Julia Rehman on who owns healthcare AI decisions

Healthcare organizations today possess an unprecedented amount of data, supported by investments in artificial intelligence (AI), predictive analytics, clinical documentation tools, and sophisticated dashboards. Yet, many leaders struggle to turn this wealth of information into actionable decisions that enhance patient care, support caregivers, and improve organizational performance.

In a recent episode of "I Don’t Care" with Dr. Kevin Stevenson, healthcare executive Dr. Julia Rehman shared insights on health data governance, AI, and digital transformation. Drawing from her experiences across the United States, India, Saudi Arabia, and the United Arab Emirates, Dr. Rehman explored how healthcare entities can harness AI without losing sight of the fundamental human element of care.

Balancing AI with human judgment

The key to effective AI integration in healthcare, according to Dr. Rehman, is to resist treating it as a catch-all solution. Instead, leaders should focus on specific operational challenges such as staffing, readmissions, bed capacity, and patient flow, where AI can significantly aid decision-making. Dr. Rehman emphasizes that while AI can help identify patterns and provide insights, "there is a human in the loop when it comes to decision making." The technology should support, not replace, human intuition and judgment.

This is particularly crucial for rural and underserved healthcare organizations. Limited resources in these areas mean that efficiency is paramount. AI can help extend capabilities, manage risk, and streamline decision-making processes. However, Dr. Rehman cautions that these benefits cannot be realized unless frontline employees are engaged during the selection and implementation process. Without buy-in and proper training, staff may circumvent the tools, leading to ineffective outcomes and wasted investments.

The necessity of governance

As AI embeds itself deeper into healthcare operations, robust governance structures become essential. Dr. Rehman insists on critical questions such as: Who is accountable for AI tools and their decisions? What processes exist for addressing questionable outputs? How will organizations monitor for bias and performance drift, and ensure that there is a reliable audit trail? These questions assign responsibility across clinical, operational and technical teams. For external context, NIST’s voluntary AI Risk Management Framework offers a structure for incorporating trustworthiness into AI design, use and evaluation. Its framework and companion playbook, linked in Sources, are general risk-management resources; they do not validate a particular clinical system or the interview’s claims about adoption.

Dr. Rehman notes that few healthcare organizations are adequately prepared to govern AI to the extent necessary. Many are already adopting AI features in electronic health records and other tools without a comprehensive framework for oversight. This could lead to adoption outpacing accountability, with potentially adverse effects on patient safety and organizational efficacy.

From data richness to strategic insight

A prevalent issue in healthcare is not the lack of data but the challenge of converting data into strategic insights. Dr. Rehman describes many organizations as "data-rich and insight-poor," where data is often relegated to a reporting function rather than being leveraged as a strategic asset. Leaders can sometimes dismiss data as inaccurate without assigning ownership and accountability for its validity and operational relevance.

Moreover, while data can enhance decision-making, it cannot replace the need for human judgment informed by experience, organizational culture, and community understanding. As the healthcare landscape becomes increasingly driven by technology, it remains crucial to remember that the core of healthcare is fundamentally about people, not just metrics.