How Health Systems Scale Virtual Care Using Clinical AI

Executive Summary

Four major health systems in Georgia—Piedmont Healthcare (17 facilities), Emory Healthcare (13 hospitals), Tanner Health (5 hospitals), and Grady Health System—have moved AI and virtual care from pilots to production scale. Their combined results reveal what’s actually working in healthcare’s technology transformation.

The proven results: 200,000+ automated clinical tasks completed in 10 months. A 20% reduction in uncontrolled hypertension through remote monitoring. Zero falls in AI-monitored behavioral health units. 130 minutes per 12-hour shift returned to bedside nurses. 3,500+ clinical interventions through proactive monitoring. 2,700 virtual care endpoints deployed in less than 1 year.

But success came with hard lessons. Infrastructure costs ran 30-40% above vendor estimates. Twenty-year-old cameras couldn’t support 2025 AI requirements. Workflow design mattered more than technology selection. Training had to preserve clinical judgment alongside AI adoption. And Georgia Medicaid still won’t reimburse remote monitoring for the sickest patients.

The competitive window is narrowing. Organizations that started 18 months ago now achieve scale competitors cannot quickly match. The technology works. The business case is proven. But execution requires sophistication beyond vendor selection.

This analysis presents operational realities—not aspirational roadmaps—from executives managing multi-facility deployments at scale.

When a hospitalist at Northeast Georgia Health System faced documenting a 207-day hospital stay on a patient’s discharge day, the traditional approach would have consumed three hours and produced what he described as “a very poor job of summarizing it.” Instead, artificial intelligence synthesized the entire encounter in minutes, producing “several paragraphs of all the salient points” and even using sophisticated language the physician had never employed. The patient “was awaiting repatriation to their home country.”

Within weeks, the technology that seemed novel had colleagues “lining up outside doors” demanding access.

This snapshot from a recent Georgia healthcare conference reveals how quickly AI and virtual care technologies are moving from pilot programs to operational imperatives. At four panels featuring executives from health systems ranging from five to 17 facilities, leaders disclosed results that should capture any healthcare C-suite’s attention. They have completed 200,000 automated clinical tasks in ten months, achieved 20% reductions in uncontrolled hypertension, recorded zero falls in monitored units, and returned 130 minutes per shift to bedside nurses.

Yet the same discussions exposed implementation realities that distinguish winners from those who will struggle. Twenty-year-old cameras cannot support 2025 AI requirements. Georgia Medicaid will not reimburse remote monitoring for the sickest patients. Young physicians at one organization could not discharge patients when computer systems failed because they had never learned to write prescriptions manually.

For healthcare executives navigating what one panelist called “a fast-moving cycle,” the message is clear. The technology works, the business case is proven, but execution requires sophistication that goes far beyond vendor selection.

The Business Case

Scale Achieved

Piedmont Healthcare moved from 130 monitoring endpoints in January 2024 to 2,700 by year-end. This represents a twentyfold increase across 17 facilities serving 2,719 beds. The system completed its buildout in just ten months using a rolling implementation strategy that installed roughly six rooms daily while keeping facilities operational. When the Masters golf tournament in Augusta blocked hotel availability, contractors simply shifted to other locations without breaking stride.

Emory Healthcare, the academic medical center with 13 hospitals, has 425 beds live on virtual nursing and sensor technology. The system is targeting 1,000 by year-end and 2,000 by fiscal 2027. Emory already operates 200 e-ICU beds and deploys 200 virtual nurses in ambulatory settings, giving it substantial experience to build upon.

Tanner Health, a nimble five-hospital system west of Atlanta, achieved zero falls during its med-surg pilot with a false alert rate below 1%. This created such staff confidence that “other units are requesting technology, asking ‘when are you installing this for us?’” Staff from units without the technology reported feeling “mistreated because they don’t have anything.”

These are not aspirational roadmaps. These are operational realities with measurable returns.

Clinical and Financial Outcomes

Piedmont’s remote patient monitoring and chronic care management programs enrolled 13,000 patients in care coordination and 3,000 in device-based monitoring within one year. The results: a 12% decrease in patients with A1C above 9 (uncontrolled diabetes), a 20% decrease in patients with uncontrolled blood pressure, nearly 3,000 appointments facilitated for patients struggling with access, and 3,500+ clinical interventions through proactive monitoring calls.

