Let’s talk about transforming healthcare. I had the chance to have some great discussions with Dr. Denise Basow, Chief Digital Officer at Ochsner Health, and Dr. David Carmouche, EVP and Chief Clinical Transformation Officer at Lumeris over the past few weeks. Lumeris is deploying its AI-enabled primary care as a service platform Tom at Ochsner Health.
So today’s send is a behind the scenes, exclusive first look on how the first of its kind partnership came together, and why Ochsner is ahead of the curve when it comes to primary care transformation, and what other health systems can learn from Ochsner’s enterprise AI strategy (and why the system prioritized primary care).
Primary care is on a collision course with multiple forces at play:
- A $50B federal bet on rural health.
- An aging, polychronic population that will completely outstrip today’s primary care labor capacity.
- And a new class of AI-enabled platforms that can finally take work off clinicians instead of piling more on.
If you want to see how those forces fit together, you end up in Louisiana, talking to forward thinking leaders like Dr. Denise Basow and Dr. David Carmouche, learning everything there is to know about their AI-enabled primary care as a service platform, Tom, and the partnership between the two organizations – of many to come – that will transform primary care for the better.
Let’s dive in.
The 27-Hour Day: Why AI Enabled Primary Care Is No Longer Optional, and what Ochsner and Lumeris are Doing About It
When Ochsner recruited Dr. Denise Basow as Chief Digital Officer, they were not selling comfort.
Louisiana has some of the highest rates of obesity, diabetes, and chronic disease in the country. Medicaid mix is heavy. Social risk is everywhere. Rural hospitals are fragile. Traditional fee for service primary care does not work well in that environment, and the usual answer of recruiting more PCPs and APPs was already faltering.
Dr. Basow came in as a former practicing PCP who then spent 25 years building UpToDate into one of the most widely used clinical tools on the planet. She knew that guidelines and content were necessary, but nowhere near sufficient, in places where the basic math of primary care was broken.
Today’s primary care math is brutal.
A typical PCP panel sits around 2,000 patients. And if you actually try to deliver guideline based preventive, chronic, and acute care while handling the never ending inbox, the work adds up to ~27 hours per day. That’s…more hours than we have in a day. Stating the obvious.
This level of burden is the baseline for any state and the simple reality of the economics of primary care (just ask Wal-Mart – sorry, Dr. Carmouche. I kid). Now layer on Louisiana’s burden of heart disease, obesity, diabetes, and the underappreciated wave of older adults with multiple chronic conditions who are about to show up in even larger numbers. The result is a perfect storm for primary care access and an unsustainable economic trajectory for both Louisiana but across the nation and particularly in rural settings.
In New Orleans or Baton Rouge, Ochsner can blunt some of that pressure with scale. In rural parishes, resources are stretched thin. Distances are longer. Transportation is weaker. Broadband is spotty. The nearest primary care visit can mean hours on the road.
Ochsner understood very clearly that they could not “out hire” the problem. They had to figure out how to care for more people without simply throwing more people at it. Which brings us to two developments converging right now:
- First: AI platforms are finally moving from ambient scribing pilots to more complexity: in this case, actual clinical workflow transformation in primary care.
- Second: CMS just dropped $50B for rural health transformation, and all 50 states are scrambling to figure out how to deploy it before year-end deadlines.
The companies that can credibly answer both problems simultaneously are positioning themselves at the center of where healthcare is headed. And what gets built for rural America will not stay there. Because the 27-hour problem exists everywhere, and health systems across the nation are trying to figure out how to expand primary care panel sizes and solve for labor constraints and access problems in primary care.
The platform architectures and payment models that work for patients two hours from the nearest provider will scale to suburban and urban populations facing the same fundamental primary care capacity constraints. Rural health is just the beginning, fam
So with that in mind, here’s how one of the country’s most innovative health systems is thinking about AI transformation, why they partnered with Lumeris and are deploying Tom for AI-enabled primary care as a service, and how rural health transformation could lead to something much greater in healthcare: a sustainable path forward.
Building the Groundwork Before AI Got Sexy
Well before Tom, Ochsner was already behaving like a system that assumed clinicians were the scarce resource.
They built digital medicine programs for hypertension and diabetes that live largely between visits. Patients measure from home. Remote teams monitor trends, adjust meds under protocol, and escalate when something is off. The PCP still owns the relationship, but a lot of the day to day chronic disease work runs on a separate track, offloaded to alternative teams or means of communication.
They turned telehealth and virtual care into normal access points, not side projects. Patients in rural communities can see Ochsner specialists on screens inside local hospitals instead of driving across the state. Asynchronous e-consults let rural clinicians get specialist input without giving up their patient.
