AI Agents

Context

Built for banking within the Retail mortgage underwriting stack, this agent is used early in the file to confirm that the property being financed is correctly identified, owned, and free of undisclosed encumbrances. It runs when the purchase file or refinance pack includes registry extracts and deeds, so underwriting proceeds on verified parcel, ownership, and lien information.

What it does

The agent reads land-registry extracts, title deeds, cadastral/parcel records, and any attached searches, then validates ownership, legal description, and encumbrances against the application and appraisal. It reconciles names (seller/borrower vs. registered owner), legal descriptions and parcel IDs, easements and rights-of-way, mortgages/charges and their priority, restrictions, and pending actions. Discrepancies—name mismatches, parcel/lot inconsistencies, unrecorded liens, expired or missing pages—are flagged with a concise rationale and links to the exact clause, line, or folio so reviewers can confirm quickly.

Core AI functions.

Document classification for deeds, registry folios, charges, easements, and maps; OCR and field extraction for owner names, parcel/lot/folio numbers, legal descriptions, charge numbers, amounts, priorities, and dates; normalization across formats and jurisdictions; cross-document checks against the loan application and appraisal; exception logic for missing annexes or map/description conflicts; and reason-code generation with page-level lineage to the source.

Problem solved

Manual title checks are slow and error-prone. Inconsistent descriptions, undisclosed encumbrances, or stale searches are discovered late, triggering rework, delays to clear-to-close, or post-funding risk.

Business impact

Files move faster and safer to clear-to-close: ownership and parcel identity are evidenced, encumbrances and priorities are transparent, and ask-backs are precise. That reduces rework, shortens cycle time, protects collateral quality, and improves audit readiness through clause-level evidence.

Integration and adjacent use cases

Integration complexity: Medium. The agent ingests registry/deed PDFs and searches from the broker/borrower portal, email, or DMS and writes validated ownership, parcel, and encumbrance findings (with evidence links) into the underwriting workbench; core systems are unchanged.

It is commonly combined with:

• Purchase agreement & contract validator (to align parties, property, and dates),

• Appraisal & valuer/license/method validator (to reconcile legal description and value basis),

• EPC validity checker (where energy certificates are required),

• Income & employment verifier and Self-employed income analyzer (for borrower income evidence),

• Debt & liability cross-checker (to align bureau and declared debts),

• Collateral insurance & pledge verifier (to confirm coverage and loss-payee wording), and

• Final completeness checker (to catch missing pages or expired searches before sign-off).