March 2026 Summaries

Monks, a global digital services company, faced challenges with its legacy observability platform, which hindered its AI scalability and increased operational costs due to data sampling and management inefficiencies. Three Tree Tech helped Monks transition to Axiom's schema-less, serverless architecture, which provided comprehensive data visibility, reduced observability costs by 40%, and improved incident resolution time by 65%. The new system supports petabyte-scale data ingestion, ensuring secure and efficient AI workload handling, and was selected after a rigorous evaluation using Five Filters methodology, which assessed technology, process, financial viability, team alignment, and business context. The shift transformed Monks' data from a financial burden into a strategic asset, allowing the company to enhance its client-facing security services while maintaining robust data governance and security protocols.

Axiom's recently launched MetricsDB, now generally available, offers a unified platform for managing logs, traces, metrics, and events, enabling teams to query all machine data from one place without the traditional "metrics tax" associated with active time series. This release introduces a production-ready environment for ingesting, querying, and monitoring metrics with a pricing model that charges based on data volume rather than the number of dimensions tracked, thus facilitating greater data visibility without financial penalties. Axiom's Metrics Processing Language (MPL), designed for time-series operations, allows for code-first querying, making it suitable for AI agents that continuously monitor system anomalies. With MPL, users can easily transition from existing languages like PromQL, aided by a migration guide provided by Axiom. By leveraging object storage and ephemeral compute, MetricsDB efficiently manages high-cardinality data, eliminating the need for pre-aggregation, while its approach to pricing and management ensures scalability and cost efficiency for large-scale operations.

Axiom has introduced a new feature called Review and Issues, designed to streamline the process of identifying and addressing problems in AI-generated outputs by utilizing a queue-based workflow for reviewing conversations. Reviewers work through conversations, making binary judgments on their quality and documenting observations in their own words, which Axiom then uses to match with existing issues or create new ones without predefined categories. This process allows domain experts to capture nuanced insights that automated systems might miss, such as incorrect actions or outdated references. The tool builds a taxonomy of issues naturally from these reviews, logging every categorization decision for traceability and correction if necessary. Over time, the system helps teams identify recurring patterns, such as specific tool integrations causing incomplete actions, thereby guiding where to focus testing and improvements. The introduction of this feature completes a comprehensive toolkit that includes capability instrumentation, test case evaluations, production feedback collection, and live traffic scoring, thereby enhancing the ability to manage and resolve AI-related issues effectively.