To build trust and deliver meaningful technological advancement, Ochsner went right for the most pressing clinician pain points.
Ambient documentation with DeepScribe was not a “cool pilot.” It was a lifestyle change that said, ‘we see your 10 p.m. note writing and we are going to fix it.’ Rolling that across thousands of clinicians built equity with the medical staff. Pairing that effort with inbox reduction sent a clear message that Ochsner was willing to use technology to give time back, not just to extract more productivity.
That trust became the foundation for more complex initiatives.
Behind the scenes, Dr. Basow and her team were getting more pragmatic about what to build and what to buy.
Epic’s AI roadmap is real. Every health system in the country seems to have gone to UGM and came away thinking Epic is doing everything now. Waiting on the core vendor is the safest political move, and in some domains, that is the right answer. Nobody gets fired for waiting on Epic, right?
But Ochsner only has a handful of data scientists (7, compared to hundreds at larger systems) and a long list of urgent problems. For strategically important areas where Epic will take years or where switching costs are low, they made the call to partner with earlier stage companies, accept some risk, and keep the EHR as the system of record rather than the sole source of innovation. Denial management is a current example. Epic’s robust RCM AI is probably 2-3 years from prime time. Ochsner has an urgent need now, so they partnered with an early stage company. If Epic’s solution catches up and equals or surpasses it, they are willing to switch.
The lesson from all that work was simple.
Technology is a small part of the job. The hard part is reworking workflows, operating models, and trust. Once you have done that, the specific technology becomes relatively fungible.
That mindset set the stage for Tom and created the perfect cultural alignment for a partnership to blossom.
Tom: AI as the Ultimate Primary Care Extender
While Ochsner was building those muscles, Lumeris was going through its own evolution.
After more than a decade in MA and population health, the company had seen the same pattern across markets. Everyone talks about primary care as the front door. Very few organizations have the operating model or economics to make that front door actually work.
Carmouche and his team decided to wrap everything they had learned about risk, outcomes, and next best actions into a new operating layer for primary care. That layer became Tom, and I wrote about the launch of Tom here.
Tom is built on a simple premise.
If an activity in that 27 hour primary care day does not require human judgment, AI or a non physician team member should be doing it. If an activity is repetitive outreach, tracking, reminding, or routing, a conversational agent should be on point, not a physician or APP.
In practical terms, Tom uses data from the EHR, claims feeds, labs, pharmacies, HIEs, RPM programs, and consumer sources. The platform builds a longitudinal view of each person and uses that view to generate and prioritize actions.
Those actions might look like: reaching out to a patient who is overdue for a screening, checking in between visits on a diabetic whose numbers are drifting, coordinating follow up after an ED visit, or surfacing a social barrier and connecting the patient to a local resource.
Some of those actions go through humans. Others are handled entirely by the platform. All of them are meant to keep the care team working at the top of their license.
From Ochsner’s perspective, the differentiation was not that Tom could run agents. Plenty of companies can spin up agents now. Agentic AI is becoming somewhat commoditized. No, Lumeris’ differentiation lies in the simple fact that they have real scars from operating under value based contracts and are bringing a decade-plus of population health data to train those agents on what actually works.
Just as importantly, Ochsner and Lumeris shared a vision for what primary care ought to look like.
Primary care is still relational. The physician and core team own the complex decisions and the human side of care. Tom is a named member of that team, taking work off the plate every day, not a novelty feature bolted onto a visit.
Dr. Basow’s team was not looking for the cutest agent demo. They were looking for a partner that sees primary care the way they do and is willing to co-design the operating model.
That alignment made the partnership feel less like vendor selection and more like plugging a new operating system into a model Ochsner had already been building toward.
The $50B Rural Tailwind
While these conversations and developments formed over time, Congress and the administration dropped a bombshell on the rural landscape.
The Rural Health Transformation Program created a $50B, 5-year fund that will pay states to rethink rural care models. You guys know HITECH. The stimulus that pushed EHRs into almost every hospital was roughly $26B. This rural health fund represents nearly 2x that infrastructure investment, targeted at 26M Americans who face materially worse health outcomes simply because of where they live.
Every governor was asked to designate an entity to submit a statewide plan. States had just six weeks to assemble coalitions, inventory rural needs, and propose a five year budget. An approved plan comes with a baseline allocation of roughly $100M per year, with the chance to draw more from a discretionary $25B pool if the proposal is strong and aligned with CMS priorities.
Here is where it gets operationally messy. Those dollars have to be committed by the end of 2025 and flow out on a fairly tight schedule. States are now figuring out distribution mechanisms in real time. Procurement processes? Grants? Bid out projects? The application language is submitted, but the implementation playbooks are still being written.
