may

Discover why 40% of financial AI initiatives fail, and learn how a unified data platform enables secure, real-time AI scaling in regulated environments.

AI is becoming core to financial services. But as adoption scales, so does risk. Without secure, real-time data, strong governance, and cyber resilience, AI can amplify threats as quickly as it drives innovation.

Most enterprise AI initiatives stitch together four to five systems for memory, search, and state. The Elasticsearch Platform handles all of it in one engine — and the architecture is simpler than you think.

After a year and one million messages, Elastic's Field Technology team shares five lessons from building production AI agents: why logs matter most, how retrieval thresholds shape quality, and what high token counts actually signal.

Read about the updates and bug fixes that have been included in this release.

PHAROS is a 4-agent system built with Elasticsearch that automates pharmacovigilance. It detects adverse drug safety signals using WHO-standard analysis, generates regulatory reports, and sends critical alerts in under 60 seconds.

A multi-agent duplicate detection system built with Elastic Agent Builder addresses the 56% failure rate of manual duplicate detection in Kenya's HIV patient records. It uses tiered risk scoring and specific recs to provide $195K annual savings.

The Elasticsearch Agent Builder Hackathon spotlighted Gauntlet, a top-three project that uses mocking agents to mutate tool calls, expose hidden failures, and improve AI agent reliability.

The Elasticsearch Agent Builder Hackathon showed AI agents built with Elastic tools to solve real-world problems. Winners include PHAROS (drug safety in <60s), Gauntlet (adversarial testing), and a duplicate detection agent for Kenya’s HIV program.