AI and the Future of Southern Oregon Medicine, Part 3 of 3
When AI augmentation arrives in Southern Oregon healthcare — and it will arrive, because it is already arriving at every major medical center in the country — there are two versions of what that looks like for a working family in Medford or Grants Pass.
In the first version, AI makes the existing system more efficient. Hospital billing systems get faster. Prior authorization workflows get automated. Administrative costs drop. Margins improve. The waitlist stays at six months. The family that can’t get a primary care appointment still can’t get one. The technology arrived, and the people who needed it most didn’t benefit.
In the second version, AI augments the workforce that exists. Nurse practitioners in this region are supported by tools that give them the diagnostic backing of a specialist. La Clinica’s waitlist shrinks because AI-augmented providers can handle more complex cases without specialist referral. The child whose tethered cord syndrome went undiagnosed for two years gets an answer in one session. The Medicaid enrollee whose undiagnosed autoimmune condition has been generating emergency room visits for three years gets identified and treated. The working family that has been using the ER as primary care because there’s no panel to join has somewhere else to go.
Both versions are possible. Which one happens depends on choices that specific institutions in this region are going to make, and on whether the people those institutions serve understand what’s at stake and demand the second version.
Isaac Kohane — chair of biomedical informatics at Harvard Medical School, editor-in-chief of the New England Journal of Medicine AI, and someone who has watched two previous generations of AI in medicine come and fall short — is direct about what he fears most. It isn’t rogue AI. It isn’t catastrophic diagnostic errors. It’s something more familiar and more likely.
His biggest fear is that hospitals use AI to pour concrete over existing practices. To absorb the technology in ways that entrench the current system rather than transforming it.
For Southern Oregon families, that fear has a specific shape. And understanding it is the first step toward demanding something better.
How Institutions Protect Themselves From Solutions That Would Help You
Large health systems are not, as a rule, malicious. Asante Health System spent $177 million on community benefit in fiscal year 2022 — 202 percent of the floor assigned by the Oregon Health Authority, the highest percentage in the state. That is not the behavior of an institution that doesn’t care about the community it serves. Providence operates with a mission rooted in care for the poor and vulnerable. AllCare and Jackson Care Connect exist specifically to coordinate care for the region’s most underserved populations.
But all of these institutions, regardless of mission, operate under the same structural pressure: large revenue, very small margins — typically one to two percent for major health systems — and enormous organizational complexity. When your margin is that thin, risk aversion is rational. Change is expensive. Disruption threatens the financial stability that allows the institution to function at all.
The result is a pattern that Kohane describes with precision: institutions adopt technologies that improve operational efficiency and defer or resist technologies that would restructure how care is delivered. AI that makes billing faster, that automates prior authorization, that reduces administrative overhead — that AI gets adopted because it improves the institution’s financial position without threatening its revenue model. AI that enables a nurse practitioner to do what a physician previously did, or that enables a new care delivery model to compete with the hospital’s primary care panel, threatens the revenue model and encounters resistance.
This is not a conspiracy. It is a predictable institutional response to a genuine structural pressure. But its effect on working families is real and measurable. Every year that AI augmentation of primary care delivery is deferred is another year that the waitlist stays at six months. Every year that the region’s claims data sits unanalyzed is another year that the population health patterns nobody has looked for go unfound. Every year that nurse practitioners enter the workforce without AI augmentation training is another year of care delivered at a lower ceiling than the tools would allow.
The Regulatory Lock-In Risk
The more sophisticated form of institutional resistance isn’t outright opposition to AI. It’s shaping the regulatory environment around AI in ways that protect incumbents.
Kohane watched this play out nationally when large AI companies — OpenAI working with Microsoft, Google with its own models — began publicly calling for AI regulation. On its face, this looks like corporate responsibility. In practice, Kohane identifies a second interpretation he calls regulatory lock-in: well-funded companies proactively shape regulation to require compliance mechanisms that only well-funded organizations can afford. The legal infrastructure, the liability frameworks, the certification requirements become barriers to entry that protect the biggest players from smaller, more nimble competitors who might otherwise displace them.
The same dynamic can play out in regional healthcare. If Asante and Providence and the major CCOs shape Oregon’s emerging AI governance framework — through their lobbyists, through their relationships with OHA, through their participation in legislative processes — they can build a compliance environment that makes it very difficult for the new care delivery models that would most benefit working families to get off the ground. Not by opposing AI. By defining what “safe” AI looks like in ways that only they can afford to demonstrate.
