Sales outreach · Production

Better leverages ElevenAgents to build Betsy, reducing mortgage origination costs by 41% and doubling lead-to-lock conversion

The problem

Mortgage origination requires state-licensed consultants whose time is scarce and expensive, with most of that time consumed by repetitive qualification calls, outbound dialing, and follow-ups. Most financial institutions cannot deploy AI meaningfully because eligibility, pricing, and servicing data sit across fragmented LOS, CRM, and internal tools.

First attempt

Better's initial voice offering used a Speech to Speech model but lacked the naturalness, low latency, control, and reliability needed for high-volume financial services conversations.

Workflow diagram · grounded in source
1
Outbound or inbound call initiated
trigger
“Betsy placed 1.89 million outbound and inbound phone calls”
2
Betsy voice agent communicates
ai_action
“Better has programmed Betsy to verbally communicate with borrowers and streamline the mortgage application process through a voice agent”
3
Credit, pricing, rate lock by voice
ai_action
“Credit pulls, pricing, and rate locks can be initiated by voice through ElevenLabs, with Tinman executing each step and returning the results. Controls are explicit: pricing and eligibility are computed by Tinman and communicated by the …”
4
Compliance validation and consent
validation
“Regulated steps - including credit pulls, locking, authentication, and transfers to humans - require consent and follow state licensing requirements”
5
MCP orchestration executes Tinman tools
integration
“Orchestration runs inside Better's environment via an MCP-based multi-agent system connected to Tinman tools. Calls reach a single public ElevenAgents endpoint, while internal agent switching and tool execution happen within Better's stack.”
6
Route to licensed consultant
routing
“route qualified interest to state-licensed consultants”
Reported outcome

Betsy automates 35.5% of mortgage inquiries end to end, saves loan officers over 1,666 hours monthly, reduced average cost to originate by 41%, and doubled lead-to-lock conversion in 2025, while handling nearly 100,000 monthly calls.

Reported metrics
Monthly calls handlednearly 100,000
Monthly mortgage conversations (enterprise quote)over 100,000
Mortgage inquiries automated end to end35.5%
Outbound and inbound phone calls in 20251.89 million
Show all 9 reported metrics
monthly calls handlednearly 100,000
monthly mortgage conversations (enterprise quote)over 100,000
mortgage inquiries automated end to end35.5%
outbound and inbound phone calls in 20251.89 million
loan officer hours saved monthlyover 1,666 hours
average cost to originate41%
lead-to-lock conversion increase100%
lead-to-lock conversion (CEO quote)2x
total loan volume fundedover $110 billion
Reported stack
ElevenAgentsElevenLabsTinmanSpeech to TextLLMText to Speech
Source
https://elevenlabs.io/blog/better
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Betsy automates 35.5% of mortgage inquiries end to end, saves loan officers over 1,666 hours monthly, reduced average cost to originate by 41%, and doubled lead-to-lock conversion in 2025, while handling nearly 100,00…

What tools did this team use?

ElevenAgents, ElevenLabs, Tinman, Speech to Text, LLM, Text to Speech.

What results were reported?

Monthly calls handled: nearly 100,000; Monthly mortgage conversations (enterprise quote): over 100,000; Mortgage inquiries automated end to end: 35.5%; Outbound and inbound phone calls in 2025: 1.89 million (source-reported, not independently verified).

What failed first in this deployment?

Better's initial voice offering used a Speech to Speech model but lacked the naturalness, low latency, control, and reliability needed for high-volume financial services conversations.

How is this sales outreach AI workflow structured?

Outbound or inbound call initiated → Betsy voice agent communicates → Credit, pricing, rate lock by voice → Compliance validation and consent → MCP orchestration executes Tinman tools → Route to licensed consultant.