Why AI Agents Fail Without Operational Memory – BigPanda Triple-T
Every enterprise deploying AI agents for IT operations hits the same wall: agents without grounded operational context hallucinate, misroute, and can’t learn from outcomes. This session presents the architectural pattern behind an operational knowledge graph that ingests structured ITSM data alongside unstructured tribal knowledge — chat transcripts, bridge calls, runbook fragments — and turns it into a compounding data layer that AI agents consume in real time.
– Why retrieval-grounded agents outperform fine-tuned models on enterprise operational data
– How a tiered sensing funnel (rules, AI correlation, context refinement) drops 80-90% of event noise before agents ever see it
– What “”compounding”” means in practice: the system gets measurably smarter without retraining
From starting his career as a United States Marine Corps helicopter pilot to becoming the Chief Innovation Officer of BigPanda, Jason Walker brings skill and professionalism to any situation regardless of scale, speed, or complexity. Jason is driven to help customers accelerate their AIOps journey by leveraging BigPanda’s practical applications to enhance operations performance and ignite transformative growth. He previously served as field CTO and then Chief Customer Officer for BigPanda where he led the customer success team. Previously, Jason spent 10 years at Activision-Blizzard where he led Blizzard’s IT operations transformation to support multi-product, always-online global gaming services with an AIOps-driven centralized OpsCenter. His positive experience as a BigPanda customer at Activision-Blizzard convinced him to become a Panda himself.
