Glossary
Trajectory Quality
An evaluation of the intermediate steps in a multi-step agent workflow — tool selection, reasoning quality, error recovery, escalation decisions — not just the final output. A tool-calling error early in a trajectory silently corrupts every downstream step, which final-answer accuracy alone won't catch.
An agent can produce a correct final answer while its reasoning path silently corrupted an intermediate step. Trajectory evaluation is the only way to catch that before it compounds.
Can an agent have a correct final answer but poor trajectory quality?
Yes — that's exactly the failure mode trajectory evaluation exists to catch. An agent can call the wrong tool, recover from an error by luck, or take an inefficient path, and still land on the right final answer. Final-answer accuracy alone won't surface any of that.
Where this comes up
AI Agent Observability: What to Track
Need more than a definition?
If this term describes a problem you have, find out where you stand with a free scored diagnostic.