~/panel

Panel Discussion: Money Moves at Machine Speed - Trust, Risk, and Quality in Agentic Finance

on YouTube 01h 04m
Panelat TestMu Conference 2026 · 2026-08-19..08-21· [ai-assisted-development]

speakers

name:Trincy Thomas
linkedin:trincy-thomas-38399b69
name:Preetham Sunilkumar
linkedin:preetham-sunilkumar-3207388a
name:Veena Setlur
links:none on file
name:Krystyna Sylyvonchyk
links:none on file

resources

video:https://youtu.be/OmYgOY18O1Q
slides:none on file

summary

Banks and insurers are already running agents that autonomously handle account maintenance (address and beneficiary updates), quietly revert a reversible duplicate-transaction, and compress advisor prep from 20 hours to minutes while FINRA marketing review moves from 5–10 days to minutes. Governance requires shadow-mode deployments, golden datasets and rubrics, tracking override rates, and logging the full decision context as a flight recorder plus a model biography. Release criteria hinge on testability, risk appetite, and whether failures are containable.

AI-written summary · about

chapters (52)

  1. 0:00Money Moves at Machine Speed - Trust, Risk and Quality in Agentic Finance
  2. 1:36Panel Intros - Trincy Thomas, Preetham Sunilkumar, Veena Setlur, and Krystyna Sylyvonchyk
  3. 3:08Where Agents Are Truly Acting vs Only Recommending
  4. 3:38The More Real Money and Regulation, the More Agents Just Consult
  5. 4:38A Reversible Duplicate-Transaction Claim as a Safe Delegation
  6. 5:39Autonomy Isn't Binary - It's a Spectrum
  7. 6:43Earning More Responsibility Through Code Review First
  8. 9:45Reframing: The Most Important Thing an Agent Acts On Isn't the Transaction
  9. 10:16Automation Existed Since the 1990s - What's New Is Reasoning
  10. 10:46Layer One: Account Maintenance Running End to End
  11. 11:17Layer Two: Giving Advisers Back Their Most Precious Resource
  12. 12:20Twenty Hours of Meeting Prep Compressed Into Minutes
  13. 12:51Layer Three: FINRA Marketing Review From 5-10 Days to Minutes
  14. 13:52Agents Earn Operational Trust Before Consequential Trust
  15. 15:24Governing When the Regulation Itself Is Still Evolving
  16. 16:56Every Wave of Financial Innovation Outran the Rule Book
  17. 17:57Where Regulation Actually Sits Today: SR 11-7 and FINRA Guidance
  18. 18:28Institutions Building Governance Now Are Writing the First Draft
  19. 18:59Approach One: Treat Every Agent as a Regulated Actor
  20. 19:31Approach Two: Adopt the Existing Supervisory Framework
  21. 20:03Approach Three: Define the Boundaries of Acceptable Drift
  22. 21:06The Accessibility Compliance Angle From the QA Seat
  23. 22:37Documenting What the AI Was Allowed to Do, and What It Couldn't
  24. 24:09Testing a System That May Not Give the Same Answer Twice
  25. 24:40Certification Doesn't Disappear - the Criteria Change
  26. 25:10Golden Data Sets, Rubrics and Acceptable Ranges
  27. 25:41Tracking Override Rates as a Validation Mechanism
  28. 27:43Schema and Property Validation Instead of Value Matching
  29. 29:15Keeping Human Insight Meaningful Instead of a Checkbox
  30. 30:17Why Reviewers Get Bored, Not Lazy
  31. 30:50The Courtroom Model: AI Prosecutes, AI Defends, a Human Judges
  32. 32:25Testing the Human: Deliberately Injecting False Results as a Fire Drill
  33. 34:00The Fire Drill Nobody Looks Up For
  34. 35:01Label-Driven Scoring and Reducing Reviewers as Confidence Grows
  35. 36:33Shadow Mode Deployment as a Mandatory Event
  36. 38:07Audit Readiness: What Gets Logged, What Evidence Is Retained
  37. 39:08Reconstructing What Happened Is Easy; Why Is Hard
  38. 39:40Build a Flight Recorder, Not Just an Audit Trail
  39. 40:41Logging the Full Decision Context Plus Every Human Touch Point
  40. 41:42The Model Biography as the Second Evidence Layer
  41. 42:12Continuous Production Monitoring as the Third Layer
  42. 42:42Three Things to Have Ready Before an Examiner Asks
  43. 43:44Build the Evidence Trail at Architecture Time, Not Examination Time
  44. 45:50The Bar for Releasing an Agent That Can Act on Its Own
  45. 46:51It Isn't Whether You Trust It - It's Whether You Can Contain It
  46. 47:25Three Requirements: Can You Test It, Does It Match Appetite, Can You Defend It
  47. 49:56Agents Inside Engineering Teams as the Under-Discussed Impact
  48. 51:28Lessons Learned and What Never to Do Again
  49. 52:31Build the Constraint Spec Before the Capability Spec
  50. 53:32Majority Voting, Confidence Scores and Learning by Falling
  51. 57:11Never Train Agents on Clean, Perfect Data
  52. 58:44Q&A: The Critical Skills for the Next Five to Ten Years

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.