For working families in Southern Oregon, this risk is not theoretical. Oregon’s price transparency legislation — SB 1060, which would have codified existing federal hospital price transparency rules into state law — failed in the 2025 session because the hospital lobby organized opposition faster than diffuse public interest could be aggregated. The AI governance conversation is earlier in its development, which means there is still time to shape it in ways that serve community interest rather than institutional interest. But that window will not stay open indefinitely.
The Data Problem Is Your Problem
The most concrete obstacle between working families in Southern Oregon and the care that AI augmentation could provide is not technology. It is data access.
AI systems improve with data. The more comprehensive the training data, the more reliable the clinical augmentation. Building AI tools calibrated to Southern Oregon’s specific patient population — its disease burden, its demographics, its comorbidity patterns — requires access to the electronic health records, claims data, and clinical documentation that currently live inside Asante, Providence, AllCare, Jackson Care Connect, and the Oregon Health Authority.
Those institutions have strong incentives not to share that data. The patient who moves from Asante’s primary care panel to an AI-augmented nurse practitioner practice is a patient Asante no longer bills. Kohane describes health systems as the most data-hostile industry he has encountered — he pays hundreds of dollars per patient to extract records for his undiagnosed disease research because hospitals deliberately make extraction difficult. He is clear about why: “They’re keeping you captive.”
But here is what working families in Southern Oregon need to understand: you have a legal right to your own health data.
The 21st Century Cures Act, passed by Congress in 2016, established that patients have the right to access their health data programmatically — meaning not just the right to request a paper copy, but the right to have it delivered in a usable digital format. If you are a patient at one of the approximately 800 hospitals connected to Apple Health, you can authenticate your records and download your labs, medications, diagnoses, procedures, and wearable data directly to your iPhone. That is already available for tens of millions of Americans, including patients at Asante and Providence.
The gap between that legal right and practical reality is large, and it is not accidental. Most patient portal interfaces are not navigable, not searchable, and don’t display trends over time. The APIs that are supposed to make data portable frequently return unusable formats. This is the difference between a right that exists on paper and a right that patients can actually exercise.
For working families who want to take advantage of AI diagnostic tools — who want to bring their full longitudinal health record to a new provider, or input it into an AI system for a second opinion, or share it with a researcher working on the undiagnosed condition that’s been making them sick — the practical path is this: download your records from your patient portal using Apple Health or a similar aggregator app, keep a personal copy, and know that sharing it with a clinician using HIPAA-compliant AI tools (including the enterprise versions of GPT-4 running on Microsoft’s Azure cloud, which Asante’s Epic system already integrates) is legal and becoming standard at major medical centers.
This matters because it is one of the few places where individual action, right now, can change the data landscape that determines whether AI tools trained on Southern Oregon’s patient population ever get built.
The Falling Asleep at the Wheel Problem
There is one risk in AI augmentation that Kohane raises that working families need to understand, because it affects what you should demand from your providers and not just from health system executives.
When AI systems take on more of the cognitive work of clinical decision-making, clinicians face a new hazard: automation complacency. The tendency to defer to the system’s output, to stop applying independent judgment, to let the AI’s recommendation become the actual decision without interrogating it.
Kohane illustrates this with a non-AI example. In the 1990s, there were extensive debates about the appropriate dosing of ondansetron, a common anti-nausea drug. When electronic order entry systems became standard, the system’s default dose became, in practice, what 95 percent of physicians ordered. The clinical debate ended not because it was resolved but because the default replaced the reasoning.
AI will create this dynamic at a vastly larger scale. The clinician who follows an AI diagnostic suggestion without examining whether you, specifically, fit the pattern that AI was trained on is more dangerous than the clinician who reached the same conclusion through their own reasoning. The AI’s output reflects the population it was trained on. You are an individual with a specific history, specific circumstances, and potentially specific presentations that don’t fit the training data pattern.
For working families, this means something practical: AI augmentation is a reason to expect more from your clinical encounters, not less. A clinician supported by AI should be able to tell you what the AI flagged, why they agree or disagree with it, and what in your specific situation they considered that the AI couldn’t. If a clinician is using AI tools and can’t explain their reasoning beyond “the system recommended it,” that is a failure of the human-AI workflow, not a feature.
Oregon institutions building AI augmentation into clinical practice need to design human oversight into the workflow from the beginning, not retrofit it after the fact. That includes training clinicians to interrogate AI outputs, building systems that flag when a clinician has passively accepted a recommendation without active engagement, and measuring whether AI adoption is improving diagnostic accuracy or just shifting liability.
