Shadow AI Report: Actionable Insights

Addressing unsanctioned tool usage starts at the policy level. Establish clear risk policies for AI usage and tool approvals that reflect the organization’s ethos and business goals, which can be understood across roles and teams. Additionally, include clear processes for regular policy updates as AI technology advances and privacy requirements change.

Building a multidisciplinary team of providers and administrators can break down silos within organizations and create a culture of learning and greater collaboration. This can foster greater understanding between administration and providers about key point-of-care challenges, and help gain early buy-in for solution testing and advocating.

Not all AI tools are created equally or with the same industry-specific attention. Identify how small workflow efficiencies can be achieved with tools, and consider whether a more general administrative tool can suffice or whether a purpose-built solution tailored to the healthcare industry's needs and concerns is necessary.

To support adoption and mitigate shadow tool usage, policy communication must be widely disseminated and accessible in multiple locations for teams to reference. Additionally, users must clearly understand the reasoning for avoiding unsanctioned tools and the black box model— returning to core shared values of safety and security.

Training and educational sessions can help users clearly understand why it’s critical to avoid unsanctioned AI tools that can introduce bias and impact patient safety. This is especially important as younger providers enter the workforce and haven’t yet developed their professional clinical expertise to critically analyze AI outputs.

Long-term adoption and success require staying updated on the latest tool features and workflow needs. It’s crucial to maintain an open line of communication with providers on how AI tools are addressing—and not addressing—current challenges to stay ahead of potential unauthorized tool usage. Consider how this monitor and feedback step can inform AI policy refinement.