Dr. Basow has been in direct conversations with Louisiana leadership and CMS about deployment. Her read: CMS wants transformation, not subsidy for the status quo. Dr. Oz made it clear to state leaders that simply pouring money into traditional models that have not worked will not fly. Innovation is not optional. It is the entry ticket for funding.
What Louisiana needs to prioritize, per Basow: baseline technical infrastructure first (EHR modernization, broadband, wired facilities), then technology that enables synchronous and asynchronous care delivery to rural communities. Not displacing local hospitals, but augmenting what those hospitals can offer. The play is partnership.
The Coalition: Platform Meets Provider Network
Lumeris recognized immediately that the problems the Rural Health Transformation Fund was trying to address are the same problems Tom is built for, just turned up to 11.
In rural America, primary care is often hours away. Chronic disease is more prevalent. Cancers are diagnosed later. Mortality is higher. The absence of continuity and prevention shows up in the hardest metrics.
If you could design primary care from scratch for someone who lives two hours from the nearest clinic, you would never start from brick and mortar visits. You would start from something that looks a lot like an AI enabled, virtual first, team based model that ties into local assets when needed.
That is exactly what Lumeris and a set of partners proposed when they pulled together the Collaborative for Healthy Rural America.
- Teladoc brings a national virtual primary, urgent, and behavioral health workforce that can reach into every rural zip code. But a virtual urgent care visit is episodic. There is no longitudinal relationship, no personalized outreach between encounters. Livongo helps but is not a comprehensive primary care platform.
- Deloitte helps states design, implement, and manage large programs. They were already working with many states on their rural health applications.
- Nuna (a John Doerr portfolio company) plugs in chronic disease apps.
- Unite Us ties the whole thing into community and social services.
Tom sits in the center as the connective tissue, orchestrating outreach, follow up, and navigation across all those touchpoints, and integrating any local providers and hospitals that want to plug into the network.
The coalition is not a new joint venture. It is a blueprint. It gives states a way to turn abstract budget lines about “technology enabled primary care access” into a concrete program that can be implemented and measured.
When the group sat down with CMS leadership, the reaction was strong. One official reportedly said something to the effect that a group of them had been sitting around earlier in the summer trying to envision what a technology solution for rural access could look like, and the coalition just showed them it. Another suggested every state should have language for something like this in their application.
What resonated was not just the AI platform. It was the comprehensive approach from a company with VBC operating history, not just a tech vendor. CMS still believes deeply that payment models will ultimately reward outcomes. Having a partner that understands how to administer risk and has the data infrastructure to prove clinical value matters for the sustainability question that hangs over every infrastructure investment.
Sustainability and the Next Version of Primary Care Economics
The Rural Health Transformation Fund is being discussed publicly as a lifeline for struggling rural hospitals and clinics. Privately, everyone understands that the money is also offsetting Medicaid cuts and other payment changes that will squeeze margins.
If the dollars are used to buy more of the same, rural communities get a temporary reprieve and then drift back into the same pattern of closures and access gaps.
If the dollars are used as infrastructure capital to stand up new primary care models that can live on VBC contracts and modernized fee schedules, the story changes.
That is the bet Lumeris is making.
Use the five year funding window to deploy AI enabled primary care at scale. Prove that the model improves outcomes, reduces avoidable ED and inpatient utilization, and stabilizes rural hospitals through better managed downstream volume. Then work with payors and CMS to wrap new or adapted payment models around that infrastructure so it can support itself after the initial funding ends.
Here is the limitation Carmouche keeps running into: current billing codes require human time or human action. If you look at TCM, RPM, and CCM codes, the requirements specify that a human did something. AI generated activities do not qualify.
So what has to change?
That might mean updating existing codes so that AI can count toward certain requirements instead of everything being defined as human minutes. It might mean hybrid PMPM payments that explicitly pay for the Tom layer as part of primary care. It might mean VBC contracts where AI enabled teams take on performance risk with tight guardrails.
Carmouche floated an even more provocative idea in our conversation: Could an agentic AI platform have its own NPI number? Payments flow to the platform, distributed to participating providers and technology partners based on contribution to outcomes. Wild to contemplate, but he says those conversations are happening with CMS and CMMI. The door is at least cracked open for rethinking what constitutes a care team member.
Regardless of the exact mechanism, the direction is the same.
For the first time, the financial model has to acknowledge that AI is not a bolt on. It is a core part of the labor stack in primary care.
The Policy Wildcard: State Fragmentation Risk
The regulatory environment remains messy and unsettled. Some states (Illinois, California notably) are moving to restrict AI in clinical care, particularly behavioral health. Meanwhile, CMS is signaling aggressive openness to AI enabled solutions.
