We built Sawmills to fix observability's two biggest problems: cost and quality. Here's what year one looked like.

Ronit Belson
CEO & Co-founder

A year ago today, we stepped out of stealth with a team of enterprise software veterans, and a conviction that the observability industry had a fundamental problem nobody was solving.

The thesis was simple: enterprises are spending millions every year on telemetry data they never use. Observability has become the second-largest engineering line item after cloud. Costs spike unpredictably. Tool sprawl is the norm. A single misconfigured service can rack up hundreds of thousands in overage fees overnight. 

And underneath all of it, there's a quality problem. Incomplete fields, inconsistent naming, and noisy signals delay detection, confuse alerts, and stretch every incident. As telemetry volume grows, these gaps multiply into fragmented standards that can't keep pace with change. Without quality telemetry, teams are paying to collect, process, and store massive volumes of data that never yield a useful insight.

For years, the tradeoff felt locked in: pay the bill, or lose visibility. Accept low-quality telemetry, or spend a fortune trying to fix it. There was no third option. That is what Sawmills set out to change. And after twelve months of building, shipping, and working alongside some incredible engineering teams, I can say with confidence that we did.

The introduction of AI for telemetry management 

Sawmills is the first telemetry management platform that uses AI to change the equation entirely. We sit between your engineering teams and your observability tools. Keep Datadog. Keep New Relic. Keep Splunk, Grafana, whatever you love and rely on. Sawmills works upstream, analyzing your telemetry data in real time as it flows, cutting the waste before it ever hits your observability bill.

The result is something that shouldn’t have to feel radical but does: dramatically lower costs, cleaner and higher-quality telemetry, and observability platforms that actually perform better because they’re processing signal, not noise. No rip-and-replace. No vendor lock-in. 

For the first time, companies can maintain full visibility with the tools they already trust while slashing costs, improving data quality, and safeguarding the reliability of their entire observability stack.

What Year One Proved

The numbers speak louder than anything I could write here.

Across our customer base, we’re delivering 50–80% observability cost reductions. We’re saving companies millions in annual observability spend. And we’re catching the things that used to cause real damage — runaway integrations, noisy services, surprise bills that could blow a quarterly budget in a day — and stopping them with a single click, before they ever hit.

We're also helping teams get their telemetry house in order, enforcing consistent standards, eliminating incomplete and unreliable data, and giving engineering teams the instrumentation quality they need to detect and resolve incidents faster. The data is cleaner. The alerts are sharper. And incidents that used to stretch on are getting cut short.

Companies like BigID, HiBob, BigPanda, Via, and others we can’t name yet trusted us early. From fast-growing startups to large enterprises, they’re all proving the same thing: “collect everything, figure it out later” is dead. Smart telemetry management is what comes next.

What We Shipped

We've shipped a lot. We launched Smart Insights, AI that works directly inside the data stream, detecting waste patterns and surfacing one-click fixes as telemetry flows through the system. 

We shipped HotReload, the first tool for live telemetry processor updates with zero deployments, zero restarts, and zero delays, an industry first. 

We released AI log transformation, which automatically detects unstructured logs and converts them into structured, queryable fields without requiring any regex or redeployment.  

Each of these features was born from real conversations with real engineering teams wrestling with real problems. That’s how we intend to keep building.

Thank you

Year one was proof. Year two is about a massive leap forward that I believe will reshape how the entire industry thinks about telemetry management.

To every customer who trusted us early, you pushed us to build faster, think bigger, and stay relentlessly focused on what actually matters. To my co-founders Amir and Erez, and our entire team, building a new category with you has been the ride of a lifetime, and we’re just getting started. To Team8, Mayfield, and Alumni Ventures, we thank you for betting on us before we had a logo.

The world generates more telemetry data every single quarter. Most of it is noise. We exist to cut the noise and keep the signal, and to give engineering teams the power, the visibility, and the control they deserve.

Year one: proof. Year two: we change the game.