Cost surprises emerge after agent behavior has already become routine.
Engineers need an early signal for costly trajectories while trace ingestion must remain fast and trustworthy.
Who you’d be doing this for
“Finance asked why Tuesday cost twice as much, and I had to sift through traces all afternoon.”
Izumi Wang · AI Application Engineer
Operates a production agent that handles internal procurement requests across several model and tool calls.
What is at stake
A successful procurement agent workflow is consuming twice its expected token spend on some days. You must weigh early warnings against noisy alerts and added ingestion cost.
Why it isn’t already fixed
Every obvious fix costs something else. That’s the part you’d have to decide.
- early cost warning vs alert credibility
- forecast quality vs ingestion latency
- generalization across agents vs interpretable signals
- model capability vs predictable serving cost
Why Datadog
LLM Observability connects agent traces, evaluations, and production telemetry where AI application engineers investigate unpredictable token spend.
Written with these in mind
Not your kind of problem? 40 more at Datadog, or browse every organization.
This is the setup. The work is inside.
Running it puts you in the room: the full situation and its constraints, stakeholders who push back in their own words, and the decisions that are yours to make. What you produce becomes a Day One Plan — work you can show someone instead of describing.