Clinicians should participate in every phase of developing and training new AI tools to guarantee the technology meets real clinical needs, according to Jay Anders, chief medical officer at Medicomp. Anders delivered the message during a HIMSS TV interview at HIMSS26 on June 26, 2026, where he argued that leaving clinicians out of the process leads to tools that frustrate users and fail patients.
"Clinicians know what they need AI tools to do, so they should be involved at all stages of development and training of new AI technology to ensure it works well for them and for patients," Anders said.
Building AI that clinicians actually want
The call comes as health systems deploy more AI-powered clinical decision support, ambient scribes, and diagnostic aids. Many of these tools are built by engineers with limited exposure to clinical workflows. The result is often technology that adds friction rather than reducing it. Anders' statement underscores a shift in thinking: effective AI is not just about model accuracy, but about fit within a busy, high-stakes environment.
Medicomp, which produces clinical relevancy and decision-support systems, has long advocated for physician-led design. The push for AI tools that reflect real clinical needs is a growing theme in AI for Healthcare. At HIMSS26, multiple sessions addressed how to bridge the gap between data scientists and end users. Anders' remarks place an explicit emphasis on training - not just development - as a shared responsibility.
Training as a continuous loop
Involving clinicians in training means more than asking them to label data. It means using their feedback to refine model behavior, adjust alert thresholds, and ensure the AI's outputs align with the reasoning providers use in practice. Without that loop, AI tools risk automating outdated or irrelevant patterns. Anders pointed out that clinicians are the best judges of what information is actionable and what is noise.
Why this matters for healthcare professionals
Healthcare professionals who evaluate or purchase AI tools should insist on seeing evidence of clinician involvement throughout the product's lifecycle - from initial design through ongoing training updates. The absence of that involvement is a red flag. Anders' message is a practical reminder that the people who use the technology are the ones best positioned to make it work safely and effectively at the bedside.
Your membership also unlocks:
