments. She added that AI-based lease audits identify “revenue lost that is then found” and flag “potential liabilities” where overcharges or compliance errors might arise.
These tools have shifted leasing staff from repetitive communication toward qualified-lead conversion. As Versin summarized: “We only filter the qualified leads to be given to the onsite team… the handoff is really to triage at the top of your funnel.”
2. Internal Operations: Training, Talent, and Simulation
While leasing is the most visible use case, Cortland’s Saharsh Chordia highlighted gains in training and talent development:
“Where we’re reaching a lot of value with AI is not necessarily efficiency gains—it’s doing things that we just didn’t have time to before.”
Cortland uses conversational training simulations to let staff “do reps” on workflows that would be too time-intensive for human trainers. “It’s not real-estate-specific,” Chordia said, “but we’re using it across leasing and centralized operations.”
Anne Baum added that her team has begun experimenting with avatar-based video training using HeyGen. “Our HR team loved it,” she said. “They thought we were going to come back with PowerPoint animations, and instead we came back with an avatar that explained open enrollment.”
Though early, these experiments point to agentic systems that personalize learning and feedback at scale—an area with immediate cost and consistency benefits.
3. Investment and Back-Office Applications
Harshit Shah, CTO of Divco West and its credit affiliate Loan Core Capital, framed AI evolution over the past decade: from data warehousing and BI, to generative text, and now to agentic systems that “have a brain of themselves to a certain degree.”
Divco’s custom agentic tooling focuses on investment diligence and data ingestion:
- Deal memos and classification: “Once you have the pipeline, how do you get to the entire memo writing? How do you have the model classify deals into different ratings and drive actions based on that?”
- Document extraction: “Information was trading as papers—Excel, PDFs, T-12s. We’re trying to tap into that and build a custom agentic approach to extract and load into our warehouse.”
These initiatives aim to “do more with less” while standardizing outputs across equity and credit underwriting. Shah emphasized that Divco keeps “humans in the loop” but is actively pushing for end-to-end automation of data flows between third-party managers, acquisitions teams, and asset managers.
4. Measuring Impact: Productivity, Not Headcount
Despite fears of displacement, none of the panelists reported near-term headcount reductions. Versin explained:
“We are really removing the mundane task and the repetitive task from our onsite teams… as long as we increase productivity per full-time employee, there is no reason to cut.”
Chordia agreed that change is coming more in hiring expectations than layoffs. Shah elaborated:
“The conversation on new hiring is changing. When you pitch an idea to hire an analyst or a marketing person, leadership asks, are you sure you can’t extract more work by leveraging technology?”
Baum confirmed the same trend: “It was the first time I’d ever been asked whether AI could accomplish some of the tasks in my hiring plan.”
The consensus: agentic AI is redefining job scope and process design, not immediate staffing levels.
5. Keeping the Human Touch
Versin cautioned against over-automation: “Sometimes we trust a leasing AI to really do the whole leasing process without the human interaction, and I think this is where the recalibration is very important.”
Her prescription: insert human contact at key touchpoints.
- A confirmation call 24 hours before a tour “builds that relationship and prevents no-shows.”
- A personalized follow-up email after a tour—referencing specific units viewed—signals authenticity: “That’s where the prospect feels, I’m not talking to a bot anymore.”
For operators seeking to balance efficiency with empathy, these low-tech gestures remain essential to conversion and retention.
6. Common Implementation Mistakes
The panel identified recurring pitfalls in early deployments:
- Insufficient training data. Baum admitted, “We made the assumption that 80 percent of questions would be about availability… We may have lost some high-quality prospects because we inadvertently created friction.”
- Poor data hygiene. Versin stressed, “It’s trash in, trash out… making sure that you have accurate, updated knowledge in the bank the AI is using.”
- Going too small. Chordia warned, “There’s diminishing returns if you focus on too small of a use case. It’s better to go big because it’s just as hard to get the right output.”
- Compliance blind spots. Versin added, “Make sure your AI doesn’t violate any rules and regulations… Any affordable-housing question should go to a human being.”
- Lack of integration planning. Shah noted, “You have to have programmatic capability to tap into different systems… if your system isn’t open, it’s not going to make it happen.”
Each lesson underscores that AI success depends less on algorithms than on governance, data discipline, and process scope.
7. Build vs. Buy: Differentiation vs. Speed
A key strategic question was whether to develop custom agentic systems or rely on vendors.
Cortland’s Chordia summarized the company’s approach:
“Buy for speed, build for differentiation.”
Off-the-shelf platforms like Elise AI and Funnel accelerate deployment for commoditized tasks, but internal teams focus on areas that define competitive advantage—such as proprietary underwriting or resident analytics.
Brookfield’s Versin echoed the priority on integration: “Having a vendor that can be adaptable, connected, and already integrated with legacy systems is a win-win.”
Towne Properties relies on a centralized technology-vetting team to ensure interoperability: “Any new technology has to go through them… it allows us to make sure all of our technologies are speaking together.”
Divco’s Shah added that “industry vendors will get you to a certain point, then you’ll need customization on top.” The firm treats vendor selection as an ongoing sourcing exercise—balancing external innovation with internal data strategy.
