Grant Lunney, Product Lead at EverHealth – Interview Series

Grant Lunney is Product Lead at EverHealth, with nearly two decades of experience in healthcare technology, clinical operations, and product strategy. His career spans Boulder Community Health, OTTO Health, NextGen Healthcare, and CrowdHealth, with a focus on ambulatory care and the practical challenges medical practices face. At EverHealth, he works on applying AI across clinical documentation and practice workflows to give clinicians more time for patient care.

EverHealth provides healthcare technology spanning electronic health records, practice management, medical billing, and patient communications. Its EverHealth Scribe uses ambient AI to turn patient-provider conversations into draft clinical notes within the EHR for clinician review and approval. In this interview, Lunney discusses behavioral health documentation, patient consent and privacy, clinical oversight, and how AI can reduce administrative work for independent practices.

Your career has spanned nearly two decades in healthcare technology, from clinical operations and strategy at Boulder Community Health to product leadership at OTTO Health, NextGen Healthcare, CrowdHealth, and now EverHealth. How have these experiences shaped your perspective on the role AI should play in healthcare, and what attracted you to the opportunity to lead product innovation at EverHealth?

I’ve spent the better part of two decades working in healthcare technology, predominantly in the ambulatory space. During that time, I’ve watched a remarkable progression in the tools we’ve given physicians to help them practice medicine. I remember when electronic prescribing was a big deal. The idea that a physician could send a prescription electronically and have it ready by the time a patient arrived at the pharmacy was a meaningful advancement. We’ve since introduced increasingly sophisticated EHRs, process automation, interoperability, and all sorts of tools designed to make healthcare more efficient.

I think about AI much the same way. Whether we’re talking about large language models, generative AI, or predictive models, these are additional tools in our toolbox. The objective remains the same: make clinicians more efficient and ultimately improve the quality of care they can deliver.

Consider a primary care physician managing a panel of 2,500 or 3,000 patients. Their days are consumed by acute visits, medication management, and caring for patients with multiple chronic conditions. But I don’t believe for a second that those physicians wouldn’t love to spend more time working upstream with their patients. Talking about nutrition, exercise, behavioral health, and the social factors affecting someone’s health. Helping people make meaningful changes that might prevent the next chronic condition from developing in the first place. That’s why they went to medical school.

That’s what excites me about AI. We’re beginning to develop technology capable of taking meaningful chunks of administrative work and information processing off the clinician’s plate, giving them more time to practice medicine.

What attracted me to EverHealth is our ability to apply this technology incrementally across an entire medical practice. I believe strongly in solving one problem exceptionally well, then moving to the next. We can improve clinical documentation, then diagnosis recognition, then coding, orders, prescriptions, and practice operations. Over time, those improvements begin working together.

EverHealth Scribe uses AI to transform patient-provider conversations into structured clinical documentation. Can you walk us through how the technology works, from capturing a session to generating a clinical note, and explain what distinguishes your approach from other AI-powered medical documentation solutions?

At its simplest, an AI scribe listens to a conversation between a patient and clinician, uses speech recognition to understand what’s being said by whom, and then leverages a large language model to organize the clinically relevant information into a structured note. The clinician reviews that information, makes any necessary corrections, and approves the final documentation.

I think we’re reaching a point where most good ambient AI scribes are becoming remarkably capable of doing that fundamental task. Models differ in sophistication and accuracy, but I believe the bigger opportunity for differentiation will increasingly come from everything around that process.

That’s where DrChrono comes in. DrChrono runs the whole independent practice in one system, from the first phone call to the paid claim. AI such as Scribe works in every role, from the front desk to the exam room to the billing office, and the clinician always has the final word.

In the exam room, it starts with how much more of the clinician’s work we can complete once the note is generated. We’ve already added diagnosis and CPT code recommendations, and we’re expanding into orders and prescriptions, all built directly into the EHR. We’re also pulling in context from previous visits, medical history, and pre-visit documentation, because a 15-minute conversation is only one small part of a much larger clinical story. We’re also developing annotations that capture the observations and notes clinicians make that never get said out loud.

We’ve built reliability into all of those capabilities, too. If a clinician spends 30 minutes with a patient and then finds out the recording failed, that’s a catastrophic experience, so we have to account for everything from microphone placement to spotty Wi-Fi.

If we get all of that right, the clinician should finish an appointment with virtually all the documentation work already completed. Their remaining responsibility is to review the clinical content, exercise judgment, and act on the recommendations that make sense.

Mental health consultations often involve deeply personal conversations, emotional nuance, and sensitive disclosures. What are the unique technical challenges of developing an AI scribe for behavioral health, and how do you ensure the system accurately captures clinically relevant information without misinterpreting context or documenting unnecessary personal details?

