The Challenge
At United Wholesale Mortgage, loan officers and processors were spending hours on repetitive tasks - manually updating pipeline stages, sending status notifications, and validating data across systems. The mortgage pipeline moved fast, and manual processes were creating bottlenecks and errors.
My Approach
Process Mapping
I shadowed loan officers and processors for a week to understand their daily workflows. I documented every manual step and identified the highest-impact automation opportunities:
- Pipeline stage transitions - Manual updates across 8 pipeline stages
- Notification workflows - Status emails sent manually at each milestone
- Data validation - Cross-referencing loan data between Salesforce and underwriting systems
- Document tracking - Manual checklist management for required documents
Building the Automation Suite
I built a modular automation framework using a combination of Flow Builder and Apex:
Flow Builder Automations
- Auto-stage progression - Triggered by key data changes, automatically advancing loans through pipeline stages
- Smart notifications - Context-aware emails and alerts based on pipeline stage, loan type, and broker preferences
- Document checklist generation - Dynamic checklists created automatically based on loan product type
Apex Trigger Framework
- Validation engine - Real-time data validation against business rules before stage transitions
- Batch processing - Nightly reconciliation jobs to catch any data discrepancies
- Integration handlers - Automated data sync with external underwriting and compliance systems
Testing & Rollout
I built a comprehensive test suite with 95% code coverage and ran a two-week pilot with a small group of loan officers before rolling out to the full team.
Results
- 60% reduction in manual processing time per loan
- 85% fewer data entry errors in the pipeline
- 2x faster average time from application to closing
- 95% code coverage with automated regression tests
Key Takeaways
The key insight was building automations that were modular and configurable. Rather than hard-coding business rules, I created custom metadata types that allowed operations managers to adjust thresholds and routing rules without needing a developer. This made the system sustainable long after I moved on.
This case study covers work completed at UWM in 2021. Specific metrics and implementation details have been generalized to protect proprietary information.