Uploaded April 2026 | Updated September 2026, 3 days ago

---

How do you know if your AI is actually working? In this episode from the WorkOS booth at HumanX 2026, Michael Grinich sits down with Ameya Bhatawdekar, Field CTO at Braintrust, to discuss the critical infrastructure needed to move AI from prototype to production.

Ameya dives into why "Evals" (evaluations) have become the unit tests of the AI world, and how Braintrust helps teams avoid regression in non-deterministic systems. They explore the shift toward sophisticated multi-turn agentic workflows and introduce Braintrust’s new "Topics" feature—an inbuilt agent that analyzes 100% of production logs to find hidden insights and failure modes.

Key Topics Covered:
• Why "Evals" are essential for shipping AI with confidence.
• Moving beyond prompt engineering to dynamic orchestration.
• Harnessing production data to build robust test sets.
• The Braintrust CLI: Integrating observability into the coding workflow.
• "Topics": Automated log analysis for petabyte-scale data.

Learn more about WorkOS: workos.com Evaluate your AI with Braintrust: https://www.braintrust.dev/