Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

AI won’t replace humans in healthcare

Four questions for leaders before you embed AI.

[Photo: Getty Images]

Profile photo of Krinx Kong

Chief Commercial Officer · Cognivia

Every conversation about AI seems to land in the same place: fear. Fear that AI will replace jobs, replace creativity, or even replace what makes us human.

I think that fear misses the point. AI won’t replace humans, but will replace low-value cognitive labor, which keeps us busy but rarely improves judgment.

In practical terms, from what I’ve seen working in clinical trials and healthcare analytics, AI can help compress tasks like:

  • Review a 20-page clinical publication and extract key endpoints and implications.
  • Transcribe and organize a 60-minute strategy call into structured action items.
  • Triage long email threads to isolate what actually requires a decision.
  • Rebuild status decks from existing dashboard data.

Research from McKinsey’s 2023 report on the economic potential of generative AI suggests that a significant portion of current work activities could be automated or augmented by AI, and that professionals spend a large share of their time searching for or recreating information. Generative AI can meaningfully reduce that friction.

The real breakthrough isn’t productivity but understanding, because AI excels at pattern recognition.

What AI can’t do

Yet to reach that understanding, we need human intervention. AI can detect patterns in data, but interpreting human behavior and context still requires human judgment.

In boardrooms and product reviews, I often hear some version of the same assumption: “If the model is accurate, users will adapt.” In clinical design discussions, non-adherence is frequently framed as a reminder problem. In enterprise deployments, resistance to new systems is often dismissed as a training gap.

The assumption is simple: If the system is correct, behavior will follow. It rarely does.

Years ago, I was involved in a pharmaceutical product launch. On paper, the product offered a sound strategy, clear positioning, and defensible forecasts. But when the plan was presented to sales leadership, the reaction in the room was more cautious than the confidence reflected in the launch materials.

The issue wasn’t the product itself. It was credibility. Sales leaders wanted an honest discussion about where the drug worked well and where it had limitations.

Once the conversation shifted from polished marketing language to a more transparent discussion of the product’s strengths and trade-offs, alignment improved, and the launch gained traction.

AI’s promise as a tool


The same pattern appears in healthcare AI. AI-driven adherence models and clinical trial monitoring dashboards can flag patients at risk of dropping out of a study based on declining engagement metrics.

First, the system might identify missed appointments. But it doesn’t automatically identify the dominant behavioral drivers behind them.

Yet this is important, as contributing psychological factors often surface before the missed visit, including perceived time burdens, doubt about whether the treatment is helping, disappointment when clinical expectations change, or erosion of trust with the study team.

When dropout is reduced to logistics, the underlying psychology is ignored. But when behavioral signals are quantified early, they can guide targeted, low-burden interventions without changing the protocol itself.

For years, organizations have treated human behavior as too inconsistent or subjective to measure reliably. It’s easier to model transactions than trust, or optimize workflow rather than motivation.

AI most often breaks down in the gap between intention or enrollment enthusiasm versus action, or sustained commitment. But behavior isn’t noise. It’s measurable if you choose to look for it.

When integrated properly, AI systems can surface patterns that were previously invisible. Workflow tools show teams consistently overriding automated recommendations, trial dashboards may reveal dropout increases after specific milestones, such as when patients experience early side effects or when the visit schedule becomes more demanding than expected, and feedback analysis may detect increasingly passive language from patients or slower response times to study communications.

These are behavioral signals, and AI can help expose them. It makes visible the hesitation, friction, and trust dynamics that shape outcomes that rarely show up in clean datasets.

4 questions for leaders

The next breakthrough in AI won’t come from bigger models alone. It comes from systems designed with behavioral awareness at their core, which account for how people actually think, decide, hesitate, and adapt.

Leaders embedding AI into their organizations should ask:

  1. Does this system account for human motivation, or just task execution?
  2. Where are users working around it, and why?
  3. What gap exists between stated preference and revealed behavior?
  4. What does resistance tell us about trust?

AI will continue to improve technically, but technical accuracy does not guarantee that clinicians, researchers, or patients will trust or use the system. Optimization does not guarantee adoption by clinicians or researchers, and predicting behavior does not mean patients will actually follow through.

The real question isn’t whether AI is intelligent. It’s whether it understands the humans it influences.

Krinx Kong is the chief commercial officer at Cognivia.