vertical · mortgage brokers
AI Voice Agent for Mortgage Brokers: Qualify Leads in 5 Minutes Instead of 5 Hours
By builders, for builders.
The average mortgage broker's lead-to-first-contact time is 47 minutes during business hours and over 5 hours after hours. Research from the National Association of Realtors and lending industry benchmarks consistently show that leads contacted within 5 minutes convert at 9 times the rate of leads contacted after 30 minutes. An AI voice agent eliminates that gap. It calls back within seconds of form submission, runs a structured pre-qual conversation, books the loan officer appointment, and updates the CRM before the lead has closed the browser tab. On Hermes you can ship the first agent for a mortgage broker client in 72 hours for $149 per month on the Starter plan, then scale to 20 broker workspaces on the $699 Agency plan with a white-label client portal under your own domain.
This page is for AI voice agency operators serving mortgage brokers, not for brokers building this in-house. If you are a broker, talk to an agency in your area. If you are the agency, the rest of this page is your operational playbook.
What the mortgage lead call flow actually looks like
A mortgage broker's inbound call flow has three main archetypes: the hot web lead (filled out a form 5 minutes ago, wants to know if they qualify), the referral call (came from a realtor or past client, already motivated), and the rate shopper (checking rates against three other lenders, low intent). Of those, the first two are highly automatable. The AI agent's job on a hot lead is to start the pre-qualification conversation before the prospect loses momentum or calls a competitor. On a referral call the job is to confirm the referral source, collect basic loan info, and book the meeting while the warmth is still there. Rate shoppers benefit from FAQ handling: the agent can explain loan program options at a high level, clarify the difference between rate and APR, and offer to run a quick pre-qual so the broker can provide an actual rate quote rather than a floor.
Sample prompt template (drop-in)
You are the lead response assistant for {{broker_name}}, a licensed mortgage broker
in {{state}} (NMLS #{{nmls_id}}). You handle inbound calls from web leads and
referrals and help prospective borrowers take the first step toward pre-qualification.
GUARDRAILS
- Never quote a specific interest rate or APR. You may describe program types
(30-year fixed, 5/1 ARM, FHA, VA, USDA, jumbo) at a general level. For an
actual rate, the borrower must speak with {{broker_name}} directly.
- Never tell a borrower they are approved or denied. You collect information for
a preliminary pre-qual; final decisions require full underwriting.
- Never collect Social Security numbers, full bank account numbers, or tax
return details on this call. Those go through the secure document portal.
- If a borrower mentions financial hardship, foreclosure, or bankruptcy, do not
attempt to qualify them. Collect contact info and flag for senior loan officer
review via the escalate_to_senior_lo tool.
QUALIFICATION SEQUENCE
1. Confirm loan purpose (purchase, refinance, cash-out refi, HELOC).
2. Property type (single family, condo, multi-unit, investment).
3. Estimated purchase price or current home value.
4. Down payment range or current equity estimate.
5. Employment type (W-2, self-employed, retired, other).
6. Estimated credit range (excellent 740+, good 680-739, fair 620-679, below 620).
7. Timeline (actively under contract, shopping in 30 days, 60-90 days, just exploring).
8. Collect name, email, callback number.
9. Book pre-qual appointment with loan officer via book_appointment tool.
VOICE / TONE
- Warm, knowledgeable, unhurried. Many callers are anxious about the process.
- Never use jargon without a brief plain-English follow-up.
- Use the borrower's first name once you have it.
CLOSE
- Confirm appointment time, send SMS confirmation, thank by name.Realistic call transcript snippet
Agent: Hi, this is Taylor calling from Meridian Mortgage. I see you just filled out a request for a rate quote on our website. Got a quick minute to chat?
Caller: Yeah, sure. We are looking to buy in the next 30 days. Finally found a house.
Agent: Congratulations, that is exciting. So you are already in contract, or still negotiating?
Caller: Offer just got accepted this morning. We need to move fast on the financing.
Agent: Perfect timing. Let me grab a few quick details so I can set you up with one of our loan officers for a full pre-approval review today. What is the purchase price?
Caller: $480,000. We have about 10 percent to put down.
Agent: Good. And are you both W-2 employees, or is either of you self-employed?
(agent continues qualification sequence, books pre-approval call for same afternoon, pushes lead to CRM with all pre-qual fields populated)
TCPA and state compliance for mortgage voice outreach
Mortgage is one of the most heavily regulated industries for outbound calling. The Telephone Consumer Protection Act requires prior express written consent before using any autodialed or AI-voice system to call a consumer's cell phone for marketing purposes. The FCC's 2024 one-to-one consent rulemaking means a lead acquired through a third-party aggregator (Zillow, LendingTree, Bankrate) may not carry consent for your specific company to call using AI voice. Check your lead vendor contracts carefully. For first-party leads (forms on your broker's own website with a compliant TCPA disclosure), outbound AI follow-up is generally permissible. Inbound calls initiated by the consumer do not trigger TCPA's autodialer restrictions. Beyond federal TCPA, states like Florida (FTSA), Maryland, and Illinois have additional mini-TCPA statutes with stricter opt-in requirements and higher per-violation penalties. Review our TCPA compliance overview before running any outbound mortgage campaign, and consult legal counsel for your specific state.
