How Adobe is Building Agentic Loop to Detect an Issue, Experiment with Various Solutions

Session track: AI Engineering

Session time: 08:00 pm - 08:30 pm UTC November 12

Session description:

At Adobe, our Data Governance system leverages Neo4j AuraDB to persist and retrieve the graph object model. Although our business is multi-tenant, we currently use a single-tenant Neo4j instance, which often leads to noisy-neighbor issues that are difficult to resolve. To address this, we are building an Agentic Loop that observes signals from multiple sources to detect problems, creates a copy of the graph, and runs experiments to validate hypotheses before sharing the solution with human developers. This approach has enabled us to reduce work that would typically take weeks or even months of correlating data across disparate systems to just minutes.