Keynote

Datadog has always been driven by a broader vision of helping teams understand and operate complex systems. In this session, you'll hear from Michael Whetten, Senior Vice President of Product Management, and Cansu Berkem, Director of Product Management, as they share the latest updates across the Datadog product suite and discuss how that vision continues to shape the platform's evolution and support the next generation of AI-driven applications.

From expanded agentic capabilities and new security and compliance features for global enterprises to improvements in data reliability for AI applications, join us for a first look — live from the heart of the AI industry — at how Datadog helps teams understand and operate complex, distributed environments.

As software engineering shifts from predictable code to non-deterministic AI agents, traditional development playbooks and rigid technical specifications fall short. Navigating this evolving landscape requires engineering teams to fundamentally relearn how software behaves and discover entirely new ways to manage volatile user interactions.

In this session, Sumedha Goyal, Senior Software Engineer at Adobe, will share how her team navigated this paradigm shift while building conversational AI agents for Frame.io. She will discuss how cultivating a "let's try it" culture of rapid experimentation allowed her team to develop a new technical intuition for AI systems. Sumedha will then dive into the practical architecture they deployed, detailing how they leverage evals and Agent Observability to handle complex error states, optimize costs, track goal completeness, and detect when users attempt to steer agents in the wrong direction.

You will walk away with tips to transform your team's engineering mindset and implement the observability guardrails necessary to confidently ship production-ready AI agents.

Engineering velocity has accelerated by orders of magnitude and the only way to keep up is to close the loop by giving agents observability tools. Cognition is the company behind the AI software engineer Devin, and they use Devin to build Devin.

When observability alerts fire, they route to an internal channel where Devin immediately investigates: inspecting metrics, analyzing error logs, distinguishing real issues from expected noise like provider rate limits, and suggesting or shipping fixes. Devin uses Agent Observability, Datadog’s MCP server, and AWS to give it access to the crucial insights it needs to repair, improve and iterate on itself. Devin even set up and architected all the observability infrastructure such as metric plumbing, Terraform-managed dashboards, and monitors.

Jared Zoneraich, an engineer at Cognition, walks through how the team instruments production AI agents at scale (inference failures, provider fallback behavior, latency, token usage, cost), and more importantly, how they've closed the loop so that observability becomes an input to an autonomous engineering workflow rather than a passive dashboard humans stare at. Join to learn how to enable your own AI tooling to self-monitor and self-improve over time.