Taming Uncertainty via Automation: Observing, Analyzing, and Optimizing Agentic AI Systems
Dany Moshkovich, Sergey Zeltyn
TL;DR
The paper introduces AgentOps, a six-stage automation pipeline designed to observe, measure, detect, analyze, optimize, and automate agentic AI systems that rely on multiple LLM-powered agents with evolving memory and tool use. It maps these capabilities to four user roles—developers, testers, SREs, and business users—and details how instrumentation, metrics, RCA, and automated remediation can be integrated to tame, rather than eliminate, uncertainty. Key contributions include a taxonomy aligning AgentOps components with roles, a comprehensive pipeline for behavior observation through automated instrumentation, and concrete automation strategies that enable self-improving AI systems in enterprise contexts. The work underscores standardization, graph-based analytics, and self-healing execution as practical pathways to safer, more reliable, and cost-efficient agentic deployments.
Abstract
Large Language Models (LLMs) are increasingly deployed within agentic systems - collections of interacting, LLM-powered agents that execute complex, adaptive workflows using memory, tools, and dynamic planning. While enabling powerful new capabilities, these systems also introduce unique forms of uncertainty stemming from probabilistic reasoning, evolving memory states, and fluid execution paths. Traditional software observability and operations practices fall short in addressing these challenges. This paper presents our vision of AgentOps: a comprehensive framework for observing, analyzing, optimizing, and automating operation of agentic AI systems. We identify distinct needs across four key roles - developers, testers, site reliability engineers (SREs), and business users - each of whom engages with the system at different points in its lifecycle. We present the AgentOps Automation Pipeline, a six-stage process encompassing behavior observation, metric collection, issue detection, root cause analysis, optimized recommendations, and runtime automation. Throughout, we emphasize the critical role of automation in managing uncertainty and enabling self-improving AI systems - not by eliminating uncertainty, but by taming it to ensure safe, adaptive, and effective operation.