A journey that leads without ever overtly leading.
Your borrower typed their address into a “what’s my home worth?” ad. They want to know their options, and they don’t want a phone call. We built the journey to feel like a conversation with a seasoned loan officer. It just happens to be software, so it’s patient, precise and identical every single time.
It runs on deterministic rails, and it isn’t a chatbot. We fixed the steps, every figure comes from one calculation engine, and the borrower can open the math and check it. No model decides anything a regulator would ask about, and a continuous integration (CI) gate fails our build if a model endpoint becomes reachable, by import, from any module that decides a regulated outcome. How we govern artificial intelligence (AI) →
Three journeys, one set of rails
Home equity
We meet the borrower in the home-equity mental model they arrived with. The journey pulls a live value estimate for their address, runs the honest side-by-side, and only reframes into a refinance where the math genuinely wins. Where keeping the low first mortgage wins, it says so and captures the lead for you.
Cash-out refinance
Debt consolidation with the arithmetic in view: current payments in, new payment out, cash in hand. A “see exactly how this math works” panel shows every step, including the cap rules, the winner rule and the assumptions we state.
Purchase pre-approval
We point the same rails at a purchase: program selection (conventional, Federal Housing Administration (FHA), and Department of Veterans Affairs (VA) loans), a realistic monthly payment of principal, interest, taxes and insurance (PITI), with mortgage insurance (MI) where it applies, and a pre-approval file your loan officers can pick up without re-keying a field.
What the borrower actually experiences
- Their real numbers, immediately. They see a live value estimate for their property on the first screen. A soft credit pull, with their explicit authorization and no score impact, finds their actual balances, and every slider on the workbench recomputes honestly.
- Consent as a first-class step. We capture e-consent, credit authorization and e-sign per person, including co-borrowers, who complete their own consent on their own device. We don’t pull credit or sign anything on someone’s behalf.
- Checkable math. The journey shows its work. A borrower who wants to verify the consolidation arithmetic can open the panel and follow every step, and that’s why they believe the result.
- A real finish line. The cash-out and purchase journeys end with generated disclosures delivered for e-signature and a complete application pushed to your loan origination system (LOS). There’s no “someone will call you” dead end. Until you connect your LOS sandbox, that push and delivery run simulated, and we label them that way. A borrower who wants a human at any point gets one, with their file intact.
We made the honesty rule structural. The recommendation engine can’t present a losing option as a winner, because the same deterministic math that renders the workbench decides the recommendation. That’s how the experience stays persuasive and defensible in front of a regulator.
Built for how your leads actually arrive
We built lead-partner intake as an adapter architecture. Signed-token redirects and webhook models normalize into one attribution spine, so every journey outcome ties back to its source. We built attribution to carry from first click through to the funded loan once you connect your LOS milestone feed, and that’s what makes the intelligence layer’s per-source economics possible.