Nov 9, 2026
By Nima Roohi
Why the convergence of public urgency and responsible AI could reshape the next decade of care.
Something important happened in American healthcare this week, although it was easy to miss amid the normal post-election analysis. The most interesting signal for me was not simply which party won more races. It was that voters elevated a striking group of people with deep healthcare experience into consequential public offices at the same moment that affordability, access, and the basic cost of living were among the issues weighing most heavily on them.
Ohio elected Dr. Amy Acton, a physician and former state health director, as governor. California elected Xavier Becerra, a former U.S. Secretary of Health and Human Services, as governor. South Carolina elected pediatrician Dr. Annie Andrews to the U.S. Senate, and Michigan elected physician and former public-health leader Dr. Abdul El-Sayed. Georgia sent Republican neurosurgeon Dr. John Cowan to Congress. They do not share a healthcare ideology, and we should not pretend that they do. What they share is firsthand exposure to a system that is expensive, fragmented, and often unnecessarily difficult for both patients and the people trying to care for them.
That matters because the election also reinforced something more basic. Americans are struggling with the cost and complexity of everyday life, and healthcare sits squarely inside that experience. In a Reuters/Ipsos survey conducted immediately before the election, 48% of registered voters identified cost of living as the most important factor in their vote. When voters were asked which everyday expense Congress should prioritize, healthcare ranked first at 26%, ahead of housing, gasoline and food. KFF polling before the election similarly found that 82% of voters considered healthcare costs very or extremely important for candidates to address.
I would be careful about calling that a mandate for any single healthcare policy. Elections are never that simple. But I do think it is fair to say that the public is sending a clear message about the problems it expects leaders to solve: care has to become more affordable, easier to access and easier to navigate. That expectation extends beyond the walls of a hospital. Food insecurity, unstable housing, transportation barriers and lapses in insurance coverage all shape health outcomes, which is why federal health agencies increasingly describe them as health-related social needs rather than unrelated social-policy issues.
The second force: a technology wave arriving at the right moment
At almost exactly the same time, healthcare is entering a new technology era. The public conversation about artificial intelligence often gravitates toward the most dramatic use cases: autonomous diagnosis, replacing clinicians, or allowing algorithms to make consequential coverage decisions. Those questions deserve serious caution. Healthcare is not the place for uncontrolled experimentation, and high-stakes decisions require strong clinical validation, security, privacy protections, auditability and clear human accountability.
But focusing only on the most controversial edge of AI risks missing the much larger near-term opportunity. Some of the most expensive failures in American healthcare are not scientific problems at all. They are operational problems. A patient loses Medicaid because a renewal notice was missed. A health plan calls the same member repeatedly because data is fragmented across systems. A nurse spends hours finding documentation that should already be available. A family qualifies for food or housing assistance but never completes the process. A physician finishes notes late at night because documentation has become another shift of work. These are mundane problems, but at national scale they are extraordinarily costly.
The size of that administrative burden is hard to overstate. The 2025 CAQH Index estimated that electronic transactions and automation helped U.S. healthcare avoid roughly $258 billion in administrative costs in 2024, while another $21 billion in annual savings remains available through fuller automation of work that is still manual or partially manual. CMS has estimated that prior authorization alone can consume roughly 13 hours of provider time each week. And we are beginning to see real evidence that well-bounded AI tools can return capacity to the system: a 2025 JAMA Network Open study across six health systems found clinician burnout falling from 51.9% to 38.8% after 30 days of ambient AI scribe use, alongside reductions in after-hours documentation and cognitive workload.
To me, that is the more useful way to think about the first wave of AI in healthcare. The breakthrough may not be replacing expertise. It may be removing the enormous amount of friction surrounding expertise, so clinicians can spend more time practicing medicine, public agencies can serve more people without proportionally increasing headcount, and patients can complete basic healthcare and benefit processes without becoming experts in bureaucracy.
