As Fraud and Agentic Risks Mount, Data Provides Continuity

The episode highlights a fundamental shift in payments risk: fraudsters are no longer limited to isolated transactions but are leveraging AI‑driven scripts, virtual machines, and agentic bots to launch coordinated, high‑volume attacks. Traditional rule‑based systems that flag a single suspicious purchase are increasingly ineffective because the malicious activity is distributed across thousands of seemingly benign accounts. Participants stress that organizations must move beyond transaction‑level checks and adopt network‑wide analytics that map email creation patterns, IP overlaps, and device fingerprints to expose hidden criminal rings.

A second theme centers on data quality and continuity. As Dermit explains, AdData processes up to 2.4 billion email interactions monthly, yet the value lies in the contextual history that links those interactions over time. Fresh, accurate data feeds machine‑learning models, credit‑report‑style identity graphs, and real‑time risk scores. Stale or static data, as Jen recalls from her banking experience, can misclassify good customers and let bad actors slip through. The conversation underscores that a robust data pipeline—combining verified sources, social signals, and continuous profile evolution—is essential for distinguishing synthetic identities from legitimate changes.

Finally, the panel warns that agentic AI is eroding the clear divide between bots and human behavior. Automated purchases once triggered automatic denial, but today AI assistants can act on behalf of real users, making “automation” an ambiguous signal. This raises new challenges for chargeback disputes, deep‑fake impersonation, and trust in executive‑level transfers. Companies are therefore investing in multi‑signal identity graphs, enriching data with positive behavioral cues, and rethinking authorization models. By embracing a layered, data‑driven approach, firms can maintain continuity while adapting to the evolving fraud landscape.