What Southern Oregon Families Should Demand
This series has documented the structural nature of the provider shortage, the tools that already exist to address it, and the institutional dynamics that determine whether those tools reach working families. The closing argument is specific: here is what working families in this region should expect from the institutions that serve them, and here is what it would mean in practice.
From Asante Health System: A public commitment to deploying AI-augmented primary care in ways that increase panel capacity, not just administrative efficiency. Asante’s AsanteForward2030 community investment campaign, with its $100 million goal and its 10,000-donor base, describes an institution that has framed its relationship to this community as something more than transactional. The question is whether that framing extends to using the AI integration already available through Epic to expand who can get an appointment and when. A concrete measure: how does wait time for a new primary care patient change over the next three years as AI tools are deployed?
From AllCare Health and Jackson Care Connect: A commitment to deploy the longitudinal claims data they hold for the region’s Medicaid population in ways that serve population health analysis. AllCare and JCC together hold some of the most comprehensive longitudinal health records in the region for the working families and lower-income households most affected by the access shortage. A partnership with OHA, a university research program, or a population health analytics company to mine that data for the undiagnosed patterns that current practice is missing would be a direct use of existing assets for the community that generated them. A concrete measure: within two years, what population health insights derived from regional claims data have been published, shared with clinical partners, and translated into changed care protocols?
From Oregon legislators and OHA: Substantive enforcement of the 21st Century Cures Act’s data access provisions. The difference between a nominal right to data and a practical one is the difference between a hostile API and a usable one, and OHA has both the regulatory authority and the community mandate to push health systems toward the latter. Additionally, the Oregon All-Payer Claims Database represents a regional asset that should be accessible for population health research that serves communities. The framework governing that access should be built in public, with community input, before the lobbying effort arrives to shape it in ways that serve institutional interests. A concrete measure: within one year, what percentage of Southern Oregon patients can download a complete, usable copy of their health record in under ten minutes?
From Southern Oregon University and Oregon Institute of Technology: Health sciences curriculum that prepares nurse practitioners, physician assistants, and community health workers to use AI tools from their first clinical placement. The OHSU nursing school on SOU’s Ashland campus and OIT’s clinical training programs are the regional workforce pipeline. Graduates entering practice in Jackson and Josephine counties who don’t know how to work with AI augmentation tools are entering the workforce at a lower ceiling than the academic medical centers they’ll be compared to. A concrete measure: by 2027, what percentage of health sciences graduates from regional programs have had hands-on training with clinical AI tools as part of their required curriculum?
The Decision Is Yours Too
Kohane’s closing advice to clinicians is to start using these tools now. To collect patient data — with consent — that can inform longitudinal analysis. To treat AI as what it is: a very smart, comprehensive assistant that can augment clinical judgment without replacing it.
For working families in Southern Oregon, the parallel advice is this: know your rights, use them, and make your expectations clear to the institutions that serve you.
You have the legal right to your health data. Download it. Keep a copy. Use it.
You have the right to ask your clinician what AI tools they’re using and how they’re using them — and to expect an answer that goes beyond the default.
You have the right to participate in public comment processes at OHA, in county health planning, and in the community governance mechanisms that determine how healthcare is organized in this region. The AI governance conversation is happening now, at the state and regional level. The people best positioned to shape it in the interest of working families are the working families themselves.
The technology to meaningfully change the healthcare access picture in Southern Oregon exists today. The data to train it on this community’s specific patient population exists today. The institutional infrastructure to deploy it exists today.
What has not yet happened is the decision — by health system leaders, CCO executives, state legislators, and the communities that hold all of them accountable — to use it for the people who need it most rather than the purposes that are most convenient.
That decision is not made in a boardroom alone. It is made, in part, by the people who show up, who ask the questions, and who make clear that the second version — the version where the technology actually changes what happens to the working family that can’t get an appointment — is the only acceptable one.
This is the third article in a three-part series on AI and the future of medicine in Southern Oregon. Part 1 examined the structural nature of the provider shortage. Part 2 addressed what AI augmentation can concretely do for working families in this region. ReImagine Healthcare is a subsidiary of Flourish Charity, a 501(c)(3) nonprofit. We publish research and analysis on healthcare system design in Southern Oregon. We welcome responses, corrections, and partnership inquiries at reimagine-healthcare.org.