A 50-minute psychotherapy session, a 15-minute behavioral health consultation, and a seven-minute internal medicine visit are three very different conversations. And the documentation that should come out of each encounter will look different, too.

So we start with the clinician and work backward. What information needs to be captured? How should it be organized? What level of detail is clinically appropriate? And what does the finished note need to look like for the provider to feel confident that it accurately represents the encounter?

Once we’ve established that, we can start configuring and refining the AI to produce documentation that aligns with those expectations. The goal is to take what might be a very long, complicated, and deeply personal conversation and distill it into the clinical information that’s actually relevant to the patient’s care.

And I think that’s particularly important in behavioral health. Just because something was said during a session doesn’t automatically mean it belongs in the patient’s permanent medical record. The model also has to interpret clinical meaning accurately, and in behavioral health, a single statement can mean very different things depending on context. That’s why the clinician’s review and judgment remain essential.

We spend a lot of time talking about how intelligent these models are, but I think some of the most important work happens before the model ever generates a note. It’s establishing what a successful clinical output looks like and then continuously improving how reliably we produce it.

Simply put, our process is to define what good looks like, work backward from there, and let the clinician be the final authority on whether we’ve achieved it.

Research from KFF found that 77% of U.S. adults are concerned about the privacy of medical information shared with AI tools. How does EverHealth address these concerns, particularly around patient data retention, third-party access, and the use of sensitive information for AI model training or improvement?

That concern is legitimate, and I don’t think patients should have to trade privacy for better care. At the same time, a technology company should not ask them to take its safeguards on faith. Patients and practices deserve clear answers about what information we use, who can access it, and how long we keep it.

With EverHealth Scribe, patient information is processed to support clinical documentation and continuity of care. We work with technology partners, including CarePilot, under contractual safeguards such as business associate agreements that govern how protected health information is handled. We also use encryption and access controls, and identifiable patient information is not used to train AI models. Scribe transcripts are available for up to 180 days, helping clinicians revisit relevant context across encounters, but practices can also request transcript deletion.

Ultimately, we can’t bolt privacy onto innovation after the fact. It has to shape the decisions we make about what we build, what we connect, and what we retain.

Mentavi Health, which has implemented EverHealth Scribe, obtains patient consent both during intake and before each visit. How important is this ongoing consent model in establishing patient trust, and what should happen when a patient declines AI-assisted documentation or withdraws consent during a session?

Mentavi’s approach of obtaining consent during intake and again before each visit reinforces that the patient has a say in how their care is delivered.

I spent a considerable part of my career in telehealth, so I believe patients are generally receptive to technology when they can see the value it creates. Ideally, patients experience more eye contact, more direct engagement, and a clinician who’s actually focused on the conversation instead of sitting behind a laptop or typing into a workstation. That’s a pretty compelling value proposition.

Of course, that value needs to come with choice. Patients should understand how the technology is being used and feel comfortable declining it. If someone withdraws consent during a visit, the recording should stop and the clinician should be able to continue the encounter and document it through their normal workflow. Any information already captured needs to be handled according to the applicable consent and privacy requirements.

Ultimately, I think trust comes from consistently delivering value. If we’re helping clinicians spend more time engaging with patients, and those patients can actually feel the difference, we’re doing something right.

AI-generated clinical notes introduce potential risks, including hallucinated information, omitted details, and inaccurate interpretations of patient statements. What safeguards has EverHealth implemented to identify these errors, and how do you ensure clinicians maintain meaningful oversight rather than simply approving AI-generated documentation?

Clinicians work hand in hand with the scribe and are responsible for editing AI-generated notes before adding them to the medical record, to ultimately limit the likelihood of any discrepancies in the output.

The clinician needs to review and edit the information before it’s committed to the medical record. They’re still ultimately responsible for determining whether the note accurately represents the visit and whether anything important was missed or misinterpreted.

The other major component is continuous feedback. We ask clinicians after every visit whether the documentation was helpful, and more than 96% of those responses have been positive. That’s an encouraging signal that we’re consistently delivering value.

When a clinician reports a problem, our team engages directly to understand what happened, identify the underlying issue, and determine how we can improve the experience. That direct relationship with the people using the technology is an incredibly important part of how we develop and refine the product.

I think that’s ultimately how you build trust in AI-assisted documentation. You make the underlying information accessible, preserve meaningful clinical oversight, and continuously evaluate whether the technology is actually helping.

Mentavi Health recently reported a phased implementation involving more than 30 licensed clinicians and over 14,000 AI-assisted clinical notes, with improvements in structured documentation and diagnosis-code capture. What have been the most significant findings from this deployment, and how do you measure whether AI scribes are genuinely improving clinician productivity, documentation accuracy, and patient care?

