Consultant Robert Hess III says AI is administratively helpful in healthcare, but it’s yet inappropriate for making clinical decisions.
“AI is doing a lot of stuff in the back of house, and that’s the biggest safety net at this moment,” says Hess, with Hess III Consulting in Phoenix.
Eligibility redeterminations, fraud detection, program integrity, patient outreach, appointment reminders and transportation coordination keep care moving. And AI is already proving itself in those spaces.
For example, the Big Beautiful Bill requires medical providers to redetermine Medicaid eligibility every six months.
“That’s where AI can help,” he says.
DEEPER DIVE: Read all the Ranking Arizona Top 10 lists here
INDUSTRY INSIGHTS: Want more news like this? Get our free newsletter here
Hess sits on national boards like the Safe AI in Medicaid Alliance; the Harvard Business Review Advisory Panel, where he serves as a subject matter expert and edits AI manuscripts prior to publications; and the Coalition for Healthcare AI. Those groups include representatives from OpenAI and Google to CMS and state Medicaid directors.
“They are shaping how AI can be used in healthcare and where it shouldn’t be used yet,” he says.
“There are tons of studies that prove the effectiveness of text messaging campaigns and outreach, even if it’s automated. Those things really do have a significant impact on people’s access to healthcare and healthy behaviors.”
Hess has lived this, as he is a legal guardian for an incapacitated adult. He’s navigated the refill calls, the surveys, the scheduling and the endless coordination. “It was interesting to see AI stepping into that role,” he says.
But he’s equally clear about where AI is failing — and where providers are wasting time.
“Everyone and their grandma are popping up with an AI agent,” he says.
Most healthcare providers in Arizona are small practices, and they’re overwhelmed by the flood of tools with no standards, no ratings, and no shared language. Some freeze and do nothing. Others jump in too fast.
Both choices, he says, are dangerous, as a person may receive a wrong diagnosis if he/she reaches out to AI.
“Healthcare, specifically in Arizona, is hanging on by a thread,” he says.
Shrinking Medicaid funding, declining ACA enrollment, workforce shortages, and rapid population growth have created a system under enormous pressure, he says. Clinics are closing. Nonprofits are merging to survive. Hospitals in rural areas are at risk.
“Providers are doing more for free or doing more with less,” he says.
“This is exactly where Hess says AI should be stepping in — not in diagnosing patients, but in stabilizing the operational backbone of healthcare. He points to one of the most urgent examples: Medicaid eligibility and appointments.
Non‑clinical tasks like social media, market research, grant writing, financial analysis, and drafting communications are safe places to begin.
“There are tons of research that proves the effectiveness of text messaging campaigns and outreach, even if it’s automated,” he says. “Those operational things are great opportunities for AI to help us preserve resources and scale the services that folks are offering.”
In that environment, adopting the wrong AI tool can be catastrophic. Hess shares a recent case where a major insurer used AI to deny claims, only to be unable to prove the decisions were accurate when questioned by federal officials. “There’s a lot of risk in adopting AI without governance,” he says.
He says he believes the first step for any healthcare organization is building guardrails. The U.S. Department of Health and Human Services recently released an AI strategy and governance framework, which Hess says, should be required reading. Organizations need to identify an internal person who understands AI, create a committee or advisory group, write policies, and train staff on what’s appropriate.
Only then, he says, should they start experimenting. Most providers haven’t even turned on Microsoft Copilot, however.
“You’ve got to dip your toe in the water and start playing with it,” he says.
What providers should not do, he stresses, is use AI for clinical decision‑making. The technology simply isn’t ready. He learned that while caring for his brother‑in‑law during the end stages of cancer. AI tools helped explain complex conditions, but they failed to ask follow‑up questions or consider missing information. “The AI was taking the inputs we were giving it and only relying on that information,” he says. “What a human would do is ask more questions.”
The inability to probe, contextualize, or read between the lines is why AI cannot replace clinicians. It’s often said “a Google search does not replace eight to ten to twelve years of medical training; and the same is true with AI (for now)” he says.
For Hess, the path forward is measured but optimistic. AI won’t fix healthcare, but it can help stabilize it if providers adopt it thoughtfully, with governance and guardrails.
“AI has the opportunity to give us a more cohesive experience,” he says. “But providers need standards and a responsible approach.”
