Case study · HFHL · Mortgage broking · Melbourne, Australia

From a tangle of Excel files to a 20-broker loan pipeline.

HFHL was running the whole brokerage out of spreadsheets — a master pipeline, per-broker trackers, and a shared drive of PDF payslips. We shipped a single web app that holds every application end-to-end, with AI doing the heavy lifting on documents.

live · 202520+ staff usersAI document extractionEnd-to-end pipeline

The problem

Excel was holding the brokerage together.

  • The whole loan pipeline lived in Excel — a 'master' tracker, per-broker copies, plus a shared drive of PDF payslips and bank statements.
  • Version conflicts were daily: two brokers updating the same row, applications dropping off the radar between stages, no audit trail of who touched what.
  • Processors were re-keying borrower data from PDFs into the tracker. Bottlenecks (sitting in Conditional Approval for two weeks) were invisible until clients chased.

What we shipped

A single web app that runs the entire pipeline.

  1. 01End-to-end loan application management — one record per application, moving through a stage-gated pipeline from new lead to settled and post-settlement.
  2. 02AI document hub: borrowers upload payslips, bank statements and ID; OCR extracts the fields and pre-fills the file. Brokers verify rather than re-type.
  3. 03Live kanban + manager dashboards: every staff member sees their queue, every manager sees throughput and where deals are stuck.
  4. 04Stage-driven task templates: each stage spawns the right checklist (doc requests, lender submission, conditional follow-ups) so nothing gets missed.
  5. 05Audit log + role-based access — every state change is timestamped, every doc upload attributed, ready for compliance review.

Digital transformation benefits

Why moving off Excel changed the day-to-day.

Single source of truth

One application record. No more reconciling a broker copy against the master tracker — every keystroke lands in the same database.

AI-extracted client documents

Payslips, bank statements and IDs are OCR-parsed on upload. Borrower name, employer, income figures, account balances — auto-extracted, broker confirms.

Stage-gated pipeline with audit trail

New Lead → Pre-approval → Submitted → Conditional → Unconditional → Settlement → Post-settlement. Every transition is timestamped, attributed, and reversible.

Real-time manager visibility

Pipeline value, conversion-by-stage, age-in-stage, broker load — dashboards that read off the live database, not a Monday-morning Excel pull.

Bottleneck + SLA detection

Deals stuck in a stage past their SLA surface automatically. Managers chase the right bottleneck, not the noisiest broker.

Compliance-ready logs

Every doc upload, every state change, every note — captured against a user, a timestamp, an application. ASIC/NCCP-friendly out of the box.

Faster staff onboarding

New brokers learn the process by using the tool. The workflow is the SOP — no separate manual to keep in sync.

Reduced manual error

AI extraction + duplicate detection + required-field validation catches the data-entry mistakes that used to surface at the lender, not the desk.

The outcome

Live with 20+ staff on the platform.

  • 20+ staff active on the platform — brokers, processors, settlement officers and managers.
  • Excel pipeline fully retired; single source of truth for every application HFHL handles.
  • AI OCR removes the bulk of manual data entry — borrower files are populated from documents in seconds, not minutes per application.
  • Manager visibility moved from 'ask the broker' to a live dashboard showing pipeline value, conversion by stage, and stuck deals.

Where AI shows up

Not a sidebar — woven through the workflow.

Lead intake

Inbound referral emails / PDF lead forms parsed automatically into application fields — broker doesn't re-key the borrower's name, contact, loan amount.

Document hub

OCR + entity extraction on payslips (gross/net, frequency, employer), bank statements (BNPL, recurring obligations, end balance) and IDs. Auto-fills serviceability inputs.

Credit + serviceability

AI risk-flag pass — surfaces undeclared liabilities, irregular income patterns, and prompts the right lender shortlist with a rationale the broker can edit.

Pipeline

Next-action suggestions per card ('no movement 7 days — chase Doc X', 'lender expected to respond today').

Settlement

AI-drafted client comms — conditional-approval letter, settlement checklist, post-settlement follow-up cadence.

Stack

Hosted on Cloudflare, AI-augmented end-to-end.

Next.js + TypeScriptWeb app frontend
Tailwind CSSDesign system
Cloudflare Pages + D1Hosting + database
AI OCR + LLMDocument extraction + drafting
CF AccessStaff authentication

Interactive walkthrough

Open the interactive prototype

Six screens, one click-through. Each screen explains who's acting, what the AI does, and how the design reflects HFHL's workflow.

Open prototype →