Most health care leaders can explain why they’re adopting AI, but far fewer can demonstrate how their governance actually functions day to day, according to Madhu Reddiboina, founder and CEO of RediMinds, whose company recently became one of the first organizations to earn URAC Health Care AI Accreditation.
Why URAC Health Care AI Accreditation Demands Evidence, Not Belief
Reddiboina describes AI governance across the industry today as policy-heavy but operationally thin, noting that nearly any organization can produce a polished governance document, while very few can demonstrate that a specific AI output was actually reviewed, by whom, and with what authority to halt it. He explains that accreditation closes this gap precisely because it asks for proof rather than stated principles, examining governance, defined AI use cases, risk management, transparency, monitoring, and accountability across the full lifecycle of a system in production.
A High-Volume Environment Where This Mattered Immediately
Because RediMinds processes hundreds of thousands of determinations annually, Reddiboina says pursuing this accreditation was never a theoretical exercise, and he argues trustworthiness is fundamentally a property of the organizational foundation built around a model rather than an inherent quality of the model itself.
Why Accuracy Alone Doesn’t Equal Safety Within This URAC Health Care AI Accreditation Framework
Reddiboina identifies what he calls the most expensive misconception in health care AI: the assumption that a high-performing model is automatically safe to deploy. He explains that technical performance and safe, compliant real-world use are separate properties, and that risk depends heavily on how a given system reasons and what decision it’s being asked to make, meaning the same underlying model can carry very different risk levels depending on the workflow it’s placed into.
Why Treating All AI as One Risk Category Backfires
He warns that applying uniform governance across every AI use case results in over-governing tools that are essentially harmless while under-governing those that carry real consequences for patients or clinicians.
Building a Real AI Inventory Within This URAC Health Care AI Accreditation Process
According to Reddiboina, a genuine AI inventory goes well beyond a list of vendors; for each use case, it should capture the underlying purpose, the data involved, the decisions it influences, its assigned risk tier, a named owner, required oversight, and how its performance is monitored over time. He notes this exercise often surfaces meaningful gaps, such as tools with strong vendor documentation that were never actually validated against an organization’s own patient population.
Why Shadow AI Is a Useful Diagnostic Signal
Reddiboina points out that employees often turn to unauthorized generative AI tools when approved options fall short, and he cautions that banning these tools tends to reduce only reported use, not actual use, making shadow AI activity a valuable signal for identifying where official tooling is falling short of real clinical or operational needs.
What Genuine Accountability Requires Within This URAC Health Care AI Accreditation Standard
Reddiboina argues that “human in the loop” has become something of an empty phrase in the industry, and that in practice it must mean a specifically named, accountable individual with real override authority at the point where an AI output carries consequences for a patient or health plan member. Given the volume of determinations his organization processes, he explains that reviewing every single output isn’t feasible, so RediMinds instead applies tiered oversight that concentrates human review where the cost of an error is highest, while documenting override decisions at the case level so any determination can later be reconstructed.
Deciding Authority and Training in Advance
He recommends organizations decide in advance who holds authority to approve, limit, suspend, or retire a given AI system, and who bears responsibility for residual risk, so that the early hours of any incident response aren’t spent simply figuring out who is accountable. He also argues that training needs to be role-based and scenario-driven rather than generic annual compliance training, specifically addressing when human judgment should override the tool.
Why Reddiboina Frames This as a Pressure Test, Not a Finish Line
Reddiboina says the most valuable outcome of pursuing accreditation wasn’t the credential itself, but the shared language it created across clinical, legal, compliance, and operational teams, along with a clearer picture of which processes were genuinely repeatable versus dependent on a single person’s institutional memory. He notes that an external reviewer working from a defined standard uncovers issues an internal audit typically misses, and since models drift and vendors update their products between review cycles, ongoing monitoring and revalidation need to become standard operating practice rather than a one-time compliance project.
The Six Questions Reddiboina Says Every Leader Should Be Able to Answer
He closes with six diagnostic questions organizations shouldn’t wait for a regulator or adverse event to confront: whether they know everywhere AI is being used, whether a single named person is accountable for each use case, whether they can demonstrate how each use case’s risk was evaluated, whether employees are trained for the actual decisions they face, whether patient and member disclosures are clear and appropriate, and whether they can detect a problem and stop a tool quickly.
What This URAC Health Care AI Accreditation Account Means Going Forward
Given Reddiboina’s emphasis that regulatory frameworks for health care AI remain unsettled but are converging around transparency, auditability, human oversight, and documented attention to bias, organizations pursuing similar accreditation now may find themselves better positioned than those waiting to retrofit governance once formal regulations solidify. Health system and health plan leaders evaluating their own AI governance maturity may find his six-question framework a practical starting point for identifying gaps before an external reviewer, regulator, or adverse event forces the issue.
What to Watch Going Forward
As more health care organizations consider pursuing URAC’s Health Care AI Accreditation or similar third-party verification, industry observers will likely watch whether this kind of evidence-based governance model becomes a standard expectation across the sector, particularly for AI systems influencing coverage determinations or clinician payment decisions. Given Reddiboina’s argument that a qualified human should always stand behind consequential AI-driven determinations even as the technology accelerates, this URAC Health Care AI Accreditation experience may serve as an influential reference point for how other health care organizations structure accountability as AI adoption continues to scale across utilization review and revenue cycle workflows.
For more healthcare industry updates, insights and news, visit DistilINFO. Click here to subscribe to stay informed.
