About

As AI agents move from experiment to production, the analytics gap is widening. Observability tools tell you if your agent is working, not whether your agent is working for your customers. This talk explores the emerging landscape of 1st party and 3rd party AI agents, draws a critical distinction between agent observability (the DevOps lens) and agent analytics (the customer experience lens), and demonstrates a practical approach to agentic self-tracking – where the agent itself emits structured behavioural events in real-time. With live examples of system prompts, tracking tool code, and real event data, you’ll leave with a concrete framework for instrumenting your own AI agents for genuine customer insight.

Takeaways

• Your current tracking misses agent activity — you’re losing visibility into real user behaviour
• If you don’t separate agents from humans, your metrics and attribution become unreliable
• Frontend-only tracking leaves gaps — you won’t see how decisions are actually made
• Without schema enforcement, your data becomes untrustworthy and hard to debug