The hardest part of bringing AI into a healthcare organization isn't the AI. It's everything that happens before the first provider ever logs in - who drives adoption, what they do day to day, and why so many rollouts stall without the right playbook and on-ground partners driving it. That's the part almost no one talks about.

So we sat down with Amanda McFayden, Director of Clinical Operations at DENT Neurologic Institute, the largest outpatient Neurology center in the country. Amanda has spent nearly 20 years at DENT, working through almost every corner of the organization before stepping into her current role. She's watched healthcare technology evolve across two decades - and over the past few quarters, she led the rollout of Marvix AI at DENT, growing usage to north of 90% adoption. 

Our conversation ranged widely: the role of AI in neurology, how to drive large-scale adoption of new software, guidelines for administrators and leadership to shape a rollout, and more. Here's her candid take. You can watch the entire podcast below:

Key Highlights from the Podcast:

AI in Neurology: Why Large Organizations Are Moving Now

For DENT, AI wasn't a nice-to-have - it was a response to a breaking point. Growing documentation demands, tougher insurance requirements, and staffing gaps that never closed after COVID had pushed providers to the edge. Leadership knew the usual fixes such as EMR triggers, manual workarounds wouldn't hold much longer.

So what tipped the scales toward AI scribes, and why was burnout the problem they cared about most? Amanda breaks down the turning point that made adoption not just smart, but necessary.

Cheat Sheet: Evaluating AI for Your Practice

1. Name the problem first. Start with the specific pain point you're solving—"general interest in AI" isn't a reason to buy.
2. Separate staffing problems from provider problems. They often look similar but need very different solutions.
3. Align leadership on the goal before you talk to any vendor. Clarity upfront prevents misdirected demos and wasted cycles.

How DENT Built a Checklist for Choosing the Right AI Solution

Before speaking to a single vendor, DENT's leadership did something most organizations skip: they defined exactly what they needed. The result was a structured checklist with two non-negotiables—provider-level customization and two-way EMR integration.

Why those two? Outpatient neurology spans dozens of subspecialties, and every provider has their own set of requirements based on the medical scenarios, their sub-specialties and their own clinical voice. So an AI that flattened that into a singlular fixed template wasn’t the right fit.

Two-way EMR integration mattered because most neurology visits are follow-ups, and without the ability to pull previous visit summaries into the current note, the AI scribe is doing only half the job leading to capped ROI.

DENT piloted other solutions before landing on Marvix. Amanda is candid about why they failed.

Cheat Sheet: Building Your Evaluation Criteria

• Define your must-haves before any vendor demo. Walk in knowing your non-negotiables so the demo works for you, not the other way around.
• Require two-way EMR integration. Pulling prior visit context isn't a bonus; for follow-up-heavy specialties, it's more than half the value.
• Test competing pilots with the same provider group. Comparing like for like is the only way to know which software actually performs better.

How DENT Ensured Adoption Across 80-Plus Providers

Signing with a vendor isn't the finish line, it's where the real work starts. DENT began the trial with a few super users, chosen deliberately across provider types and subspecialties to stress-test whether the AI could handle the full range of complex scenarios they encounter.

Once that group started seeing results, adoption grew organically. Watching colleagues take a real lunch break and stop charting at night did more than any formal campaign could.

But the super users were only part of the story. Amanda has a specific playbook for the providers who pushed back and the operational discipline behind it is what separates a software that gets adopted from the one that quietly stalls.

Cheat Sheet: Rolling Out AI in Your Practice

• Build a representative pilot group. Span provider types and subspecialties, not just one setting or specialty, so results reflect the whole organization.
• Schedule demos around clinic hours. Before or after, never during, respect providers' time and attendance improves.
• Bring operational discipline from your side. Track training completion individually and follow up. Adoption is managed, not assumed.

The Playbook DENT Built for AI Adoption

Amanda is direct about a plain truth: the cost of losing an experienced provider to burnout dwarfs the cost of a software that could have prevented it.

Her playbook didn't start with the software- it was built on the answers of three questions the team asked at the very beginning: Why do we need an AI scribe? What is the status quo today? And what will we expect of our providers post the roll-out?