Patient stories illustrate impact beyond statistics. Clinicians helped patients obtain medications at lower cost, substituted medications when access failed, arranged denture access, secured specialist appointments, and stayed on the phone with patients experiencing altered mental status until help arrived. “These clinicians are really doing wonderful things and helping improve patient lives,” explained Katie Nolan, Piedmont’s director of virtual health.

The virtual nursing implementation produced measurable operational improvements. Excess days declined over a two-month consistent trend. Discharge order-to-door times shrank. ED hold times decreased. Medication reconciliation timeliness improved significantly. And 18% of nursing shift time—130 minutes per 12-hour shift—returned to direct patient care.

Emory reported dramatic reductions in falls with injury in units equipped with LIDAR-based sensor technology. The system achieved improved patient satisfaction scores (“likelihood to recommend”), declining length of stay, and reduced readmissions through predictive model integration with virtual nursing warm handoffs.

At Tanner’s behavioral health facility, the impact proved transformative. The AI system monitors patients throughout the facility using facial recognition, not just in beds. It detects early signs of agitation through micro-facial cues like jaw clenching and facial tics, enabling staff to intervene before full behavioral crises. The system alerts staff when patients fashion anything resembling a ligature for suicide attempts, when gait becomes unsteady from medication side effects, and when patients enter wrong rooms (preventing both violence and inappropriate sexual contact).

The technology also automated 15-minute safety checks required by policy. Previously, mental health techs opened doors every quarter-hour, often waking patients and creating agitation. Staff reported, “If you want me to be grumpy, come wake me up every 15 minutes.” One tech admitted, “I couldn’t tell if the patient was really breathing, so I go over to them and use the flashlight on my phone and shine it in their face.” The AI enables unobtrusive monitoring while automatically logging checks, freeing staff from standing at seclusion room doors maintaining “line of sight” for hours.

The ROI

While panelists noted that “billing is challenging right now” for some services, the value proposition extends beyond fee-for-service reimbursement. Even without reimbursement mandates, “in the long run, it saves money because they can help prevent unnecessary admissions and improve care in regular primary care appointments,” Nolan explained.

Organizations are seeing natural productivity gains without imposing volume mandates. “We’ve decreased some of that cognitive burden and that burnout,” noted one executive. “They’re more likely to see that next patient, or just have more time to spend with patients.” This happens naturally “without having to be very explicit” about productivity requirements, which is important “because a lot of these technologies are not inexpensive.”

The cost avoidance is measurable. Each prevented readmission saves $15,000 to $30,000 depending on condition. Each prevented fall with injury avoids $35,000 in additional costs. Reducing ED hold times improves throughput and revenue capture. Decreasing length of stay by even half a day multiplied across thousands of patients creates substantial margin improvement.

Why Act Now

The Competitive Window

The pace of deployment suggests a competitive inflection point. Piedmont moved from concept to system-wide implementation in under two years. Emory’s goal of 2,000 beds by 2027 represents a strategic commitment to technology-enabled care models that will be difficult for competitors to match without similar lead time.

More telling, the technology is creating its own demand. After Piedmont’s pilots, “word of mouth” between facilities drove expansion requests. Organizations that demonstrated value found nurses asking “when are we getting this?” rather than resisting change. Virtual nurses have become recruiting tools. Facilities advertising technology-enabled practice environments gain advantage in talent wars.

Nancy Haile, Vice President for Care Delivery and Innovation at Emory, identified the “breakthrough moment” early in their journey. “We were able to see that it worked, and that it clicked those buttons of patient experience, clinician experience, and reduction in [negative outcomes].” Organizations reaching this tipping point are accelerating while laggards remain in analysis paralysis.

The question is not whether your health system will deploy these technologies. The question is whether you will be an early mover capturing competitive advantage or a fast follower spending premium prices to catch up after your clinicians have experienced these tools at other organizations and now consider them minimum requirements.

The Talent Crisis

Healthcare’s workforce crisis makes these technologies strategic, not optional. Virtual nursing directly addresses nursing shortages, burnout, and retention. Piedmont reported decreases in “after-hours pajama time” for clinicians and natural increases in patient volumes without mandates as cognitive burden decreased.

“Those of us who got into caring for patients naturally want some more time, feel a little bit less stressed, can spend more time with patients,” one panelist explained. The technology is “freeing them up to do what they got into medicine to do.”

For organizations competing for scarce clinical talent, the value proposition shifts from cost reduction to talent attraction and retention. Younger clinicians “grew up with these tools so they are more comfortable,” creating expectation that sophisticated technology support is standard, not luxury.

