Building Agentic AI Platforms
summary
Claims an autonomous mortgage pipeline handles 47 separate steps end to end, verifies income over weekends, and adapts to novel cases like crypto income by researching rules, calling external services, and creating repeatable verification paths. Architecture rests on workflow orchestration, event-driven communication, and ML pipelines with real‑time inference; includes a mainframe translation agent, versioned agent deployments, decision provenance logs and a flight‑recorder event stream. Performance notes: adaptive scaling (300% spike handling) and monitoring via decision‑reversal rates show when agents need retraining.
AI-written summary · about
chapters (8)
- 0:00Introduction to Agent AI in Mortgage Processing
- 1:01Current Mortgage Processing Challenges
- 1:39Understanding Agent AI Systems
- 2:34Real-World Examples of Agent AI
- 6:46Key Components of Agent AI Architecture
- 10:16Challenges in Implementing Agent AI
- 21:10Practical Lessons and Future Directions
- 29:58Conclusion and Final Thoughts
related talks
38mFrom Paperwork to Pipeline: Autonomous Mortgage AI
29mLeveraging AI to Enhance DevEx & Operational Efficiency
34mAI Agents: Transforming Autonomous Decision-Making
27mMulti-Agent AI: Transforming Business Operations and Value Creation
18mAI Agents for a Green Future
26mThe AI-Based Banking Era
44mBuilding Agentic AI Systems: From Transactions To Conversations
21mBuilding Agents (the hard parts!)