Independent mortgage banks spent $10,936 to originate a loan in Q2 2026 and kept $973 of profit. The cost problem is 11x the prize. That makes this a CFO conversation, not an innovation conversation.
Four segments, entered in order. Home equity is the beachhead because the proof already lives there. Independent mortgage banks are the expansion. Personal loans are the volume play. Servicing comes last, through accounts already won.
HMDA filings and NMLS registrations name every lender, their volume, and their footprint. Five channels, ranked by yield: the reference snowball first, then the trade show floor, where the live demo does the selling.
Every lending buyer has the same three fears: control, compliance, and cost. The handoff, the audit trail, and outcome pricing answer them in order. We don't sell autonomy. We sell work that keeps moving safely.
A named list of 200 accounts in week one. First pilots scoped inside 45 days, each on a scorecard agreed before it starts. The playbook written down as I go, so hire number two ramps in weeks.
It's outside-in, built from your site, our call, and public lending data. Mark it up and tell me where I'm wrong, and I'll tighten it. That's how I'd treat a real territory plan too.
The thesis in one paragraph: the pitch isn't AI phone calls. It's agents that own the workflow end to end. Take the application, chase and verify documents, work with underwriters, and get paid per lead or per loan file instead of per minute or per seat. The pricing model is the wedge because it moves the risk to us.
Brian Mays, Oakland. Built with my own agents.
The beachhead, because the proof already lives here. This corner of lending is a small town: the founders and operating execs know each other, sit on the same panels, and poach each other's people. The play is a reference snowball, not net-new selling. Every live account should produce two warm intros and one referenceable number.
Buyer: COO, Head of RevenueMid-size shops doing roughly 500 to 5,000 loans a year. Big enough to feel the cost problem, small enough to buy without a 12-month procurement cycle. Their math is public and brutal: about $11K to make a loan, under $1K kept. Document chase and application completion eat their processors alive.
Buyer: Head of Originations, COO, Head of OperationsHigh application volume, thin margins, and speed-to-lead decides who wins the loan. Outcome pricing fits this segment best of all: pay per qualified lead or per funded loan.
Buyer: COO, Head of OperationsPayment reminders, collections, loss mitigation across the first three segments. Servicing buyers need the compliance story proven first, so this is the expansion motion, not the entry. It is also the stickiest revenue: servicing runs forever.
Buyer: VP of ServicingLending rewards social proof more than almost any industry. Every live customer gets asked for two intros at the moment of their first win, not at renewal. The intros travel sideways, not head to head: an ex-colleague now at a mortgage bank, their subservicer, a founder from the same YC batch, a peer in a different segment. Nobody hands you their direct rival, and the plan doesn't need them to. Where it is head to head, the ask is a reference call, which buyers take because intel flows both ways.
MBA Annual and the regional shows, HousingWire's events, Fintech Nexus, and a seat in the MISMO working groups. The buyers there are actively shopping for tech, and nothing in the hall demos like this product does live: an agent takes a real call, chases a real document, and hands off mid-conversation. Pre-book meetings before the show, run the live demo all day, and skip the booth until a show has paid for itself once. Warm top of funnel at the exact moment the buyer is looking. The events ladder below sets the spending order.
The MISMO and handoff posts are already doing top-of-funnel work. What's missing is the harvest: every comment and DM from a lender title becomes an outreach within 24 hours. That's an AE job, and it's free pipeline.
HMDA filings and NMLS registrations give us every lender, their volume, and their state footprint. Build the 200-account list by segment, find the four buyer titles, and run outbound where Feather's own agents do the chasing. Selling Feather using Feather is the demo.
Loan origination system marketplaces (ICE Encompass, MeridianLink), point-of-sale vendors, subservicers, and the advisory firms lenders actually listen to (STRATMOR, Cornerstone). Not a launch channel, but it compounds.
MBA Annual runs October 11 to 14 in Chicago. If I'm in seat, I'm there on a floor pass with meetings booked before the doors open and the live demo on a laptop, no booth. Then the cheap circuit: state MBA association conventions (Texas, California, the regionals), where the buyers are less picked over and a pass costs a fraction of the national shows, plus a seat in the MISMO working groups. And the cheapest event of all: a co-hosted webinar with a live customer, which costs nothing and leaves a recorded demo we reuse in every sequence.
