~/panel

Panel Discussion: Driving the Agentic Shift in Banking Adoption Governance and Scale

on YouTube 49m
Panelat TestMu Conference 2026 · 2026-08-19..08-21· [ai-assisted-development]

speakers

name:Javier Solano
links:none on file
name:Mauricio Sánchez
links:none on file
name:Humberto Henríquez
links:none on file

resources

video:https://youtu.be/HXTE2aP4-lY
slides:none on file

summary

Panel maps a practical playbook: a live deployment that uses agents for deduplicating the customer golden record with a human-in-the-loop second check each day, three filters for agent vs copilot (reversibility same business day, verifiable against a policy/ledger, prefer a rules engine), testing by scenarios at scale with human-graded golden sets and shadow rollout, ongoing validation where agents testing agents detects loops and ensures task completion, and a strict fallback ladder: retry bounded, deterministic rule, human, decline plus an integration layer and curated data to prevent deepening hallucinations.

AI-written summary · about

chapters (45)

  1. 0:00Panel Opens - Driving the Agentic Shift in Banking
  2. 1:08Panel Intros - Javier Solano, Humberto Enriquez, and Mauricio Sanchez
  3. 4:43Setting the Scene: Governance, Guardrails and Non-Deterministic Agents
  4. 6:18The Most Promising Production Use Cases in a Regulated Bank
  5. 6:49Deduplicating the Customer Golden Record With Agents
  6. 8:21Two Sides of the Coin: Automation Plus Human in the Loop
  7. 9:22The Second Check That Reviews Everything Created That Day
  8. 10:25Why the First Wave Focused on Customer-Facing Chatbots
  9. 10:56The Real Value Isn't Opinion - It's Execution With a Human in the Loop
  10. 11:59Reconciliation, Back Office, KYC and Compliance
  11. 12:31Observability Gaps and Getting Back to Production Faster
  12. 13:02What Is a Real Agent vs a Copilot?
  13. 13:32If It Touches Money or a Customer, Start With a Copilot
  14. 14:03Filter One: Reversibility Within the Same Business Day
  15. 14:35Filter Two: Can You Verify the Decision Against a Policy or Ledger?
  16. 15:06Filter Three: If a Rules Engine Can Do It, Use a Rules Engine
  17. 15:36Testing Non-Deterministic Chat and Voice Agents
  18. 16:36Guardrails and Identifying When the Agent Goes Off the Rails
  19. 17:38Why a Human Must Stay in the Loop for at Least the Next Six Months
  20. 18:10Structured Minds Produce Structured Agents
  21. 19:10From Pass/Fail to the Distribution of Behaviour Across Many Runs
  22. 20:14Scenarios at Scale, Human-Graded Golden Sets, Shadow Mode Rollout
  23. 21:17You Need Different Tools to Test Agents Than to Test Systems
  24. 22:19Validating the Trajectory: Right Tool, Right Arguments, Right Sequence
  25. 23:21Testing Stop Behaviour When the Agent Doesn't Know
  26. 23:53Every Model Version Release Is a Change in Production
  27. 24:56Regular Re-Testing, Like Penetration Testing
  28. 25:28Agents Testing Agents as an Ongoing Maturity Process
  29. 26:00Making Sure the Agent Actually Finishes Its Task
  30. 26:30Detecting Loops and How the Agent Tells the User It Failed
  31. 28:02Have You Experimented With Voice Agents Yet?
  32. 29:03A Voice Agent Doesn't Think Alone - It Orchestrates Others
  33. 29:34Reading the Room and Defusing a Frustrated Customer
  34. 30:34Why Voice Is Harder: Legacy IVR, Integration and Localisation
  35. 32:36Governance: What You Let the Agent Do and Not Do
  36. 33:37A Project They Had to Roll Back by Assuming Voice Was Just Chat
  37. 34:39Observability and Fallback Protocols Are Not Optional
  38. 35:39Detection, Tracing and Correction as Three Ingredients
  39. 36:40Start From the Worst Thing the Agent Could Do, Not From Success
  40. 37:41The Fallback Ladder: Retry Bounded, Deterministic Rule, Human, Decline
  41. 39:44Technical Lessons From Deploying Into Core Banking
  42. 40:47Take Care of Your Data - It's the Elephant in the Room
  43. 41:48Legacy Isn't the Problem; The Integration Layer Is Mandatory
  44. 43:20Q&A: Automated Regression Tests When Output Isn't Deterministic
  45. 45:55Q&A: Auditing the Security Bounds of Agent-Triggered External Tools

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.