Sawmills debuts Mills, an agentic telemetry manager

Sawmills has launched Mills, a telemetry management platform it describes as the first "agentic" system designed to manage telemetry operations across software development and production environments.

Telemetry data underpins observability tools used by engineering teams to monitor applications and infrastructure. It can include logs, metrics, and traces generated by software services. As systems grow, telemetry volumes can rise quickly, driving higher spend on observability tools and making it harder for teams to spot meaningful signals in noisy data.

Sawmills says many organisations face an ownership gap between the teams that generate telemetry and those responsible for its cost and quality. Developers often create instrumentation while shipping product changes, while DevOps and platform teams typically oversee observability budgets and operational standards. The split can lead to redundant data and delayed remediation when problems emerge.

Mills is positioned as an always-on system across the telemetry lifecycle. It monitors telemetry pipelines, identifies what Sawmills calls waste, and flags data-quality issues. It then proposes fixes and routes them to the team responsible for the affected service or component. That team reviews and approves the change. After approval, Mills deploys it and maintains a rollback option.

Policy Guardrails

The workflow is built around policies defined by DevOps teams. These policies set constraints on what changes Mills can propose and deploy. Sawmills says developers can then make changes through self-service within existing workflows, reducing repeated cross-team coordination.

Mills also targets earlier stages of the software lifecycle. Sawmills says it can catch telemetry problems in code and CI pipelines before they reach production, and feed production telemetry insights back into development. The company describes this as a continuous feedback loop across telemetry generation, delivery, and consumption.

Mills integrates with existing observability platforms, and customers do not need to migrate to start using it. That may matter for organisations with established monitoring stacks and long-running contracts with observability vendors.

Agent Integration

Sawmills also points to integration with agent-based engineering workflows. Mills includes an MCP integration and a Sawmills command-line interface. The company says other agents can query telemetry data, route issues, and take actions through these interfaces without manual handling.

Ronit Belson, CEO and co-founder of Sawmills, framed the launch as part of a broader shift toward agentic systems taking on operational tasks that previously required ongoing cross-team coordination.

"We're at an inflection point in how engineering teams operate. The rise of agentic AI means that entire categories of operational work that used to require constant human coordination can now have a dedicated, always-on owner. For telemetry, that changes everything," said Belson.

Belson also linked telemetry management to cost control and engineering productivity.

"The ownership gap that's been bleeding engineering time and inflating observability costs for years finally has a solution that doesn't require more headcount or more process; it requires a better operator. And what our customers are finding is when you finally apply a dedicated agentic operator, observability waste drops dramatically, and the data actually provides real signal," Belson said.

Sawmills claims the approach can reduce observability costs by up to 80% and improve data quality and system reliability. The release did not provide customer names, deployment scale, or benchmarking methodology for those figures.

Telemetry cost management has become a bigger issue as modern applications expand the number of services that produce data. Engineers are also increasingly instrumenting systems with detailed traces and high-cardinality metrics. While these can improve incident investigation, they can also generate large volumes of data that are expensive to store and query.

Mills is available now, and Sawmills plans to expand adoption among DevOps and platform teams that manage telemetry pipelines across development and production systems.