Sumanth Kamath — chief technology officer at eXp Realty — took the stage Tuesday in Dallas at HousingWire’s AI Summit to share how his company is using artificial intelligence (AI) to drive efficiency, empower employees and scale operations.
In a conversation with HousingWire Senior Director of Data and Content Tracey Velt, Kamath described a sweeping AI strategy that includes everything from custom ChatGPT assistants and digital “twins” to document automation, voice-activated agent tools and sentiment analysis.
ChatGPT as part of the culture
Kamath said eXp had distributed 1,000 ChatGPT licenses to staff before many companies had moved past curiosity.
“I wasn’t surprised on the adoption side as much, but I was more surprised on the innovation and the creativity that came out of it,” he said. “People started out with search, basically a Google replacement, and then soon enough, the light bulbs went off.”
Staff began uploading standard operating procedures and knowledge bases into GPTs to create specialized assistants for their teams.
“We had project managers [who] all having custom GPTs that people would go to, rather than really long documents,” Kamath said. “So today, as we speak, we have 1000-plus custom GPTs.
The takeaway, he said, was that AI wasn’t an intimidating “big monster.”
“It’s like an iPhone,” Kamath said. “You put it in front of people, people will pick it up and translate it. The more you make it a part of your workflow, the more people will use it.”
Digital twins for the C-suite
One of eXp’s boldest AI experiments was building “digital twins” of its executives, including Kamath.
“How do you make a leader be available to the lowest of the employee levels, where they can bounce ideas by a leader like they were talking to them, and at the same time, create space on the leaders calendars?” he asked. “[So] they’re not running from 30 minute back-to-backs, and they can focus on strategic items and innovation and such?”
Executives created a “manual of me” that outlines decision-making styles, preferences and dislikes. The company recorded meetings to capture leaders’ behaviors — then worked to fine-tune responses, Kamath said.
“My calendar specifically is now more open, and there are so many meetings where people will start by saying, ‘Have you already run this by your digital twin?’” he told the crowd.
Empowering ‘citizen developers’
Kamath described a “citizen development” approach in which frontline employees build their own AI-powered tools.
“The person who is facing the problem is best suited to solve that problem, rather than that person documenting the problem and sending it over to the development team,” he said.
With AI-assisted “prompt coding,” staff outside of engineering have created fully functioning products. One example: The company’s new performance management system that was built in-house by HR.
Automating document review
With more than 1.1 million documents processed since launch, eXp’s in-house “Doc AI” system has transformed compliance and transaction management.
“We leveraged what OpenAI gives to us in terms of extracting data from documents, and now we intercept documents as soon as the agent loads it,” Kamath told attendees. “Then, we extract information from that document, and then we have a UX breakout wherein we now say our confidence is at 95% of how accurate what we get from this document is. The human is still in the middle.”
The system extracts data, assigns a confidence score, and still involves licensed brokers for oversight — but productivity has increased “multi fold,” he added.
To serve its 83,000 agents, eXp built a real-time, AI-powered query system that is now voice-enabled.
“Imagine an agent who’s driving from one showing to the next and they want information,” Kamath said. “They have no time to go and click on a system. They can talk to a system, and we respond, so that was very exciting.
“When you take a concept to scale, it has to be right every single time,” Kamath said. “We’ve learned a lot along the way.”
Replacing surveys with sentiment analysis
Instead of relying solely on surveys, eXp now uses AI to analyze call transcripts from its help desk to determine the sentiment of the call to determine which agents should get a phone call for more help.
“What that did for us, from an operational perspective, is it saved time,” said Kamath. “Now we could engage supervisors who would call the agent immediately to say, ‘I know you were not happy about this. Let me help you.’”
From early custom GPT experiments to scaling AI for tens of thousands of agents, Kamath emphasized that the key to success was giving employees the tools — and the trust — to innovate.
“We are an innovation-first company. We are not risk averse at all,” he told the room. “We handed all of this to our frontline employees and several different innovations came through.”
English