A six-month practitioner evaluation surfaced consistent operational gaps in AI-assisted SOC tooling including model drift, elevated false positive rates in production environments diverging from vendor benchmarks, and undisclosed human-in-the-loop dependencies. This is not a vulnerability disclosure but an operational risk signal relevant to any organization using or procuring AI-assisted SIEM, SOAR, or detection platforms. Security leaders should establish production performance baselines before and after AI deployment, identify alert categories with high AI confidence but low analyst validation coverage as structural blind spots, and require vendors to document drift detection mechanisms and retraining cadence as procurement criteria.