// 01
Borrower chatbots and lead-gen agents
The tools most 2026 roundups lead with. They answer prospect questions around the clock, qualify and route. They touch enquiry data, not the case file.
Most AI built for mortgage brokers is built to win the lead. Almost none of it is built for what happens after — the case file full of SSNs, pay stubs, bank statements and tax returns sitting on your side of the deal.
The category is not one thing. Sorting it by whose data it touches is the only sort that matters once the file is real.
// 01
The tools most 2026 roundups lead with. They answer prospect questions around the clock, qualify and route. They touch enquiry data, not the case file.
// 02
These speed up document review and flag missing conditions. They touch the case file, and they are cloud services.
// 03
Built for the lender receiving the file, not the broker assembling it.
// 04
Where the application lives once the borrower commits.
// 05
On nobody’s tool list, and the ones the trade press is warning brokers away from. This is the category’s real security exposure.
// 06
The one row here where the borrower’s documents never leave the brokerage’s own infrastructure.
A mortgage case file is not a CRM record. It is a stack of the most regulated personal data a private business ever handles: Social Security numbers, pay stubs, tax returns, bank statements, asset verifications. It is the same category of sensitivity as a family office’s private client documents — different vertical, identical exposure.
"If you wouldn't want it public, don't put it into AI."
And yet brokers are using AI anyway: in A&D Mortgage’s 2026 Broker Survey, 55% said they use AI daily or regularly and 72% expect their use to grow significantly within three years. That gap — between the guidance and the practice — is what this page exists to close.
Not safely, and not quietly. Pasting a borrower’s file into a consumer assistant moves regulated personal data to a provider the brokerage has no contract with — and the obligation does not travel with it.
It means case-file intelligence that stays in the building. This is not a chatbot for borrower intake — it is document intelligence for the broker agency’s own case files, running on the agency’s own infrastructure.
// 01
OCR and VLM document extraction over the documents already uploaded to the platform — classify, extract financials and dates, summarize.
// 02
Income and asset figures come out of the pay stubs, statements and returns themselves rather than out of a re-keyed spreadsheet.
// 03
The obligation-and-date surface, pointed at loan conditions across live cases instead of fund covenants.
// 04
Plain-language questions over every file the agency holds, grounded in the agency’s own documents with citations back to the source page.
// 05
RBAC with fine-grained document and skill access control, and per-organisation isolation.
// 06
Audit logging and OpenTelemetry traces — the evidence a Safeguards Rule conversation needs.
// 07
Private cloud or on-premise. Nothing is sent to a third-party model provider.
Four questions, asked of every bucket above. The last column is the only one where the answer to the second question is not somebody else’s data centre.
| Borrower chatbots | Cloud pre-underwriting | Consumer AI | Super Chain | |
|---|---|---|---|---|
| Touches the case file | No | Yes | Yes, unofficially | Yes |
| Where the data goes | Vendor cloud | Vendor cloud | A public model provider | Stays on your infrastructure |
| Who carries the liability | You | You | You | You — with nothing to hand over |
| Built for | Conversion | Speed to close | Nothing in particular | Confidentiality |
Industry guidance says no. Mortgage Brain warned brokers plainly: "If you wouldn't want it public, don't put it into AI." And under the FTC Safeguards Rule a broker must select service providers capable of safeguarding customer information, require those safeguards by contract and reassess them — so the broker stays responsible for how a provider handles borrower data.
Lead-gen AI touches enquiry data and is built to convert. Case-file AI reads the loan file itself — income documents, bank statements, tax returns — which is regulated personal data. They are different risk categories and should be evaluated differently.
No. It runs on the broker agency’s own infrastructure — private cloud or on-premise — with per-org isolation. Nothing leaves the room.
Extract income and asset figures from uploaded documents, track conditions and key dates, organise the case file, and answer plain-language questions across every file the agency holds, with citations back to the source document.
No. There is no borrower-facing surface. It is internal document intelligence for the brokerage’s own case files.
A first conversation, in person where we can. We listen to how the house works before we ever speak about the platform.