Your business has built a first-party AI agent. Now someone has to track it. In this clip from our session at Superweek, Jordan Peck walks through three approaches, and the tradeoffs of each:
1. Post-hoc chat log analysis: Dump transcripts into a data lake and run NLP for sentiment. Works, but it's slow and creates serious PII exposure risks.
2. Front-end & server-side tracking: Track message sends/receives on the front end, API calls on the back end. Decent coverage, but you lose the why, why did the agent choose to use a tool, or not?
3. Tell the agent to track itself: Instruct Claude or another LLM to fire events to your analytics tool. Technically challenging, and agents don't always do what they're told. How do you even know when it's decided not to track?
The real answer? Probably all three combined.
🎥 Watch the full session: https://snowplow.io/events/track-ai-agents-in-2026