One of the most important lessons from Mentavi’s deployment is that successful adoption depends on much more than the technology itself. Mentavi brought clinical, administrative, and technological expertise together to define documentation standards, establish quality measures, and introduce capabilities thoughtfully. That collaboration is exactly what responsible healthcare innovation should look like.

For Scribe, the problem we are trying to solve is straightforward: clinicians spend too much time documenting an appointment. So how much of that time and effort can we eliminate?

We look at documentation time, clinician satisfaction, and how effectively features like diagnosis recognition and CPT code recommendations eliminate additional steps in the workflow. As we expand into orders and prescriptions, we’ll apply the same thinking. Each capability should remove another meaningful piece of work from the clinician’s day.

The improvement in diagnosis-code capture was particularly exciting. In the evaluation, encounters using Scribe had a diagnosis-code capture rate of nearly 80%, compared with approximately 26% without it. That’s a substantial improvement in the consistency of structured clinical documentation, and a great example of how AI can create value beyond simply generating a note.

The number of notes generated is encouraging, but the amount of work we’re eliminating, and the quality and consistency of the documentation we’re helping produce, are the outcomes that matter.

Independent medical and mental health practices frequently operate with limited administrative resources compared to larger healthcare systems. How can AI-powered documentation help smaller practices compete, and what obstacles related to implementation costs, workflow integration, or technology adoption still need to be overcome?

I think the biggest opportunity for independent practices is optionality. By taking administrative work off clinicians’ and staff’s plates, we give practices the freedom to make better decisions about how they spend their time and resources. Maybe a physician wants to see more patients. Maybe they want to spend more time with the patients they already have. Or maybe they simply want to finish their last appointment at five o’clock, spend 15 minutes reviewing documentation, and go home instead of spending another two hours catching up. All of those are valuable outcomes, and the practice should get to decide which matters most.

As for implementation, we’re already the practice’s technology provider, so workflow integration is our responsibility, not theirs. Scribe runs on devices clinicians already own, and getting started can take as little as one or two training sessions. That said, a small practice has very little room to absorb technical problems. So our goal is setup that’s as close to self-service as possible.

Ultimately, I think AI gives independent practices access to capabilities and operating efficiencies that historically required much larger organizations. And just as importantly, it gives the people working in those practices more control over how they spend their time.

With states introducing increasingly varied legislation governing AI in healthcare, particularly behavioral health, how does EverHealth navigate this fragmented regulatory environment? What challenges arise when developing an AI product that must accommodate different requirements for patient consent, data privacy, and clinician oversight across multiple jurisdictions?

I feel like I’ve seen this film before. I spent much of my career in telehealth before COVID, working through state-by-state differences in licensure, patient location, consent, prescribing, and reimbursement. Then the rules changed nearly overnight, and we had to scale while they kept changing.

The AI landscape feels similar. We are focused on staying ahead of proposed legislation, understanding the state and federal nuances, and building the product so it can adapt.

That matters most in behavioral telehealth, where a single clinician may see patients in many states. Consent, privacy, and clinician oversight have to be configurable foundations of the service.

The real lesson from that experience is that the product must be built with regulatory adaptability from the start. That’s the philosophy we bring to AI at EverHealth. Our job is to absorb as much of that complexity as possible so clinicians can focus on delivering care.

Looking ahead, AI capabilities are rapidly expanding beyond documentation into clinical decision support, patient engagement, and increasingly autonomous healthcare workflows. Where do you believe the boundary should remain between AI assistance and human clinical judgment, particularly in mental health, and how do you envision EverHealth’s technology evolving over the next three to five years?

AI can listen, organize, surface relevant context, and suggest. But clinical judgment belongs to the clinician.

That’s especially important in mental health, where a patient’s history, a change in tone, or even what’s left unsaid can be as meaningful as the words themselves. Technology should make clinicians more present for those moments, not less.

Over the next three to five years, the opportunity is much bigger than a smarter scribe. Our vision at EverHealth is an intelligent ambulatory care journey, where technology helps connect the moments before, during, and after a visit.

Imagine a clinic that remembers what matters: relevant history is available when it’s needed, clinicians can focus on the patient instead of piecing together fragmented information, and clinical decisions flow more coherently into documentation, coding, care coordination, and follow-through. We’re not just looking to make individual tasks faster. We want to reduce the friction between them.

Our North Star in everything we do is to ensure that every patient feels fully seen and every provider has the clarity to deliver confident, connected care.

Thank you for the great interview, readers who wish to explore AI-assisted clinical documentation and connected healthcare workflows should visit EverHealth.