Automation & Integration in Toronto

Financial Services / Retail Banking

A Big Five Canadian bank (CAD $1.3T total assets, 12 million retail customers, CAD $380B mortgage portfolio — the largest mortgage lender in Canada) needed to automate its mortgage origination and servicing operations to reduce processing times, improve customer experience, and ensure B-20 compliance at scale. (1) Mortgage origination was slow: the end-to-end mortgage process (application to funding) averaged 28 business days — during which the customer and mortgage specialist exchanged an average of 18 touchpoints (phone calls, emails, document requests, and status updates). The bank processed 180,000 mortgage applications per year with 2,400 mortgage specialists and a back-office team of 850 underwriters and processors. The process involved 14 different systems — CRM, application processing, credit bureau, property valuation, title search, document management, underwriting workbench, pricing engine, commitment generation, legal, registration, funding, and core banking. (2) Document processing was manual: mortgage applications required 15-25 documents per application (income verification, employment letters, property documents, insurance, and identity) — each manually reviewed by underwriters for authenticity, completeness, and accuracy. Document review consumed 35% of the underwriting time. Common issues: documents arriving in wrong format (photos rather than scans), incomplete documents requiring re-request, and fraudulent documents requiring investigation. (3) B-20 compliance burden: OSFI Guideline B-20 imposed strict mortgage underwriting standards — stress test (qualifying rate), debt service ratios (GDS/TDS), and documentation requirements. Every mortgage required B-20 compliance verification — a manual checklist process that added 2-3 days to the underwriting timeline and was the primary source of compliance audit findings (3.8% of mortgages had B-20 documentation gaps identified in internal audit). (4) Servicing complexity: the CAD $380B portfolio required ongoing servicing — payment processing, escrow management, renewal processing, rate changes, and discharge. Renewal processing alone affected 65,000 mortgages per year, each requiring rate offer generation, customer communication, and documentation — a process that started 120 days before maturity but still resulted in 8% of renewals being processed after maturity date due to processing bottlenecks. (5) Bilingual requirement: all customer communications — application status updates, commitment letters, renewal offers, and servicing notifications — needed to be available in English and French, with the customer's language preference tracked and enforced across all automated communications.

We built a comprehensive mortgage automation platform over 16 weeks. (1) Application orchestration: event-driven workflow engine orchestrating the end-to-end mortgage process — from application intake through funding. Each mortgage application became an event stream — every action (document upload, credit pull, valuation order, underwriting decision) published as an event, triggering downstream workflows automatically. Real-time status: mortgage specialists and customers saw real-time progress through every stage — eliminating "where is my mortgage" calls that consumed 30% of specialist time. Parallel processing: activities that were previously sequential (credit check, then property valuation, then document review) automated to run in parallel — credit bureau pulls initiated immediately at application, property valuation ordered simultaneously, and document upload requests sent concurrently. (2) Intelligent document processing: ML-based document classification — uploaded documents automatically classified by type (T4, NOA, employment letter, property tax bill, insurance certificate) with 96% accuracy. Data extraction: key fields extracted from documents — income amounts from T4s, assessment values from property tax bills, coverage amounts from insurance certificates — pre-populating the underwriting workbench. Fraud detection: ML models comparing document patterns against known fraud signatures — font inconsistencies, data anomalies, and template mismatches flagged for investigation. Document completeness: automated tracking of required documents per application type — proactive reminder notifications sent to customers for missing documents in their preferred language. (3) B-20 compliance automation: automated stress test calculation (qualifying rate applied to the requested mortgage, GDS/TDS calculated from extracted income and obligation data), automated documentation check (every B-20 requirement verified against the application file — income verification, employment confirmation, property valuation, and identity documentation), compliance scoring (each application scored for B-20 compliance completeness — applications scoring 100% cleared automatically, applications with gaps flagged with specific remediation instructions), and audit trail (every B-20 calculation, verification, and decision documented with timestamps and data sources — creating an audit-ready record for OSFI examination). (4) Renewal automation: automated renewal workflow triggered 120 days before maturity — rate offer generation (personalised renewal rates based on customer segment, portfolio retention models, and competitive analysis), customer communication (renewal offers sent in the customer's preferred language through their preferred channel — email, mobile app, or mail), acceptance processing (customer acceptance through digital channels — electronic signature on renewal agreement, automatic system updates), and escalation (non-responses escalated through progressive outreach — app notification, email, phone call from specialist — with timing optimised to maximise renewal before maturity). (5) Bilingual communication engine: all automated communications (application status, document requests, commitment letters, renewal offers, servicing notifications) generated in the customer's registered language preference. Dynamic content: communications personalised based on mortgage type, customer segment, and regional requirements — not just translated but culturally appropriate for English and French Canadian markets.

Origination speed: application-to-funding from 28 business days to 11 business days (61% reduction). Parallel processing and automated document handling eliminated the sequential bottlenecks that extended timelines. Customer touchpoints from 18 to 7 — each touchpoint was meaningful rather than administrative. Document processing: manual document review time reduced by 72% — ML extraction pre-populated underwriting data, with underwriters reviewing AI-extracted data rather than reading raw documents. Fraud detection: ML models identified 340 suspicious applications in the first year — 89 confirmed fraudulent (from 45 detected annually under manual review). B-20 compliance: compliance gap rate from 3.8% to 0.3% (92% reduction). Zero OSFI audit findings related to B-20 documentation in the first annual examination following implementation. Automated compliance audit trail eliminated 6 weeks of annual internal audit preparation time. Renewal processing: on-time renewal rate from 92% to 99.2%. Renewal processing time from 45 minutes per file to 8 minutes (82% reduction). Customer retention at renewal improved from 78% to 86% — automated competitive rate offers and proactive communication reduced attrition to competitors. Staffing efficiency: mortgage specialists redeployed from 2,400 to 1,800 (25% reduction) with 15% higher per-specialist volume. Back-office team from 850 to 450 underwriters and processors. Bilingual: 100% of automated communications delivered in customer's preferred language — French-language customer satisfaction scores increased 18 points (previously, French communications were delayed due to translation bottleneck). Annual value: CAD $185M (staffing efficiency — 1,000 FTE reduction at average CAD $95K fully loaded cost, offset by technology and retained staff costs), CAD $78M (revenue acceleration — faster origination capturing time-sensitive purchase mortgages that previously lost to faster competitors), CAD $52M (retention improvement — 8 percentage point improvement in renewal retention on CAD $380B portfolio), CAD $28M (fraud prevention — 89 confirmed fraudulent applications intercepted, plus deterrence effect), CAD $12M (compliance — eliminated audit findings, reduced examination preparation, and avoided potential OSFI penalties) = CAD $355M. Development cost: CAD $620,000. Payback: under 16 hours.