Artificial intelligence-driven observability startup Sazabi Inc. is emerging from stealth mode this week with a platform that challenges conventional monitoring approaches by focusing exclusively on log data and using AI agents to automate analysis.

The company is positioning its technology as a departure from traditional observability stacks that rely on a combination of logs, metrics and traces. Instead, Sazabi argues that advances in AI make it possible to extract all necessary operational insight from logs alone, thus reducing complexity and storage costs.

“Fundamentally, logs are just events, metrics are aggregated events and traces are basically correlated events,” said founder Sherwood Callaway, whose previous ventures include financial software firm Brex LLC and phone call automation platform Opkit Inc. “We only accept logs, and we create metrics and traces from those logs on the back end.”

Callaway said the approach is rooted in his own practical experience debugging production systems, where logs are often the first source engineers consult. He described logs as “the most intuitive type of telemetry to use” and “the Occam’s razor of observability.”

Simpler approach

The company’s platform, scheduled for delivery by the end of the year, is designed to simplify observability by reducing the need for instrumentation and complex data pipelines. Rather than requiring teams to configure multiple monitoring systems, Sazabi collects log data and uses AI to interpret it at scale.

That approach “reduces the complexity around instrumentation and the complexity in our data pipeline and storage infrastructure,” Callaway said.

At the core of the platform is an AI agent that continuously analyzes log streams, identifies anomalies and determines whether issues warrant attention. The system also generates alerts automatically based on historical patterns.

“We have our agent running in perpetuity in the background, looking for anomalies, investigating them and deciding whether they’re worth your attention,” Callaway said.

Agents control how alerts are generated and routed, including what information is included and who receives the notification. The result is fewer alerts with higher relevance, Callaway said. “It results in many fewer alerts, alerts that have much higher signal and that are much more actionable,” he said.

Sazabi also features a conversational interface that allows engineers to query production systems using natural language rather than navigating dashboards. The system can answer questions such as “why is production down?” or “which commit is responsible?” and return results based on real-time data analysis.

“You can root cause in seconds what would typically require minutes or hours of digging through different screens in a traditional tool,” Callaway said.

The underlying architecture combines log ingestion with a storage-and-query layer optimized for AI workloads. Sazabi uses materialized views and summarized data representations to reduce the amount of information the system must scan.

Summary data

“We can take an hour’s worth of log data and summarize that into a much smaller package using language models,” Callaway said. “You only have to query the summary.”

AI enhances storage efficiency by determining which data should be retained, summarized or moved to lower-cost storage tiers.

Sazabi targets early-stage and growth-stage technology companies rather than large enterprises, reflecting Callaway’s background in startup environments and the fact that most enterprises already have multiple observability platforms.

Venture-backed tech startups are typically focused on rapid product development and may lack specialized observability expertise, he said.

The platform is designed to integrate with existing observability tools, allowing customers to adopt it without replacing their current infrastructure.  “Just change the endpoint and we’ll be receiving logs,” Callaway said.

While incumbent vendors are all adding AI features to their observability suites, Callaway argued that the shift toward AI-driven software development will require fundamentally new observability models.

“It will be very hard for legacy or incumbent vendors to completely retool their products,” he said.

Sazabi has not disclosed detailed funding information but confirmed it’s in the process of raising seed capital.

Image: Pixabay