The Agency is building a proprietary artificial intelligence platform that runs open-source large language models on infrastructure it controls, alongside commercial models from companies like OpenAI and Anthropic. The luxury brokerage says it will eventually put the system in front of its more than 4,000 agents worldwide.
The platform doesn’t have a name yet. Zane Burnett, The Agency’s EVP of digital strategy, isn’t in a hurry to pick a name, either.
“I feel like the act of naming this platform kind of dilutes how important it actually is,” Burnett told Inman in an exclusive interview on Friday. “The naming thing is just a little gimmicky to me.”
Instead, The Agency is exploring a short onboarding interview for each agent so the platform is tailored to how that person works. Agents can call it whatever they want.
Mauricio Umansky
According to the company, the system pulls data that currently lives in separate places into a single data lake that agents and leadership can query in plain language. Agents will get access to answers grounded in The Agency’s live data, not generic internet results. That data includes MLS and market data, transaction management, brokerage financials, listing syndication, marketing tools and internal communications, available through a single, secure and conversational interface.
“AI is the biggest shift our business will see in a generation, and we refuse to rent our future from someone else’s playbook,” Mauricio Umansky, The Agency’s founder and CEO, said in a statement. “We’re building our own platform, shaped by our culture and our data, so every Agency agent walks into every room with more insight than anyone else at the table.”
For now, it’s a work in progress. Burnett said the platform is in alpha testing with The Agency’s internal team. A beta with some leadership is scheduled in about a month, specifically with Umansky. Burnett said Umansky “has been a huge part of this and of driving what gets built.”
The Agency will formally unveil the AI platform at its leadership summit in Miami in November. All of The Agency’s leadership should have access in Q1 2027, with agents following later in the year.
“We want to be intentional about how we roll this out, so it’s intentionally phased,” Burnett told Inman. “Could we roll it out to everybody and their mother in Q1? Yes. But we don’t want to do that. We want to address hazards or issues as they come up so that by the time we reach our true end user, the agent, those bumps have been smoothed out.”
Luxury data and a hybrid model
The Agency’s press release calls this the first platform of its kind in residential real estate. That’s a big claim as Compass International Holdings rolls out its Home Platform to roughly 340,000 agents across the former Anywhere brands, and eXp and Real build out their own AI tools.
Real Brokerage recently used its RISE conference to launch Leo Pro, a library of more than three dozen AI agents inside its reZEN operating system, for 36,000-plus agents. Leo Pro takes a described task and handles it, checking in before major actions.
At eXpCon, eXp Realty recently rolled out “MCP First, AI of Choice,” which uses Model Context Protocol to connect agents’ existing AI tools, such as Claude and ChatGPT, to eXp business data instead of making them adopt a branded assistant.
eXp is still building its own technology, though. CEO Leo Pareja called MCP the first piece of a larger AGNT OS rebuild targeted for early 2027. eXp International is also rolling out Nexus, a single-login operating platform it says can cut a market’s software costs by up to 70 percent, which it eventually plans to sell to other brokerages.
Zane Burnett
Burnett acknowledged he doesn’t know the details of what competitors are building. He pointed to two things he thinks set The Agency apart.
The first is its customer base. The Agency’s global luxury footprint, combined with its media brand, gives it what Burnett described as a unique window into how ultra-high-net-worth consumers behave. He stressed that the company’s rights to that data are unambiguous.
The second is the architecture.
“I don’t know that I’ve ever read about any of them running a hybrid of self-hosted open-source LLMs alongside commercially available models,” Burnett said.
Keeping sensitive data in-house
The Agency’s AI platform setup is mostly about data governance. Under the luxury brokerage’s model, commercial models handle work like market analysis and forecasting. Sensitive client, transaction and financial data stays with models the company hosts itself.
Burnett said the uncertainty around the big AI frontier labs, particularly as OpenAI and Anthropic move toward IPOs, makes him wary of sending that data through their systems.
This October, Inman turns its focus to AI and its rapid rise in real estate. During Artificial Intelligence Month, we’re digging into the startups shaking things up and the established players folding AI into their offerings.
“We don’t necessarily want to run sensitive data through commercial models, because we have no idea what’s happening there,” he said. “So in a lot of ways, we’re creating our own doors and our own rooms within which that data lives.”
He wouldn’t say which open-source models The Agency is running. He cited DeepSeek as an example of a model a company can host and train on its own terms, but said he wasn’t confirming they use it.
Self-hosting gives a brokerage more control, but it’s expensive to build and maintain. Burnett declined to disclose how much The Agency has invested or how big the team is. He said a core group of developers works on the platform, and an outside development firm adds engineers as needed.
1 interface for every tool
Burnett framed the build-vs.-buy question around what he sees as the industry’s assembly problem.
“If there are five CRMs available, three transaction management solutions and so on, how different are you if all you’re doing is a variation on assembling the same products everybody else is using?” he said.
The Agency isn’t ditching vendors entirely. It uses Rechat today and will keep some existing tools. The goal is for agents to work through those tools without logging into them, for example by running comps in Rechat or pulling an office P&L from inside the platform.
Burnett compared the old approach, a dashboard of tiles agents single-sign-on into, to the knobs on a 20-year-old car stereo. The new approach is the controls on a modern steering wheel.
“Your hand never leaves the steering wheel, but you’re interacting with several different components,” he said. “We’re past dashboards.”
Making it sticky
Burnett said the business case for building its own AI platform comes down to recruiting and retention, a problem brokerages have been chasing for decades. Burnett said the platform isn’t meant to predict which agents are likely to leave. Its value is supposed to come from making agents’ work easier.
“Going back to the car stereo metaphor, it lets them keep their eyes on the road while doing five times what they normally could,” he said. “That truly answers the question of what makes this sticky.”
