AI is now built into most SIEM, XDR and SOAR products, and every security team is being asked what it is getting for it. This course teaches the work underneath the marketing. It starts with where AI genuinely helps a SOC and where it fails, how LLM security assistants are wired to your data, and how they differ from the machine learning anomaly detection that SOCs have used for years. You then apply LLMs to the core analyst loop: triaging and enriching alerts, summarizing incidents into timelines, and keeping every claim grounded in evidence so hallucinated indicators never reach a block list.
The middle of the course is hands-on query and detection work: turning analyst questions into KQL, SPL and ES|QL, validating generated queries before trusting their results, running hypothesis-driven hunts, drafting Sigma rules with an LLM, and testing them against true-positive and benign data. The final modules survey the vendor landscape as of October 2026 (Microsoft Security Copilot, Gemini in Google Security Operations, CrowdStrike Charlotte AI, SentinelOne Purple AI, Splunk's AI capabilities and open-source options), show how to put LLM steps into SOAR playbooks behind human approval gates, cover prompt injection through logs and emails and the handling of sensitive telemetry, and explain how to measure MTTD, MTTR, analyst time and false-positive rates without fooling yourself. A capstone lab ties it together on a synthetic phishing-to-endpoint intrusion dataset you can build and run on a laptop.