AI Exchange @ Tessl

Dave explored how Wayve is using AI agents to support code review, with a focus on making sure their output is reliable, accurate and actually useful. He showed how the team measures agent performance, gathers feedback and continuously refines prompts to improve results.

His through-line: observability, evaluation and data-driven decision making are what make an AI agent deployable in production. Rather than replacing engineers, agents enhance software development when paired with the right guardrails and oversight.

Nick took on the growing security challenges of autonomous AI agents, demonstrating why traditional security practices alone are no longer enough.

He shared practical approaches for securing AI systems through strong guardrails, zero-trust principles and AgentOps — how engineering teams can deploy agents safely, reducing risk and keeping control as those agents become more autonomous.