Grepper cuts observability spend by 90% with AI signal extraction

Grepper has developed a solution to address the rising costs of observability, a sector where companies typically allocate 5% to 20% of their infrastructure budgets. Jeb Naus, the company's representative, noted that the shift to microservices architectures caused telemetry data volumes to grow faster than the marginal reliability gains, rendering traditional observability strategies expensive. Grepper’s initial product addresses this by reducing telemetry costs by 90%. It achieves this by extracting signal from noise upstream in the data stream, rather than storing all raw data. Naus described the implementation as plug-and-play, allowing data to pass through the system to achieve significant cost reductions.

The company is now building a second layer of its platform that feeds these extracted signals into AI agents for troubleshooting. This approach differs from traditional AIOps tools, which typically start from lagging indicators like alerts and backtrack to root causes. Instead, Grepper uses predictive models to detect new patterns or anomalies in logs. These specific signals are then provided to AI agents to form hypotheses and verify causal chains. Naus argued that traditional AIOps fails because it attempts to solve complex root-cause analysis from lagging indicators. He stated that Grepper’s method has successfully prevented production incidents in internal and customer environments by identifying early warning signs that human operators might otherwise ignore.