About the client
Our client is a nationally recognised Australian mortgage and lending provider that helps everyday Australians achieve their property goals. With a large network of brokers across suburban and regional areas, the organisation combines national scale with genuine local insight.
The broker network is built around locally based small business owners who understand the needs of their communities and provide tailored advice throughout the home loan journey. As a licensed credit provider operating under the National Consumer Credit Protection Act 2009 (NCCP Act), the organisation maintains rigorous compliance with responsible lending obligations, privacy requirements, and anti-money laundering regulations — making operational accuracy and auditability essential to every loan processed.
Challenge
The client’s operations team was manually processing 20 to 40+ documents for every loan application — including borrower financial statements, payslips, ATO notices of assessment, identification documents, and compliance declarations. Staff had to extract information, reconcile it across statements and forms, check for inconsistencies, and validate data against internal business rules, serviceability requirements, and regulatory standards.
This manual approach created a growing set of issues:
Slow, inconsistent turnaround times — each application required approximately 60 minutes of manual processing, impacting settlement timelines and broker productivity.
Compliance exposure — responsible lending obligations require consistent, demonstrable validation of borrower information. Manual processes introduced interpretation variability and made it difficult to evidence compliance decisions at audit.
Limited scalability — the team could process only ~53 applications per month, constraining growth of the broker network as application volumes increased.
Rising per-loan costs — at $38 per application, manual processing was eroding margins and limiting reinvestment in customer experience and broker support.
Variable document quality — broker-submitted application packs arrived in inconsistent formats with missing or ambiguous information, creating significant friction for staff performing borrower verification and serviceability assessments.
As application volumes increased, the organisation faced declining throughput, higher rework rates, and growing risk of compliance breaches — making intelligent automation essential to achieving operational efficiency while maintaining regulatory consistency.
Stack Highlights
The Approach
Leveraging our AWS-native Intelligent Document Processing (IDP) accelerator, Codex delivered a fully functional automation solution within 6 weeks. Our approach was aligned to the document verification and record-keeping requirements of Australian mortgage lending:
1. Conducted workshops with lending, operations, and compliance stakeholders to understand document flows, regulatory context (NCCP, Privacy Act, AML/CTF), and key pain points in the borrower verification process.
2. Mapped the end-to-end mortgage document lifecycle to identify the highest-value automation opportunities, including serviceability documents, identity verification, and financial reconciliation.
3. Designed a modular AWS-native IDP architecture deployed in the AWS Sydney region (ap-southeast-2) for Australian data residency, with IAM least-privilege access controls scoped per Lambda function and borrower PII stored in a dedicated S3 bucket organised by processing stage.
4. Embedded automated validation and cross-document consistency checks directly into the processing pipeline — including income reconciliation, identity matching, missing-document detection, and conflict resolution — with provenance metadata recorded in stage outputs for processing auditability.
5. Orchestrated a traceable workflow in Step Functions with Map-state parallelism, retry logic on stages 3–6, and email-based reviewer notification for exceptions and missing documents.
We then implemented broker document ingestion via email polling into Amazon S3, enabling automated processing without requiring migration of existing document repositories for the pilot.
The resulting workflow automated document ingestion, extraction, validation, and triage — reducing dependency on manual review while improving consistency and traceability across lending document operations.
The Outcome
Codex delivered a scalable, AI-driven document automation workflow purpose-built for regulated mortgage processing.
Key results:
Processing time per application from 60 mins to 15 mins (75% reduction)
Annual processing cost reduced from $24,168/year to $6,042/year (75% reduction)
Monthly throughput of 53 applications increased to 212 applications (300% increase)
Zero dditional staffing required
Overall, the solution made the document processing workflow nearly 75% faster, reducing task time from 60 minutes to 15 minutes. This enabled the organisation to process 4× more tasks with no additional staff, significantly reducing operational cost while strengthening compliance and improving customer experience.
Why This Matters for Financial Services
Mortgage lending is one of the most document-intensive, compliance-heavy processes in financial services. The combination of high document volumes, strict regulatory requirements, and the need for speed creates a unique challenge that manual processes cannot sustainably address.
This engagement demonstrates that intelligent document automation — when purpose-built for the regulatory landscape — can simultaneously improve speed, reduce cost, strengthen compliance, and scale operations without compromising the human oversight that responsible lending demands.
Intelligent Document Processing Highlights
DocumentIQ automates extraction, validation, and triage of financial documents — including Australian bank statements, payslips, ATO notices, and identity documents — providing consistent, accurate, and fast results in place of manual processing. The system interprets ambiguous fields, detects missing information, and identifies anomalies that rule-based logic would miss.
Every aspect of the solution is designed for the compliance requirements of Australian mortgage lending. Responsible lending rules are embedded directly in the processing pipeline, not applied as an afterthought. Human-in-the-loop review is maintained for exception cases, ensuring regulatory accountability.
Deployed in the AWS Sydney region (ap-southeast-2) with least-privilege IAM policies scoped per Lambda function. Borrower document artefacts are stored in a dedicated S3 bucket with prefix-based access controls. Lambda's ephemeral compute model ensures no borrower data persists in memory between invocations.
DocumentIQ connects securely to the organisation's existing systems, enabling automation without data migration or disruption to established workflows.
The IDP framework serves as a long-term capability, enabling the lender to expand automation across additional document types, multi-applicant scenarios (joint borrowers, guarantor arrangements), and broader loan processing workflows.
Talk to Us
We would love the opportunity to connect and understand more about the problems you are trying to solve.
Adrian Campbell
Chief AI Officer
Martin Campbell
Managing Partner
Get in touch to coordinate a meeting with one of our technical experts.
Australia: +61 7 3132 3002.
