BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//hacksw/handcal//NONSGML v1.0//EN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20260722T235850Z
DESCRIPTION:Click for Latest Location Information: http://dgiq-aigov2026.da
 taversity.net/sessionPop.cfm?confid=166&proposalid=16767\nMulti-agent AI sy
 stems are becoming the architecture of choice for complex financial operati
 ons — but they introduce a governance challenge our frameworks weren&
 #39;t built for. When a pipeline of autonomous agents collaborates on a dec
 ision, who owns the outcome?\n\nThis presentation shares findings from orig
 inal applied research on accountability in multi-agent AI. We built and ins
 trumented a multi-agent loan approval pipeline, discovering that errors com
 pound rather than self-correct, generic risk frameworks miss system-level f
 ailures, and where you place human oversight matters more than whether you 
 have it.\n\nThe key insight: the unit of accountability is not the individu
 al agent — it is the transition between agents. We introduce HandoffP
 ackets and the ATTAIN Framework (Auditability, Traceability, Task Attributi
 on, Intervention Points, Normative Compliance) — five measurable dime
 nsions that provide a complete accountability surface for any multi-agent s
 ystem.\n\nAttendees will leave with: a diagnostic lens for identifying acco
 untability gaps, an instrumentation pattern for post-hoc attribution, an ev
 idence-based human oversight placement strategy, and a regulatory preparati
 on playbook. Organizations that solve accountability first will deploy fast
 er, shape regulation rather than react to it, and scale agentic AI with con
 fidence.
DTSTART:20261118T093000
SUMMARY:Accountability in Multi-Agent AI: The ATTAIN Framework
DTEND:20261118T101459
LOCATION: See Description
END:VEVENT
END:VCALENDAR