They also looked hard at cost, and not just the software’s price tag but also the integration costs. At the end numbers told the story behind real impact.

Cheat Sheet: Building Your AI Adoption Playbook

• Start with three questions. Why are we doing this, where are we starting, and what will we ask of providers? Answer these before anything else.
• Budget for total cost, not just the software. Factor in integration and the behavior changes needed to make the investment worthwhile.
• Measure the outcome you care about. Survey providers on fatigue and workload throughout the pilot, don't assume the AI is working, verify it

What Has and Hasn't Worked in AI for Healthcare

Neurology is one of the hardest environments for any ambient AI documentation tool. The clinical language is complex, subspecialty requirements are diverse, note templates vary, and providers differ enormously in the phrasing preferences for their documents. Add to it the complexity involved in making the documentation insurance compliant. The softwares that failed DENT's evaluation simply couldn't handle that variation.

So what made Marvix AI different? Out-of-the-box clinical quality in documentation was part of it but Amanda points to something less obvious as the real differentiator between an AI company worth working with and one that isn't.

Cheat Sheet: How to Evaluate an AI Vendor

• Test against your hardest cases. Complex terminology and varied provider styles expose weak softwares fast. Evaluate on your real world cases, not mock calls in demos
• Demand provider-level customization. Organization-wide settings aren't enough; each provider needs to shape notes to their own style.
• Check for a real feedback loop. The vendor should have a clear process for turning client feedback into product updates, not just promises to "look into it."

Where Does Integration Fit in All of This

Turns out integration sits right at the center. Oftentimes, integration gets treated as a technical checkbox, but it's actually what determines how much value an ambient AI software can deliver in practical terms.

Uprooting the providers from their existing workflows to try a new shiny object has never worked. The software should fully integrate with their existing workflows not the other way round.  

Amanda talks about DENT’s experience with rolling out Marvix AI integration with eClinicalWorks (ECW) and how the pre- and post-charting dynamic changed for follow-up visits after integration.

There was an unexpected win, too - one that turned out to matter most for newer providers. And Amanda calls the integration itself one of the smoothest DENT has done over its EMR.

Cheat Sheet: Integrating AI with Your EHR

• Treat integration as value, not plumbing. How deeply the software connects to your EMR determines how much it can actually do. Don't relegate it to a checkbox.
• Prioritize two-way data flow. Pulling prior visit summaries into the current note is what transforms follow-up charting. Pushing notes within templates matters
• Prior vendor experience with the EHR matters: Integrations are tricky and EHR-specific. Vendor’s prior success with such roll-outs matters more than you think.

Advice for Administrators and Clinical Operators Evaluating AI

Amanda's core point cuts against how most organizations approach this: AI adoption in a large practice is an operational challenge, not a technical one. It takes sustained, hands-on involvement from administration and senior leadership - far more than most teams anticipate. Providers need relationships, not memos. They need someone to turn to when they can't quite articulate what they need from the software. They need training that bends around their schedule, not the other way around.

Cheat Sheet: Advice for Administrators and Clinical Operators

• Stay close to providers throughout. Not just at kickoff but sustained proximity through evaluation and rollout is what keeps adoption alive.
• Designate an administrative owner. Someone needs to own vendor communication and training coordination; it doesn't have to be a director, but it has to be someone.
• Set realistic internal expectations. Be honest about how long the first few months take. Underestimating the timeline is how momentum dies.

Parting Thoughts on AI in Healthcare

Amanda's view is that healthcare organizations in 2026 don't have the luxury of waiting. Payer reimbursements aren't getting easier, the staffing environment hasn't stabilized, and the administrative load on providers won't reduce on its own.

Her advice: Re-evaluate your processes and softwares at least once a year, and be honest about whether what you have is still the right fit.

The gap between a successful AI rollout and a failed one is never only the technology.

The best software rollouts are partnerships. The vendor brings the product and the responsiveness to improve it; the frontline administrators bring the relationships, the operational discipline, and the daily proximity to providers. Rollouts succeed when both sides share the same goals and maintain close coordination from evaluation through full adoption.

If you are interested in exploring Marvix AI for your practice, book your demo today.