Shawn Malhotra talks Rocket’s vision for AI-powered, people-centered mortgages

Some results are already showing. In home search, AI models are predicting listing prices with less than a 2% error rate. For purchase loans, Rocket customers using AI-driven chats convert at three times the rate of other channels, and 2.5 times for refinances — helped by the fact that 80% of customers prefer chat.

On the origination side, Rocket’s production teams are handling 50% more clients per person year over year. In servicing, AI assists customers with account questions and identifies refinancing opportunities when rates change.

Rocket’s data footprint will grow with Redfin and Mr. Cooper, giving its AI tools — including the proactive Authentic AI system and the emerging Model Context Protocol (MCP) — a stronger base. It will put the company in a “good position,” Malhotra said. 

Launch and learn

Malhotra said Rocket’s philosophy is to deploy early and refine in the wild, rather than wait for a perfect product — especially as AI innovation accelerates.

“Every day you’re waiting on the sideline is a day you’re not bringing value — and importantly, others in your industry are getting that value,” Malhotra said. “You can’t build a perfect product in planning; you build the perfect product by getting something out, listening to your clients, iterating all that feedback.”

As an example, Rocket recently rolled out an AI-texting tool for bankers. There was initial concern about whether messages would be sent automatically without oversight, so Rocket built it in a review loop — bankers can see every message, confirm the recipient and retain full control.

According to Malhotra, some messages are excellent and some are flawed, but the system learns from feedback and improves with each iteration. If Rocket had waited to launch until the tool consistently matched or exceeded a banker’s quality, they’d still be waiting. By releasing early, the company has already seen a 20% increase in client outreach from bankers compared to previous years, he added. 

Rocket has also experienced a “huge democratization” by allowing its employees to write code and develop apps. All activity is secure and private, with thousands of weekly active users across the company — including underwriters, HR staff, finance teams and bankers — to create new innovations.

Balancing human and AI touch

In the battle between human and AI interaction, Malhotra believes customers want AI to reduce redundancies, handle less emotional tasks and save time. But some parts of the process still require a human touch, since the homebuying journey is not like shopping on Amazon — it’s a significant life event.

“There’s a human touch that some of them really need, because when it’s your first home purchase — or just any home purchase — I want to talk to someone at some point. You don’t want to throw it over with paperwork,” he said. 

“This past quarter, we did our first end-to-end refi where the banker was basically just presenting the final qualifications and the final package to the client, but they completed everything through our chat. It took them 30 minutes to get to that point, and we have the goal to make that 10 minutes.”

The benefit of having Mr. Cooper and Redfin, Malhotra added, is that Rocket can use customers’ past interactions with those companies to deliver more personalized experiences.

“What’s nice about having that end-to-end view is, you’re the same person coming through the system. Depending on how you behave during home search and how you interact with your agent, that might tell me what kind of experience you want when it comes time to refinance your mortgage. I don’t have to relearn that now.”