Uploaded July 2026 | Updated November 2026, 17 hours ago
Learn how to debug and improve AI agents in production with Arize AX. In under 3 minutes, see how AI agent observability, tracing, LLM evaluations, Signal, and Agent Swarms help turn production failures into measurable, reviewable fixes.
π Start tracing and evaluating your agents for free:
app.arize.com/auth/login?utm_source=youtube&utm_medium=video&utm_campaign=devrel&utm_content=intro-to-arize-ax
AI agents can work perfectly during development and still fail in productionβhallucinating, calling the wrong tool, getting stuck in loops, or becoming slow and expensive.
This quick Arize AX walkthrough shows how AI engineering teams can:
β’ Trace complete agent trajectories, including subagents, LLM calls, tool calls, decisions, latency, and cost
β’ Run LLM-as-a-judge evaluations on every agent run
β’ Evaluate tool selection, instruction following, response quality, and custom criteria
β’ Create evals with Arize Skills or Alyx
β’ Use Signal to identify and group recurring production failures
β’ Run managed agents through Agent Swarms to investigate issues and propose fixes for human review
Together, these workflows create an agent improvement loop: observe what happened, measure performance, find recurring problems, and make informed improvements over time.
Chapters:
00:00 Why AI agents fail in production
00:24 What is Arize AX?
00:37 Trace every agent trajectory
01:21 Evaluate every agent run
02:00 Find recurring failures with Signal
02:23 Automate monitoring with Agent Swarms
02:39 Close the agent improvement loop
π Explore Arize AX: arize.com/ax
π Read the docs: arize.com/docs/ax
π Subscribe for more videos about AI agents, LLM evaluations, observability, and AI engineering:
youtube.com/@arizeai?sub_confirmation=1
#ArizeAX #AIAgents #AIObservability
Learn how to debug and improve AI agents in production with Arize AX. In under 3 minutes, see how AI agent observability, tracing, LLM evaluations, Signal, and Agent Swarms help turn production failures into measurable, reviewable fixes.
π Start tracing and evaluating your agents for free:
app.arize.com/auth/login?utm_source=youtube&utm_medium=video&utm_campaign=devrel&utm_content=intro-to-arize-ax
AI agents can work perfectly during development and still fail in productionβhallucinating, calling the wrong tool, getting stuck in loops, or becoming slow and expensive.
This quick Arize AX walkthrough shows how AI engineering teams can:
β’ Trace complete agent trajectories, including subagents, LLM calls, tool calls, decisions, latency, and cost
β’ Run LLM-as-a-judge evaluations on every agent run
β’ Evaluate tool selection, instruction following, response quality, and custom criteria
β’ Create evals with Arize Skills or Alyx
β’ Use Signal to identify and group recurring production failures
β’ Run managed agents through Agent Swarms to investigate issues and propose fixes for human review
Together, these workflows create an agent improvement loop: observe what happened, measure performance, find recurring problems, and make informed improvements over time.
Chapters:
00:00 Why AI agents fail in production
00:24 What is Arize AX?
00:37 Trace every agent trajectory
01:21 Evaluate every agent run
02:00 Find recurring failures with Signal
02:23 Automate monitoring with Agent Swarms
02:39 Close the agent improvement loop
π Explore Arize AX: arize.com/ax
π Read the docs: arize.com/docs/ax
π Subscribe for more videos about AI agents, LLM evaluations, observability, and AI engineering:
youtube.com/@arizeai?sub_confirmation=1
#ArizeAX #AIAgents #AIObservability










