NextFin News - In a strategic move that highlights the growing friction between massive data generation and enterprise cloud budgets, industry leaders Datadog and Figma have emerged as key backers for Sawmills AI, a San Francisco-based startup developing specialized tools for the next generation of AI-driven infrastructure. According to The Information, the investment comes as part of a broader $10 million seed funding round led by Team8, with participation from Mayfield and Alumni Ventures, aimed at addressing the "observability tax" currently hampering AI development.
The startup, co-founded by CEO Ronit Belson, CTO Amir Jakoby, and CPO Erez Rusovsky, officially emerged from stealth this February to tackle a specific bottleneck: the uncontrollable surge of telemetry data—logs, metrics, and traces—produced by modern software, particularly systems powered by AI agents. As U.S. President Trump’s administration continues to emphasize American leadership in artificial intelligence through deregulatory frameworks, the private sector is racing to solve the efficiency paradox where AI agents generate so much diagnostic data that the cost of monitoring them threatens to eclipse their operational value.
The core of the Sawmills platform is an AI-driven middleware layer built on the OpenTelemetry standard. It sits between a company’s applications and its observability vendors, such as Datadog or Splunk. By utilizing large language models (LLMs) and proprietary machine learning algorithms, the platform analyzes data streams in real-time to consolidate, deduplicate, and trim junk data. Belson noted that while enterprises often spend millions on observability, as much as 70% to 90% of the data transmitted is effectively "noise" that provides no actionable insight during a system failure.
This investment is particularly noteworthy because Datadog, a primary beneficiary of high data ingestion fees, is backing a technology designed to help customers send less data. This suggests a pivot in the SaaS ecosystem: vendors are beginning to prioritize long-term customer retention and "data quality" over short-term ingestion revenue. For Figma, the interest lies in the design and management of complex, collaborative systems where AI agents are increasingly used to automate workflows, requiring a more surgical approach to system health monitoring.
From an analytical perspective, the rise of Sawmills represents the birth of the "Telemetry Management" category. Historically, observability was a binary choice: either store everything at a high cost or risk missing the "smoking gun" log during a crash. However, the introduction of AI agents has broken this model. Unlike human-triggered events, AI agents can perform thousands of micro-actions per second, each generating telemetry. Without an intelligent filter like the one developed by Jakoby and his team, the cost of observing an autonomous agent could be 10 to 100 times higher than observing a traditional microservice.
Data from recent industry surveys indicates that 98% of companies have experienced unexpected spikes in observability costs, with many organizations now spending 20% to 30% of their entire infrastructure budget just on monitoring. By implementing "one-click" AI recommendations to convert millions of log lines into a single metric, Sawmills claims it can reduce data volumes by orders of magnitude. This concept of "telemetry sovereignty"—where the customer, not the vendor, decides what data is valuable enough to store—is becoming a central theme in 2026 enterprise architecture.
Looking forward, the success of Sawmills and its peers will likely trigger a wave of consolidation. As AI agents become the primary users of enterprise software, the infrastructure to monitor them must become as intelligent as the agents themselves. We expect to see major cloud providers integrate similar "pre-processing" AI layers into their native stacks. For now, the backing of Datadog and Figma provides Sawmills with the market validation needed to scale across mid-to-large enterprises that are currently drowning in the data exhaust of the AI revolution.
Explore more exclusive insights at nextfin.ai.