8. Risk, Compliance, and Change Management
Jamestown’s technology chief asked how panelists handle legal and compliance pushback.
Chordia’s tongue-in-cheek response—“Just don’t ask legal and you’ll be fine”—drew laughs, but he quickly clarified that the key is “open dialogue” and transparency about boundaries around bias and fair-housing rules.
Shah described deeper organizational friction:
“We have encountered challenges with certain departments where even though it might be helpful, the changes are hard because they’re so integrated into Excel. The way to overcome that is making sure the executive team sets the agenda.”
His takeaway: progress requires both executive sponsorship and service-oriented technology teams who “co-lead” experimentation while respecting each department’s risk tolerance.
9. Beyond Automation: Toward True Agency
A question from the audience asked whether any current deployments qualify as truly agentic, with adaptive decision-making beyond scripted tasks. The consensus: not yet, but the direction is clear.
Versin’s wish list:
“I want agentic AI to do my budgets next year… have goals, have all the criteria, and have the agent divide it by market and build plans automatically.”
Baum predicted a shift from vendor-led automation to enterprise-led AI strategy: “We’re going to start seeing AI strategy come from internal leaders… looking across operations, accounting, tech, and marketing. That’s when truly agentic workflows will take shape.”
Versin also envisioned resident-experience agents that act as “one brain behind everything—maintenance tickets, payments, concierge services—saving residents time and integrating all our platforms.”
Shah pointed to Divco’s custom deal-review agents and monthly-reporting bots as early steps toward autonomy: systems that parse third-party data, generate analyses, and trigger alerts “without much prompting.”
10. Talent and the Next Generation
Amherst’s Richard Rodriguez asked whether lack of AI tools might hinder recruitment.
Chordia’s view: for technical teams, yes. “You would have a hard time getting someone in the door now if you don’t have one of the coding agents like GitHub Copilot or Cursor.” But for site operations, “the push and pull is different—we still need those human touchpoints.”
Versin added that hiring now favors candidates “used to working with automations… because you can see resistance to letting automation do its work.” Familiarity with CRM systems and AI-assisted workflows is becoming a baseline expectation.
11. Integration and the Orchestration Layer
Another question concerned how firms are connecting multiple AI systems.
Chordia said Cortland relies on out-of-the-box orchestration tools provided by major cloud platforms: “They’re building AI-agent factories where you can orchestrate your agents and do setup and configuration in one place… We’ll rely on the PhDs doing the crazy stuff and apply it to our use cases.”
Shah emphasized Divco’s long-term data-layer investment:
“We’ve been building for eight years. Initially SQL Server, then Azure. We have a clean data repository. For orchestration we’re big into LangChain and LangGraph.”
That foundation allows the firm to connect proprietary agents with external models while maintaining governance and observability—essential for institutional investors handling sensitive data.
12. The Next Two to Three Years
Looking ahead, panelists forecasted several high-potential frontiers:
a. Customer insights and persona simulation.
Chordia cited research showing that AI personas can replicate survey results with “85 percent accuracy to what the actual human says.” He predicted rapid adoption of virtual focus groups that replace slow and expensive resident surveys.
b. Underwriting assistance and high-throughput deal review.
Shah expects agentic systems to “fast-track or just make the underwriting process quicker than it is today,” especially in credit origination where only “3 percent of the loans you look at actually fund.” Automating early-stage screening could multiply throughput without increasing staff.
c. Online-reputation intelligence.
Versin hopes for tools that go beyond responding to reviews: “Analyzing sentiment, forecasting our reputation based on P&L indicators, and helping marketing be more proactive.”
d. Cross-department orchestration.
Baum imagines agents coordinating multi-team processes such as property onboarding or marketing optimization: “An agent that identifies a low conversion rate, changes messaging, maybe even pricing and ad copy.”
Her broader prediction: “A couple years ago, AI conversations were about automation. Now we’re talking about agentic AI. In two or three years we’ll be back saying, here’s how we’ve implemented it and how it’s made our organization better.”
e. Continued value from explainable machine learning.
Chordia cautioned against abandoning traditional analytics: “There’s a lot you can do now with machine learning where agentic AI isn’t maybe there or you don’t want the black box… You can unlock value with more explainability.”
13. Conclusion: Toward an Augmented Organization
Across leasing desks, training simulators, and investment pipelines, agentic AI is reshaping how real-estate companies operate. The panel’s consensus was pragmatic: human oversight remains critical, but every repetitive or low-judgment workflow is fair game for automation.
Key themes emerged:
- Integration first. AI’s value depends on clean, connected data.
- Human in the loop. Empathy and accountability remain differentiators.
- Go big, not small. Broader workflows yield better ROI than isolated tasks.
- Build for differentiation. Buy for speed.
- Leadership mandate. Adoption rises when executives set clear goals and budgets.
In moderator Brad Hargreaves’s closing words, this conversation marked “the first of many” exploring how real-estate operators can harness truly autonomous systems. Agentic AI may still be nascent, but as Versin put it, the ambition is clear: “One platform, one solution, one brain behind it all.”
See the full recording of the call here.
