Human-in-the-Loop: AI in Clinical Data Management

AI is already collapsing timelines that used to take clinical teams weeks. EDC builds that once took a data manager days now take minutes. Yet speed alone doesn’t make trial data trustworthy, and teams that treat AI as a replacement for expertise risk losing the judgment that keeps evidence decision-ready.

Koneksa’s Michael Mendoza joins the SCDM Podcast to make the case for the opposite approach. As AI takes over configuration and routine setup, the value of domain expertise doesn’t shrink, it shifts, toward real-time quality review, oversight, and the judgment calls no model can make alone.

Why Human Oversight Still Decides Trial Quality

AI has a translation problem of its own. For example, a model can stand up an EDC system in minutes, but it says little about whether the data that flows through it will hold up under regulatory scrutiny months later. Expert oversight closes that distance. Real-time quality review, exception handling, and clinical judgment show data managers doing what the model can’t, and human-in-the-loop review keeps automated output interpretable and audit-ready. Meanwhile, shifting data managers out of manual configuration and into oversight roles raises the ceiling on what a lean biometrics team can monitor at once. This matters most where trial complexity is rising fastest. In multimodal, multi-site, and adaptive trial designs, data volume and decision points multiply, so oversight has to scale with them. Keeping experts in the loop is what keeps that growing data stream consistent and trustworthy from first patient in through database lock.

Inside the Conversation

Why EDC build times are dropping from weeks to minutes, and what that actually frees data managers up to do
How the data manager role is shifting from configuration toward quality assurance and real-time oversight
What “human-in-the-loop” looks like in practice when AI is analyzing trial data as it comes in
Why domain expertise is the constraint that keeps AI-accelerated trials reliable, not just fast

Listen to the episode, or schedule a scientific consultation to talk through where AI and expert oversight fit into your own data management strategy.