Cost: floor passes and gas moneyMBA's Independent Mortgage Bankers conference in early February is segment two in one hotel: the 500-to-5,000-loan shops the math points at. HousingWire's events and Fintech Nexus cover segments one and three. Still no booth: a suite, pre-booked calendars, and the agent demoing live. By now the demo has customer numbers from the first pilots in it.
Cost: mid, funded by pipelineOnce revenue is recognized, the big show gets the real treatment: a booth where the agent takes live calls on the floor all day. Nothing in that hall will demo like an agent chasing a real document in front of the buyer. The small shows prove the math; the big show scales it.
Cost: booth money, paid by closed revenueEvery account enters with a reason: HMDA volume in the ICP band, a job posting for processors or loss-mit staff, a rate event, a new state license, a comment on one of the founders' posts. Four titles mapped before the first touch: Head of Originations, COO, Head of Operations, and whoever owns servicing.
Exit: a reply, a referral, or 21 days of silenceOne workflow, quantified on the call: files per month, processor headcount, hours lost to document chase, current speed to lead. I do the math live against the MBA benchmark so the gap has a dollar sign before we hang up.
Exit: one workflow and one number agreed in writingNot slides. The agent calls a staged borrower matching their loan type, chases a missing document, verifies it, and escalates mid-conversation with full context to a human. The buyer watches the handoff, because the handoff is what they are afraid of.
Exit: pilot scorecard drafted, dates on the calendarScorecard agreed before anything starts: files touched, docs collected, contact rates, escalations handled cleanly. Weekly check-ins, one-page mutual action plan, and a reference call with a peer lender in week two. The CFO sees per-file math, not per-seat pricing.
Exit: scorecard hit and signed off, or we walkOutcome-priced contract on the piloted workflow. At the first win, two things happen the same week: the intro ask (two names, sideways) and the servicing conversation gets scheduled for next quarter.
Exit: signed, referenced, expansion scopedTwelve touches over three weeks, run on two titles per account, staggered. Feather's own agents do the sequencing and the chasing. I do the conversations.
[Name], MBA's Q2 numbers: independent lenders spent $10,936 to originate a loan and kept $973. Your HMDA filings put [lender] around [volume] loans last year, so call it [$X]M of origination cost, most of it people chasing documents and playing phone tag with borrowers.
Feather's agents take the file end to end: application in, documents chased and verified, underwriter handoffs, nights and weekends included. You pay per funded file, not per seat. If files don't move, you don't pay.
Worth 20 minutes to see it run on one of your workflows?
[Name], Unison had seven people dialing borrowers and still couldn't keep up with follow-up. Feather's agents run that outbound now with instant follow-up, and the humans work the conversations that need judgment.
Same shape as your funnel: high volume, and every minute of delay is a borrower answering someone else's call.
20 minutes this week? I'll bring your numbers, not a deck.
[Name], the fair question about any AI vendor is how long until it actually works. Nada went live with Feather in two weeks and the agents handled 5,000 calls in month one.
The pilot comes with a scorecard we agree on before anything starts: files touched, docs collected, contact rates, clean escalations. If the scorecard isn't hit, you've spent 30 days and nothing else.
[Name], three emails and four calls says the timing is wrong, so I'll stop here.
Two things first: when MBA's next quarterly cost data drops I'll send you the [state] cut, no ask attached. And if document chase gets painful before then, this stays a 20-minute conversation.
Good luck with the quarter.
Fire it: COO cold email, any capacity objection, and as the peer reference call for home equity prospects.
Fire it: implementation-risk objections, pilot scoping, and every small team that says they don't have bandwidth to onboard a vendor.
Fire it: VP of Servicing conversations, compliance review, and the segment-four expansion pitch inside won accounts.
Fire it: every opener, recomputed per account from HMDA volume so it reads as their number, not an industry statistic.
Become a lead on their own website and time the response, truthfully. Then the note: I filled out your form Tuesday at 2pm, the first human reached me Thursday morning, and Feather's agent would have called in about 90 seconds. Their own funnel is the coldest opener in lending, and every word is verifiable.
No meeting ask. Reply CALL and the agent phones the buyer inside a minute and qualifies me for them, not the other way around. The demo is the speed itself, and it costs them 90 seconds instead of 30 minutes.
A one-page cost-per-funded-loan benchmark built from their public HMDA filings against their state percentile, printed, hand-signed, overnighted. Executives open FedEx envelopes. They do not open vendor email number nine.
The agent narrates itself chasing a document on a file shaped like theirs, their loan type, their state, sent as a LinkedIn DM. Takes minutes to produce with the stack I already run.
Eight seats at every regional show, one live customer as the anchor guest, no pitch until dessert. Costs a dinner, converts like a reference call, and fills the next morning's calendar.
You spend about eleven thousand dollars to originate a loan and keep less than a thousand. Our agents take the application, chase the documents, and keep files moving nights and weekends. You pay per loan file, not per seat. If files don't move, you don't pay.
A seven-person dialing team's output, automated, with instant follow-up. Whoever calls the borrower back first wins the loan.
Every interaction permissioned, logged, and auditable. The agent knows exactly what it's allowed to say and when to hand off. Priced per outcome, not per attempt.
200 accounts across segments one and two, built from HMDA and NMLS data. Four buyer titles mapped per account.
Every warm signal from founder content and the customer network worked within 24 hours, alongside cold outbound into Heads of Originations and COOs. Every live customer asked for two intros.
One workflow, one number attached. A 30-day measured pilot with the scorecard agreed before we start: files touched, docs collected, contact rates, escalations handled cleanly. Mutual action plan on one page. A reference call with a peer lender, not a PDF.
Pipeline created, pilots scoped, the first show worked demo-first with meetings booked in advance, and the playbook written down as I go so hire number two ramps in weeks, not quarters. I've done that once: my playbook at Vori onboarded reps as the team grew to 12 while I stayed the top producer.
Where have pricing bands landed per outcome so far, and which segment's cycle has been shortest?
What killed the deals you lost: budget, compliance review, or a champion who couldn't sell it internally?
Which loan origination systems do we integrate with today? That decides how much of segment two is reachable now.
What proof does the servicing buyer need that the origination buyer didn't?
Every recorded sales call, won and lost, and I'll write the objection map from them. The last ten dead deals and what killed each. Two live customer scorecards with real numbers. An hour listening to production agent calls, because I sell what I've heard.
Demo environment and agent-config training deep enough that I run demos without pulling an engineer. The LOS integration map, live versus roadmap, because it decides how much of segment two is sellable today.
Pricing bands per outcome and who approves exceptions. Which customers I can name. The compliance envelope: what the agent may say to a borrower and where the hard lines are, so I never learn a rule from a prospect.
I know real estate from years working in it; I haven't carried a lending bag. The study is mine to do: origination flow end to end, non-QM versus agency, TRID timelines, MISMO vocabulary. What I'd ask of you is one subject-matter hour a week for the first month to correct what I get wrong.
Thirty minutes weekly with a founder for the first 60 days, deal reviews in whatever format you run, and an agreed scorecard so we both know by day 45 whether this is working. I'd rather be measured tightly than trusted vaguely.
I was sales hire number one at Vori, a venture-backed operating system for grocery stores. No playbook, no brand, no SDRs. I built named account lists, made the calls, got on planes, and closed a deal my first day in the field. Year one: 128% of a $1M+ quota, number one producer, promoted three times in under two years.
The deal I'm known for closed while I was on my honeymoon on Lake Como: a $400K+ contract with a 30-location grocery network when our biggest account was 7 locations. It closed because the process carried the context without me in the room. That is Feather's product thesis, lived, before I ever heard the company's name.
Before that, Regional Sales Director at Eagle Eye. Five months in I was part of the Albertsons close, the company's largest deal in North America. Oracle and SAP integrations, IT and compliance at the table. I've run the committee gauntlet at billion dollar companies, which is where lending deals go to live or die.
Since June I run Mays Growth Partners, where the entire outbound engine is AI agents I built myself: sourcing, enrichment, sequencing, follow-up. I do the closing. I'm a one-person version of what Feather sells, and I'd rather sell the real thing. Years in real estate before all of this mean borrowers, closings, and files are not a foreign country. And I've sat on the owner's side of the table: founded a consumer brand, ran it to about $500K a year, landed a Rite Aid pilot.
The pattern across all of it: hand me an empty territory and I come back with a machine other people can run.
Live documents, not attachments. Mark any of them up and the next version comes back tighter.
The strategy layer: segments in order, channels ranked, the pitch by buyer, the motion.
Funnel math, thirteen weeks day by day, the words I'd use, the full follow-up system, and what I'd need from you.
Companies by segment with buyer titles and the workflow wedge for each.
Named decision makers with warm paths and a first play each, checked against your customer list before any touch.