A proposal in OBBA for a 10 year moratorium on state based AI regulation got stripped from the final bill. That is bad news for anyone trying to build national scale solutions. If every state implements different AI healthcare regulations, compliance complexity becomes a significant drag on innovation.
Carmouche is pragmatic about it. When you sign up to do healthcare in multiple states, you sign up for regulatory complexity. They will figure it out. But uniform federal standards would be vastly preferable to a patchwork approach.
The broader reality: the cat is out of the bag. Consumers are already using ChatGPT and other LLMs for health questions. Your friends who are not in healthcare go to AI first for medical information. The movement is happening faster than any regulatory framework can keep pace with. Policy will eventually catch up, but the experimentation is happening now regardless.
Why Leadership Is the Real Constraint
All of this sounds like a technology story. It is really a leadership story.
Dr. Basow did not mince words on what separates systems that are actually moving from systems that are just buying AI.
There has to be explicit, sustained commitment from the top that digital and AI transformation are strategic priorities, not side projects. Ochsner’s CEO made AI and digital one of the system’s four core priorities, which allows the digital team to truly innovate. To propose and defend disruptive changes. Without that executive sponsorship, inertia wins every time.
Trust has to be earned in practical ways. Early wins like ambient documentation and inbox management are meaningful not only for productivity but for clinician lifestyle. Those wins buy goodwill for more complex changes like rethinking how primary care is delivered.
Implementation and change management have to be treated as first class work. Most AI projects fail because workflows are not redesigned, governance is weak, and operational leaders are not truly bought in. Ochsner has slowed or stopped deployments when there was not enough alignment, even when the tech worked.
Lumeris is operating under similar constraints on the client side. Health systems are being bombarded by AI vendors. Boards are demanding clear ROI stories. Executives are starting to recognize the danger of collecting a dozen disconnected point solutions. Many are waiting to see a few large, credible deployments of platforms like Tom before they make their own big bets.
In that environment, you only get to transform primary care if senior leadership is willing to pick a lane and live with the discomfort that comes with it.
2026 Outlook: The Future That Is Starting in Louisiana
Put it all together and you get a picture that feels less like speculative futurism and more like an early draft of the next decade.
A health system like Ochsner that has spent a decade building digital medicine programs, remote teams, virtual access, and clinician facing AI is now plugging in a platform like Tom to expand primary care capacity even further.
A company like Lumeris that grew up in the trenches of MA and population health is recasting its intellectual property and data into an AI enabled primary care operating layer instead of trying to sell another point solution.
A federal government that once spent tens of billions pushing EHR adoption is now pointing even larger dollars at rural health and is explicitly open to AI enabled models as part of the answer, as long as they are safe, equitable, and sustainable.
And a rural landscape that has been an afterthought for decades is suddenly the proving ground for what AI enabled primary care could look like when there is no choice but to design around distance, scarcity, and chronic disease.
Carmouche is genuinely excited about what is coming. A new paradigm in primary care economics and access feels within reach. The coalition work, the CMS conversations, the health system partnerships, it all points toward something that could fundamentally change how 26M rural Americans (and eventually everyone else) access primary care.
If this combination works in Louisiana and in a handful of early adopter states, you will start to see a very different baseline for primary care operations. Panel sizes that would have been unthinkable under the old model will be manageable. Access will be defined by responsiveness and continuity, not by the number of exam rooms.
Most importantly, primary care physicians and APPs will spend more of their time on the complex, high value work they trained for, while AI enabled systems quietly handle the reminders, check ins, nudges, paperwork, and logistics that used to fill those 27 hours.
Eighteen months from concept to deployed platform at an enterprise health system. If that pace continues, who knows what we will be talking about a year from now.
Here is what I keep coming back to: we are watching healthcare, of all industries, actually ship AI at meaningful scale. Not vaporware. Not pilot purgatory. Real platforms getting deployed at real health systems with real patients. The $50B rural health fund is creating urgency and capital to accelerate that deployment in communities that need it most.
Organizations like Ochsner, led by executives like Dr. Basow and partnering with operators like Lumeris, are showing what it looks like to move first. Health systems that want to remain relevant in primary care, rural or urban, will need to catch up or get barraged by the dust storm.
For forward thinking organizations like Lumeris, It is a really good time to be building in healthcare.
Join my Hospitalogy Membership! If you’re a VP or Director working in strategy or corporate development at a hospital, health system or provider organization, you will get a lot of value out of my community as I purpose-build the content, fireside chats, and conversations for this group. Apply Here.