BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//hacksw/handcal//NONSGML v1.0//EN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20260722T235903Z
DESCRIPTION:Click for Latest Location Information: http://dgiq-aigov2026.da
 taversity.net/sessionPop.cfm?confid=166&proposalid=16738\nA global tier-1 b
 ank operating across 40+ countries did everything right on the AI side, mod
 els, infrastructure, talent, board backing, and deployed agents across reta
 il and wealth banking. Then satisfaction fell 18 points in year one, compla
 ints that &quot;the bank doesn&#39;t know who I am&quot; tripled, and high-
 value wealth clients began moving assets out. An audit found 23% of AI-driv
 en interactions were acting on mismatched identity.\n\nThe team did what mo
 st do: they assumed it was a model problem. They spent months and serious b
 udget fine-tuning, retraining, and re-prompting, chasing accuracy that neve
 r came. The real failure wasn&#39;t the model at all. It was data governanc
 e: the same customer existed as up to five records across 17 siloed systems
 , and the agents inherited the fragmentation. The fix was entity resolution
  as a real-time identity layer. Within two quarters, identity complaints fe
 ll 74%. This is that story and what it costs to debug the wrong layer.\n\nT
 akeaways:\n\n
 How a mature, well-funded AI program quietly broke on fragmented identity&m
 dash;an 18-point satisfaction drop and complaints tripling\n
 The costly wrong turn: months and significant budget spent fine-tuning the 
 model when the real failure was data governance\n
 The &quot;confident and wrong&quot; failure mode: agents faithfully acting 
 on the wrong customer (23% of interactions)\n
 The root cause and the fix: 17 siloed systems, up to five records per custo
 mer, resolved by entity resolution as a real-time identity control\n
 The turnaround&mdash;identity complaints down 74% in two quarters&mdash;and
  a way to assess your own identity foundation before the next AI initiative
 \n\n
DTSTART:20261117T120000
SUMMARY:One Customer, Seventeen Identities: How a Global Bank's AI Got Conf
 ident and Wrong
DTEND:20261117T124459
LOCATION: See Description
END:VEVENT
END:VCALENDAR