In busy Bay Area settings, diagnostic decisions may be supported by software that routes patients, highlights likely conditions, or flags “attention needed” results for review. Problems arise when those tools are treated as a substitute for clinical judgment—or when the tool’s output is incomplete, misunderstood, or not properly verified.
Common patterns we see in cases involving automated workflows include:
- Imaging or test triage delays: results get routed through a workflow that slows interpretation or follow-up.
- Risk score over-reliance: clinicians may anchor on an algorithm’s suggestion instead of considering competing diagnoses.
- Documentation gaps: the chart may reflect what the tool emphasized rather than what was actually assessed.
- Follow-up instructions not acted on: abnormal findings aren’t escalated, despite clear “abnormal” indicators.
Importantly, an “AI mistake” is rarely the only issue. In California medical negligence claims, liability often turns on how people and systems handled the information available at the time.


