“Genie Agents made our risk management more agile and efficient, helping us to identify and prioritize potentially fraudulent transactions.” - Flávio Arena, Tech Manager, PicPay
PicPay, one of Brazil’s leading digital banks, uses statistical anomaly triggers to flag unusual transaction behavior in near real time. But the sensitivity that helped surface potential fraud also created a growing queue for human review. With Databricks Genie Agents, PicPay added a coordinated four-agent triage layer that assesses flagged anomalies, prioritizes risk and keeps analysts in the loop.
Key highlights:
- 60% fewer false positives: Agent-based risk scoring reduced the number of alerts escalated to human reviewers.
- 93% early detection on disputed value: Among anomalies routed to the agents, 93% of the value later disputed by customers had already been classified as high risk during triage.
- 10x faster response at peak volume: Parallel processing replaced sequential manual review, enabling near real-time action on potential fraud.
- 15–20 minutes to 3–7 minutes: Splitting analysis across four focused agents improved consistency while shortening processing time.
- Four coordinated agents: PicPay analyzes the flagged transaction window, retrieves historical behavior, consolidates findings into a risk score and standardizes the output before human review.
- Auditable decisions: Supervisors can inspect the agent conversation, the exact query run against the database and the resulting response.
🔗 Check out the full story 👈