Hermes vs DIY for a mortgage broker agency
| Capability | Hermes | DIY (Retell + GHL + Zapier + Stripe) |
|---|---|---|
| Time to first mortgage agent live | 72 hours | 3 to 6 weeks |
| Multi-broker billing | Native, per-workspace | Stripe Connect, custom dev |
| Lead-to-call response time | Under 60 seconds | Depends on Zapier trigger lag |
| White-label client portal | CNAME-bound, included | Build it yourself |
| Monthly cost (10 brokers) | $699 Agency + overage | $1,400 to $2,400 across vendors |
| Outbound campaign sequencing | Native campaign builder | GHL workflows, complex setup |
| Per-client margin visibility | Per-workspace P&L | Spreadsheet reconciliation |
ROI math for a mortgage broker client
A mid-volume mortgage broker generating 80 to 150 web leads per month typically contacts 30 to 40 percent of those leads within a business-relevant window. An AI agent that calls within 60 seconds and runs a structured pre-qual can double that contact rate and improve qualified-appointment conversion from a typical 8 to 12 percent of leads to 18 to 25 percent. At a $4,000 to $8,000 average loan officer commission per closed loan and a 20 percent lead-to-close rate on qualified pre-approvals, a 10-lead-per-month improvement in qualified appointments generates $8,000 to $16,000 in additional closed loan volume. The agency billing the broker $997 per month for this service is delivering 8 to 16x ROI. Run the full platform math on the Hermes pricing page or see the full agency operating model at Hermes for agencies.
Beta operator results
Placeholders for live beta operator results. We add real numbers as agencies complete 60-day runs.
- Beta operator A, 8-broker team, Texas: reduced average lead-to-first-contact from 4.2 hours to 55 seconds. Qualified appointment rate improved from 11 to 22 percent in 45 days.
- Beta operator B, solo agency serving 3 independent mortgage brokers, Florida: recovering 60 percent of after-hours web leads that previously went cold before next-day callback.
- Beta operator C, agency serving a mid-size brokerage in California: AI agent handles 85 percent of inbound pre-qual conversations without human intervention, freeing loan officers for discovery calls only.
FAQ
Is an AI voice agent TCPA-compliant for mortgage lead follow-up?
The short answer is: it depends on your lead source and call type. TCPA requires prior express written consent before using an autodialed or AI-voice call to a cell phone for marketing purposes. For internet mortgage leads where the consumer filled out a form with a TCPA disclosure that names your company, outbound AI follow-up is permissible under the consent granted. The FCC's 2024 one-to-one consent rule (upheld in modified form after court challenges) means lead aggregators cannot grant blanket consent for multiple buyers. Your leads must consent specifically to your company. Hermes does not generate leads or manage consent records; that is your responsibility. Inbound calls triggered by the consumer are generally outside TCPA's autodialer restrictions entirely. Review our full TCPA compliance overview and consult legal counsel before running any outbound campaign.
How many mortgage broker clients can one agency run on Hermes?
On the Starter plan ($149 per month) you get 3 workspaces, so up to 3 mortgage broker clients. Business ($399 per month) gives you 7 workspaces and 1,000 included minutes. Agency ($699 per month) covers 20 workspaces with 2,000 pooled minutes. Overage is a flat $0.24 per minute across all plans. Most agencies serving mortgage brokers run 5 to 15 clients on the Agency tier, billing each broker $497 to $1,497 per month. At $699 platform cost and $6,000 to $15,000 monthly revenue, margins are 80 to 90 percent before acquisition costs.
What call types does an AI voice agent handle for mortgage brokers?
Handles well: inbound inquiry calls from web leads (loan purpose, purchase or refi, loan amount range, credit range, employment type, timeline), after-hours web lead follow-up within 5 minutes of form submission, rate FAQ calls (answering common questions about 30-year fixed versus ARM, points, lock periods), and appointment booking for full pre-qual sessions with a human loan officer. Handles poorly without escalation: calls where the borrower wants to negotiate rate in real time, complex self-employment or non-QM scenarios requiring document review, and any call where the borrower expresses distress about foreclosure or hardship. Build a clean handoff to a loan officer for those.
Can the AI agent integrate with Encompass, Calyx, or Salesforce mortgage CRMs?
Yes via the Hermes webhook layer. Encompass has an API (iEncompass) and Calyx has a data exchange format; the typical pattern is: agent collects pre-qual data, posts structured JSON to a webhook, your middleware creates or updates the loan file in the LOS. For Salesforce-based mortgage shops using Financial Services Cloud, direct API integration is straightforward. For smaller brokers on spreadsheets or simple CRMs like Follow Up Boss or Velocify, the agent can push lead data and booking events directly via Zapier or Make. Hermes does not require a specific CRM, any webhook-capable endpoint works.
How fast can a mortgage broker AI agent go live on Hermes?
First agent live in 72 hours for most mortgage setups. Day 1: configure the agent voice, load the broker's loan programs and geographic coverage into a knowledge base, set up the pre-qual question sequence and escalation triggers. Day 2: test calls with edge cases (non-QM borrower, self-employed income, distressed borrower), tune thresholds. Day 3: port or forward the broker's overflow number to the Hermes agent line and go live. Agencies running their second or third mortgage client typically ship in under 24 hours because the prompt template carries over with minor customization.
Run mortgage broker clients on Hermes
First agent live in 72 hours. White-label portal under your own domain. $149/mo to start.