Is the AI SOC a reality, or just vendor hype? In this episode, Antoinette Stevens (Principal Security Engineer at Ramp) joins Ashish to dissect the true state of AI in detection engineering.
Antoinette shares her experience building detection program from scratch, explaining why she doesn't trust AI to close alerts due to hallucinations and faulty logic . We explore the "engineering-led" approach to detection, moving beyond simple hunting to building rigorous testing suites for detection-as-code .
We discuss the shrinking entry-level job market for security roles , why software engineering skills are becoming non-negotiable , and the critical importance of treating AI as a "force multiplier, not your brain".
Guest Socials - Antoinette's Linkedin
Podcast Twitter - @CloudSecPod
If you want to watch videos of this LIVE STREAMED episode and past episodes - Check out our other Cloud Security Social Channels:
If you are interested in AI Security
, you can check out our sister podcast - AI Security Podcast
Questions asked:
(00:00) Introduction(02:25) Who is Antoinette Stevens?(04:10) What is an "Engineering-Led" Approach to Detection? (06:00) Moving from Hunting to Automated Testing Suites (09:30) Build vs. Buy: Is AI Making it Easier to Build Your Own Tools? (11:30) Using AI for Documentation & Playbook Updates (14:30) Why Software Engineers Still Need to Learn Detection Domain Knowledge (17:50) The Problem with AI SOC: Why ChatGPT Lies During Triage (23:30) Defining AI Concepts: Memory, Evals, and Inference (26:30) Multi-Agent Architectures: Using Specialized "Persona" Agents (28:40) Advice for Building a Detection Program in 2025 (Back to Basics) (33:00) Measuring Success: Noise Reduction vs. False Positive Rates (36:30) Building an Alerting Data Lake for Metrics (40:00) The Disappearing Entry-Level Security Job & Career Advice (44:20) Why Junior Roles are Becoming "Personality Hires" (48:20) Fun Questions: Wine Certification, Side Quests, and Georgian Food