Topic starter
11/11/2026 6:05 pm
TL;DR: 59% of security leaders expect AI to run major portions of their infrastructure autonomously within the next three years, according to Mezmo's report and supporting survey data. Observability now has to carry structured context that both humans and machines can reason over, or automation will inherit the same confusion.
Editorial analysis by NHI Mgmt Group, based on content published by Mezmo: “5 Things to Know About Context Engineering”.
Key questions
Q: How does active telemetry differ from passive logging?
A: Active telemetry is generated by an instrumented system while it is running, such as traces, metrics, and event records emitted by code or agents.
Q: Why do AI systems make observability mistakes worse?
A: AI systems make observability mistakes worse because they add non-deterministic decisions, hidden tool use, and changing context, so a missing event can conceal both the cause and the effect.
Q: How do I make telemetry reliable for AI agents?
A: Make telemetry reliable by instrumenting each agent action with correlation IDs, timestamps, input and output events, tool calls, policy decisions, and state changes, then preserve lineage, dependency, and state transition across every hop.
Practitioner guidance
- Define a context schema for operational telemetry Specify the minimum fields that carry service ownership, dependency, and change state so AI and humans reason from the same structure.
- Enrich signals at the edge Add context before telemetry is stored or forwarded, so downstream agents do not have to reconstruct meaning from raw events.
- Tag lineage and intent on critical signals Mark which service produced the signal, why it was emitted, and what state transition it represents, especially in automation-heavy environments.
Bottom line: Context engineering reframes observability as a governed source of operational truth rather than a passive event store.
Glossary terms: Active Telemetry, Distributed Context Graph, Agentic AI, Context Engineering
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
11/11/2026 6:06 pm
Context engineering is becoming a governance discipline, not just an observability pattern. Once AI systems start consuming operational telemetry, the question is no longer whether data exists but whether it is structured well enough to support correct action. That makes lineage, enrichment, and context fidelity part of the control environment rather than a tooling preference. The organisations that treat context as governed infrastructure will make better decisions, and the rest will automate confusion.
A few things that frame the scale:
- 69% of security leaders agree identity management must fundamentally shift to address agentic AI systems, according to the 2026 Infrastructure Identity Survey.
- 53% of security leaders expect AI to run major portions of their infrastructure autonomously within the next three years, according to the 2026 Infrastructure Identity Survey.
A question worth separating out:
Q: What should teams do first when context is missing from telemetry?
A: Start by identifying the minimum context needed to interpret each event, then instrument the missing fields at source, such as request IDs, actor identity, tool name, timestamps, and state before and after the action. If the source cannot provide them, add a context enrichment layer and treat the gap as an observability defect, not a logging preference.
👉 Read our full editorial: Context engineering makes observability a governable AI signal