LEVEL 05 · LESSON 43
Tracing
A trace follows one task through the system’s actual steps.
What it is
A trace connects the steps of a single task so you can follow its path through retrieval, model calls, tools, validations, and approvals.
Why it matters
When the final result is wrong, a trace helps locate the failure instead of treating the AI system as one mysterious box.
A business example
A support reply uses the wrong return policy. The trace shows that retrieval found the right document, reranking dropped it, and generation used an older candidate. The team can fix the responsible stage.
How it actually works
Assign a task identifier and record meaningful steps with timing, versions, inputs or safe summaries, outputs, and status. Link child operations to their parent task. Preserve enough context to reproduce a failure while protecting sensitive data. A trace documents observable operations; it does not require collecting private internal reasoning or treating a generated explanation as an accurate account of how the model arrived at an answer.
A common misconception
A persuasive explanation from the model is not a substitute for an execution trace showing what tools and sources were actually used.
ONE SENTENCE TO REMEMBER
A trace follows one task through the system’s actual steps.
CHECK YOUR UNDERSTANDING
An assistant gave a wrong answer and, when asked, explains which sources it “used.” How do you find out what actually happened?
Show the answer
Answer: A.
A trace follows one task through the system’s actual steps. A persuasive explanation from the model is not a substitute for a record of what tools and sources were really used.
Take it into your next meeting
Could you identify which source and model version produced a disputed answer?
One more concept, understood.
Mark this lesson complete to keep your place.