The telemetry problem starts with a basic mismatch inside engineering organizations.
According to the Ronit, DevOps teams own observability platforms and their budgets, while developers create the telemetry data and are focused on shipping code quickly rather than managing telemetry costs. That creates an inefficient manual back-and-forth process across the organization.
Sawmills was built in response to that challenge. After initially launching with LLM-based capabilities to help DevOps teams optimize data manually, the company is now introducing AI agents designed to automate a continuous feedback loop across the telemetry lifecycle, from requirement writing and CI/CD gates to production.
The episode also looks at why this matters more now. Telemetry volume was already growing quickly, and the rise of AI-written code has made the problem even harder to manage.
?>