Why this matters now
The timing is especially important because the administrative burden on patients, providers and states is about to increase. Beginning in 2027, new federal Medicaid rules will require many states to implement community-engagement requirements and more frequent eligibility verification for portions of the Medicaid population. KFF, citing Congressional Budget Office estimates, reports that the new work requirements account for 5.3 million of the projected increase in uninsured people by 2034. Separate changes to renewal frequency are also expected to increase coverage loss. Regardless of where someone stands on the underlying policy, there should be broad agreement on one principle: if a person remains eligible for coverage, losing that coverage because our systems failed to reach them, collect a document or process information efficiently is not a policy success. It is an infrastructure failure.
The same logic applies to SNAP, housing assistance, primary-care access, maternal health, chronic-disease management and care transitions. We have spent years building programs to address these needs, but the experience of accessing them is still too often defined by phone trees, paper forms, fragmented eligibility rules and manual follow-up. That is exactly the class of problem modern automation is becoming good at: multilingual outreach, repetitive screening, document collection, navigation, reminders, follow-up and escalation when a human is genuinely needed.
This is also where I think the political debate around AI needs more nuance. The choice is not between deploying autonomous systems everywhere and freezing innovation until every risk disappears. We can move at different speeds depending on the use case. We should move deliberately when technology is making a clinical, financial or eligibility decision with meaningful consequences. But when the task is helping someone understand a renewal notice, schedule an appointment, translate instructions, find the right community resource, collect missing paperwork or route a complex case to a trained human, the risk can be bounded and the benefit can be measured.
The safeguards are not mysterious. They include clear scope, strong security and privacy controls, testing before deployment, monitoring after deployment, transparent escalation paths, human oversight for consequential decisions and the ability to audit what the system did and why. Those principles are increasingly reflected in federal AI governance frameworks as well. Responsible deployment should not mean slow deployment by default. It should mean designing systems so that we know where they can act, where they cannot, and when a person must take over.
A rare window for the next decade
What makes this moment unusual is the convergence. Voters are elevating more leaders who have actually worked inside healthcare. The public is making affordability and access impossible to ignore. At the same time, the economics of solving administrative problems are changing because technology can now carry out a growing share of repetitive work that previously required expensive human labor.
There are already signs that this can become a bipartisan operating principle, even if the policy solutions remain different. Republican physician John Cowan said after his election that insurers and bureaucrats too often stand between doctors and patients. Democratic physician Amy Acton campaigned on the argument that too many people cannot both access and afford care. They may disagree on what government should fund or how healthcare markets should work, but both are describing a version of the same underlying problem: too much friction sits between people and the care they need.
That gives us a real opportunity over the next decade. Instead of debating AI only as a threat or a promise, we can ask a more practical question: where can we use it, safely and measurably, to make healthcare work better? There is no shortage of places to start. Eligibility and enrollment. Prior authorization. Documentation. Member engagement. Care transitions. Primary-care access. Chronic-disease follow-up. Maternal health. Social-care navigation. These are not futuristic use cases. They are the daily work of healthcare, and many of them are still being done with systems and staffing models designed for another era.
The outcome I hope we pursue is not an automated healthcare system. It is a more human one: a system where fewer people lose coverage because of paperwork, fewer clinicians spend their evenings documenting care, fewer families abandon benefits because the process is impossible to navigate, and scarce human attention is reserved for the moments where empathy, judgment and expertise matter most. We have the public urgency, the policy attention and increasingly the technology to make that possible. The next decade will tell us whether we used the window well.
Sources and further reading
• Reuters/Ipsos, November 2026 Midterm Survey Topline
• KFF, Health Care in the 2026 Midterm Election
• Associated Press, 2026 voter poll analysis
• Healthy People 2030, Social Determinants of Health
• KFF, Tracking Implementation of Medicaid Work Requirements
• Congressional Budget Office, Public Law 119-21 Medicaid analysis
• CAQH / DataSpring, 2025 Index findings on administrative automation
• CMS, Electronic Prior Authorization
• JAMA Network Open, Ambient AI Scribes and Administrative Burden
