Arcanna

    Authors

    Denis Stefan

    AI Engineer at Arcanna

    Denis builds the interface that lets users design agentic workflows inside Arcanna — the place where a complex investigation gets decomposed into a chain of agents, each with the right tools and guardrails.

    He spends his time where AI engineering meets the operational reality of production SOCs. When he talks about what makes agentic investigations fast, he's talking about systems he built and watched run.

    "The AI agent will not output the same result two times in a row — so you need a decision layer that decides whether the output of the agent is trustworthy or not."
    Denis Stefan
    Featured18 min watch · 7 min read
    SME Interview

    Inside an Agentic Investigation: An AI Engineer on Decision Models, Guardrails, and Trust

    Denis Stefan — AI Engineer at Arcanna and builder of the agentic workflow interface — walks through how a complex SOC investigation gets decomposed into a chain of agents, each with the right tools and decision model guardrails on the output. Why the same architecture that makes investigations fast is what makes them trustworthy.

    Full transcriptVideo + written
    Watch the Interview →

    Recent Articles

    How Agentic Investigation Cuts SOC Investigation Time From 40 Minutes to 3Agentic InvestigationJuly 17, 2026 · 6 min read

    How Agentic Investigation Cuts SOC Investigation Time From 40 Minutes to 3

    The forty minutes an analyst spends fetching logs and writing up one alert, run as a two or three minute agentic chain: grounded, traceable, and safe.

    Denis Stefan
    Denis Stefan · AI Engineer
    AI in a Splunk SOC: Agentic Investigation Without Rip-and-ReplaceAI SOCJune 26, 2026 · 6 min read

    AI in a Splunk SOC: Agentic Investigation Without Rip-and-Replace

    Adding agentic investigation to an existing Splunk stack, in your environment, grounded by decisions you control.

    Denis Stefan
    Denis Stefan · AI Engineer
    When Analysts Can't Keep Up with Low-Priority Alerts: What AI Triage Actually FixesAlert TriageMay 28, 2026 · 6 min read

    When Analysts Can't Keep Up with Low-Priority Alerts: What AI Triage Actually Fixes

    An analyst opens a queue of hundreds of low-priority alerts — what's actually going wrong, and can AI fix it? Why rule tuning, SOAR, and LLM assistants stall.

    Denis Stefan
    Denis Stefan · AI Engineer
    Inside an Agentic Investigation: The Architecture That Actually WorksAgentic AIMay 7, 2026 · 9 min read

    Inside an Agentic Investigation: The Architecture That Actually Works

    How agentic workflows are built — agent structure, tools, and getting it right in production.

    Denis Stefan
    Denis Stefan · AI Engineer

    See Arcanna in the SOC

    Decision models and agentic workflows, built on your team's own judgment.

    Get a Demo →