Five takeaways from AI Voice Agents and use cases for Mortgage Lending

Chris Barnett

Head of Product Solutions chris at marrlabs.com

April 16, 2025

Last week, Marr Labs CEO Dave Grannan joined Josh Reicher, Chief Digital Officer at Cenlar FSB , and moderator Ruth Lee for a deep dive into how AI voice technology is used today across the mortgage industry. This wasn’t another high-level discussion about chatbots or IVR. It was a practical look at how human-like voice agents solve real problems—from lead conversion to borrower support.

Here are some takeaways from the conversation: use cases, lessons learned, and where AI voice is headed.

‍
AI Voice helps lenders win the speed-to-lead game

‍
Dave Grannan opened with a stat that caught everyone’s attention: over half of borrowers go with the first lender who contacts them. That means speed is everything if you're buying leads from sources like LendingTree or NerdWallet.

AI voice agents can engage a lead within one second of submission, qualifying them, gathering details, and transferring to a licensed loan officer. The big unlock? You don’t need more staff to do it.

“You can turn on 1,000 AI agents right now and turn them off in an hour. That kind of flexibility is impossible with human teams.”

— Dave Grannan, Marr Labs CEO

‍

AI Voice Agents dont just answer calls—they have conversations

The webinar featured a demo of a Marr Labs AI Voice Agent qualifying a borrower for a home equity loan and then making small talk to keep the caller engaged until a live loan officer joined the call. Grannan noted that this warm handoff significantly reduces call abandonment.

‍

‍

AI voice agents are designed to sound natural, respond intelligently, and never get tired or frustrated, giving borrowers a more consistent experience than traditional call centers.

‍

AI Voice can fully handle routine inquiries

‍

Josh Reicher shared how Cenlar is using AI to improve operational service efficiency.

One use case: answering common borrower questions (like requests for a 1098 form) without human involvement. The key is matching the right level of complexity to the right tool.
‍

“Most digitally savvy homeowners prefer to self-serve. If they can get an answer in one call, 24/7, they’re happy.”

— Josh Reicher, Cenlar

‍

It knows when to escalate to a human

‍

Both Dave Grannan and Josh Reicher emphasized that AI voice should empower, not replace, human agents—especially in emotionally charged or complex situations. Today’s AI voice agents can recognize the right moment to get out of the way.

Here’s how that works:

Real-time sentiment analysis
AI monitors tone, pacing, and language patterns that indicate frustration, confusion, or anxiety. If it detects any of these, the system can proactively route the call to a human—without the borrower needing to ask.
Caller Intent or Preference If a borrower explicitly states, “I want to speak to someone” or employs common escalation triggers (e.g., repeating "agent"), the AI promptly honors the request. There is no struggle for control—only a seamless handoff.
Policy or Compliance Triggers
In the mortgage industry, specific topics (e.g., rates, credit decisions) necessitate human intervention. Marr’s AI agents are equipped with built-in guardrails that prompt escalation when these thresholds are reached.
Multimodal design philosophy
Grannan noted that some borrowers prefer chat, others want voice, and some want a person. The goal isn’t to push them down a path, but to match them with the right channel at the right time.

‍

“If someone gets annoyed or asks to speak to a human, we don’t fight it—we route the call. It’s about getting the task done, not pretending the AI can do everything.”

— Dave Grannan, Marr Labs CEO

‍

This hybrid approach ensures borrowers feel supported, while human reps focus where they’re most needed.

‍

There’s real ROI on the table

‍

Grannan shared results from a recent case study across 300,000 calls:

81% cost reduction per outbound call compared to humans
$39K saved in one month for a single lender
20% higher lead prequalification rate vs. human callers

Reicher echoed the value of starting small and measuring impact. “If the tool helps your team spend more time with borrowers and less time on busywork, it pays off fast,” he said.

‍

Where things are headed: Agentic AI and Task Completion

‍

Looking ahead, both speakers pointed to agentic AI as the next evolution—tools that can handle entire workflows with minimal supervision.

‍

“Voice AI is no longer about routing or menus. It’s about completing real tasks—accurately, efficiently, and with empathy. ”