As cited
Copy frozen at (site build).
ai security
Inside Elastic InfoSec's agentic SOC: How we cut AI agent LLM calls by 60%
Elastic InfoSec optimized AI agents in their security operations pipeline by building a five-step measurement and diagnosis loop that reduced large language model (LLM) calls from 19 to 7-9 per task, cutting costs by 60%. The team measured that conversation context accumulation, rather than prompt length, drives the dominant cost in agentic workflows, and that reducing unnecessary LLM calls is far more impactful than prompt trimming. The optimization methodology applies to any agent architecture and focuses on three areas: minimizing LLM call count, eliminating redundant tool queries, and ensuring behavioral consistency across similar inputs.
Why it matters: Security operations teams running AI agents for alert triage and investigation can significantly reduce LLM API costs and latency by applying systematic optimization techniques rather than guessing at prompt changes; this is immediately relevant for organizations managing dozens of agents processing hundreds of alerts daily.
- Source published
- First seen by Cybersecurity Tracker
Source attribution
Correction
Correction recorded as of .
ai security
Inside Elastic InfoSec's agentic SOC: How we cut AI agent LLM calls by 60%
No summary had been written when this copy was frozen.
- First seen by Cybersecurity Tracker
Source attribution
Correction
Correction recorded as of .
ai security
Inside Elastic InfoSec's agentic SOC: How we cut AI agent LLM calls by 60%
Elastic's InfoSec team optimized 14 artificial intelligence (AI) agents running in their security operations pipeline, reducing large language model (LLM) calls from an average of 19 to 7-9 per task through a five-step measurement and refinement loop. The optimization focused on minimizing LLM invocations rather than prompt length, since cost scales with call count and accumulated context tokens. The methodology applies to any Agent Builder agent and separates optimization into three areas: LLM call count reduction, tool-call discipline, and behavioral consistency.
Why it matters: Security teams using AI agents in their SOCs can directly reduce operational costs and response latency by applying Elastic's optimization framework to measure and eliminate unnecessary LLM calls, which represent the dominant cost driver in agentic workflows.
- Source published
- First seen by Cybersecurity Tracker