AURA: the AI that anticipates operational risk

What it is about

Mario Jara Gómez, Vice President of Operational Risk at Banco Popular Dominicano, presented at the Fourth Annual Risk Conference (September 17, 2026) AURA, a multi-agent ecosystem that the bank built from operational risk to understand changes earlier and act better.

The 23 slides tell the journey in five steps. First the challenge: more than 6,000 elements under management (more than 2,000 procedures, 1,600 risks and 2,400 controls) that change daily, with a review that arrived after the change, results that varied depending on who analyzed and effort that was spent before reaching the analysis. Then the decision: not to start with AI but to redesign the work first, with the principle that AI proposes and the analyst decides. Then the construction of a platform on Microsoft 365 and Power Platform, with 48 agents and 105 flows.

AURA operates five capabilities: continuous supervision of risks, controls and procedures; conversational capture of loss events through to regulatory reporting; a portal for first-line managers; document intelligence for relationships and economic groups; and internal control in 159 branches. Production results include 100% of the universe under continuous observation, seven times more analytical capacity and more time for value analysis. The presentation closes with the recognition from Microsoft and the next frontier: more access, more capacity and more scope.

Objectives

• Explain why the scale of operational risk management required a different capability and not just more people or more information.
• Show the method of redesigning the work first and scaling when value is demonstrated, with AI as support for expert judgment.
• Present the five capabilities of AURA and the production results reported by the bank.

Key takeaways

1. The universe under management exceeds 6,000 elements: more than 2,000 procedures, 1,600 risks and 2,400 controls, with a historical base of five years of reconciled events and losses.
2. The guiding principle is that AI proposes and the analyst decides: AI processes and generates options; expert knowledge validates, interprets, challenges and decides.
3. The multi-agent platform brings together 48 agents and 105 flows on Teams, SharePoint, Power Platform, Copilot Studio and Power BI.
4. Document intelligence analyzed more than 80,000 documents with 95% precision, and internal control connects 159 offices.
5. Reported results: 100% of the universe under continuous observation, 7 times the analytical capacity, 98% methodological precision, 40% more time for value analysis and 24-hour operation.