For years, telemetry has been the quiet backbone of IT and security operations. The MELT stack — metrics, events, logs and traces — has powered observability platforms, incident response workflows and analytics engines. Traditionally, this data has been passive: Systems generate logs and metrics that are stored, queried and occasionally enriched by downstream tools. This approach tells us what happened but rarely explains why, and rarely guides what to do next.
As we enter the age of agentic AI, those limits become critical. Data is no longer operational exhaust for dashboards; it is the lifeblood that enables autonomous agents to reason and act. Agentic telemetry redefines the role of data by preparing, contextualizing, and making it actionable by design. By fusing system signals with human-generated context — tickets, wikis, Slack conversations, CI/CD metadata — telemetry becomes fuel for AI agents capable of triaging alerts, orchestrating workflows and even remediating issues in real time. These agents, combined with a “human-in-the-loop” mindset critical in the AI era because they ensure accountability, judgment, and ethical oversight where automation alone may fall short. By keeping humans engaged in reviewing outputs, providing feedback, and making final decisions, organizations can maintain accuracy, trust, and fairness in AI-driven processes. This balance of machine efficiency and human expertise creates stronger, more reliable outcomes and prevents blind reliance on algorithms.
Why Traditional Telemetry Fails in the AI Era
The challenge is not only exploding telemetry volumes but the absence of context that explains why signals matter. Telemetry is growing ~30% annually, not including the digital exhaust created by AI itself. Without tying it to human-generated context (tickets, wikis, runbooks, Slack threads), dashboards become noise, alerts overwhelm, and analysts burn out trying to connect the dots. Legacy schema-on-read/write systems cannot scale to AI workloads, collapsing under the demand for low-latency, high-volume queries.
Key Challenges
- Context gap: Machine telemetry shows what happened, but without correlated human or AI-generated data, investigations are slow, MTTR rises, and analysts churn under the workload.
- Unsustainable growth: Telemetry grows faster than budgets and resources. Ingestion-priced tools force teams to overspend or drop data, creating blind spots at critical moments.
- Legacy architectures: SIEMs and log platforms were built for humans reading logs, not AI-driven analysis. They enforce rigid schemas, create silos, and collapse when agents generate 10–100x the queries of humans.
The Agentic Telemetry Solution
The answer is an AI-first architecture designed for both agentic and human-driven workloads. In this model, telemetry is normalized, structured, and optimized for AI-scale use, while remaining federated and schema-agnostic. AI agents fuse this telemetry with human context to deliver explainable insights at machine speed. The result is both faster investigation and lower cost — supporting 10–100x the workload at a fraction of today’s expense.
Agentic telemetry applies intelligent agents throughout the telemetry lifecycle: Structuring data at ingest, normalizing schemas, and correlating across machine and human signals. This enables reasoning, enrichment and recommendation at scale. By moving from ingest to insight at the speed of AI, organizations transform investigation efficiency, contain costs and empower operators to become “10x investigators.”
Architecture for the AI Era
Traditional schema-on-read/write systems were built for humans reading logs and cannot handle AI-driven workloads. Agentic telemetry requires a new design:
- Structured at ingest: Normalize and enrich data as it flows in, eliminating slow and brittle downstream parsing.
- Schema-agnostic and federated: Support OTLP, OCSF, ECS and custom schemas across heterogeneous data stores.
- Built for AI-speed queries: Lakehouse-style storage and query engines optimized for orders-of-magnitude higher query volumes.
- Context fusion: Integrate machine telemetry with human context to explain both what happened and why.
From Context to Insight
Telemetry alone tells you what happened. Human context — tickets, Slack conversations, CI/CD metadata, pull requests — tells you why. Agentic telemetry fuses both streams with AI agents providing the reasoning layer. This shift eliminates 90% of investigation time currently wasted hunting for context and instead surfaces explainable, actionable insights. Teams can spend time solving problems, not correlating data.
Open, Flexible, and Future-Proof
Agentic telemetry is defined by openness and adaptability:
- Support for multiple schemas (OTLP, OCSF, ECS, custom).
- Interoperability across diverse data stores.
- Compatibility with a range of AI agents.
As organizations shape long-term AI strategies, they need standards that avoid vendor lock-in, support choice, and evolve with the ecosystem.
Benefits Delivered
- Unlimited scalability: Structured storage allows agents to issue thousands of queries instantly without overwhelming query engines.
- Reduce ingestion costs: Federated, schema-agnostic architecture enables querying data in place, lowering storage and licensing costs.
- Empower operators: Agents handle enrichment, correlation and recommended actions, freeing engineers to focus on high-value decisions rather than repetitive tasks.
- Faster Investigations: AI agents can analyze alerts, correlate with past incidents and documentation, and execute actions— all while keeping human-in-control.
Conclusion
Agentic telemetry is not just about managing logs or metrics; it is about re-architecting telemetry for the AI era. This architecture transforms telemetry from passive exhaust into AI-grade fuel. It accelerates investigations, reduces costs and enables every operator to act as a “10x investigator.” In the age of agentic AI, telemetry must evolve — and agentic telemetry is the path forward.
KubeCon + CloudNativeCon North America 2025 is taking place in Atlanta, Georgia, from November 10 to 13. Register now.