One telling anecdote. After Piedmont installed virtual nursing, facilities without it found clinicians considering departures or transfers to equipped facilities. Technology became part of the employment brand. In tight labor markets where sign-on bonuses and premium pay have limited effectiveness, work environment differentiation matters.

Value-Based Care Enabler

For executives navigating value-based care contracts, these technologies provide the population health infrastructure previously impossible at scale. Piedmont’s goal of touching 50,000 lives through chronic care management and remote monitoring. This was achieved by partnering with vendors providing both technology and clinical support. It demonstrates scalability that manual approaches cannot match.

Grady Health System serves a population that is 67% insured (heavily Medicaid, Medicare, and ACA marketplace) and 33% uninsured. The system uses technology to address high-risk readmissions and maternal health disparities. Their focus on heart failure, COPD, and postpartum hypertension illustrates how safety-net systems leverage technology to manage complex populations cost-effectively.

“We foresee more value-based care continuing to grow,” predicted Nolan. “That’s a revenue stream for health systems so that we can pay for this monitoring and kind of spread out in more areas.”

The math is straightforward. A health system with 10,000 diabetic patients in value-based contracts achieving 12% improvement in A1C control avoids approximately 1,200 patients progressing to complications. At average complication costs of $8,000 to $15,000 annually, this represents $10 million to $18 million in avoided costs. Split that with payers through shared savings arrangements, and the technology investment pays for itself while improving outcomes.

What Really Works (and What Doesn’t)

Infrastructure Costs

Perhaps the conference’s most sobering lesson comes from infrastructure requirements. Do not assume existing equipment will work.

Tanner Health discovered that cameras installed in 2005, still functioning for basic security, lacked the resolution for AI-powered patient monitoring. “You cannot deploy 2025 technology and expect that those cameras from 2005 are going to give you the resolution that we needed,” explained Tammy Sprayberry, Director of Clinical Informatics. The health system had to replace cameras across facilities at unbudgeted cost.

Emory found that “most of our TVs in our patient rooms were outdated,” requiring complete re-cabling before deploying virtual nursing. This “hardware installation” requirement significantly impacted timeline compared to software-only solutions. The team discovered that making MyChart bedside standard required infrastructure investments nobody had anticipated.

Piedmont’s device footprint explosion from 130 to 2,700 endpoints in one year quickly overwhelmed initial support plans. The organization initially budgeted two technical support staff for remote troubleshooting. That proved wildly inadequate. The team had to develop a “collaborative support model” between virtual health teams and facility IT, with protocols “regularly revised” (four to five major revisions) as they learned what worked.

One unexpected issue almost derailed patient confidence. TVs in patient rooms defaulted to cable programming, meaning virtual nurses might appear on screen adjacent to inappropriate content. The team had to develop specific TV programming protocols to ensure professional presentation during clinical interactions. “Truly is one of the things that we wouldn’t have thought about,” a panelist acknowledged, “but it’s important.”

Executive takeaway: Budget for infrastructure upgrades at 30-40% beyond vendor estimates. Assume cameras need replacement, TVs need upgrading, network capacity needs expansion, and technical support requirements will exceed initial plans. Organizations that under-resource infrastructure will watch clinician enthusiasm evaporate as reliability problems mount and workarounds proliferate.

The Ordering Problem

Remote patient monitoring requires physician orders for reimbursement. This created a workflow challenge that derailed Piedmont’s initial pilot.

“If you were waiting for a provider to enter one more order, it may or may not happen,” Nolan acknowledged. Providers are “thinking about everything else, in addition to care gaps and annual wellness visits and all the things.”

The initial approach asked physicians to identify qualifying patients during visits and enter orders individually. Adoption languished below 20%. Physicians acknowledged the value but faced competing demands on attention and documentation time.

The solution abandoned “onesie-twosies” individual ordering. Piedmont developed registry-generated quarterly lists presenting all qualifying patients for provider review. Instead of remembering to order for Mrs. Smith during her visit, physicians receive lists of all diabetic and hypertensive patients and approve or decline based on clinical judgment. “No one knows their patients better than their provider,” Nolan explained, “so they can go in and say Mrs. Smith is not compliant, she won’t use this.”

This single workflow innovation accelerated enrollment from roughly 20 patients monthly to several hundred. Combined with pre-call messaging from providers legitimizing vendor outreach (“you want to make sure that you’re legitimizing the service because a lot of times patients think someone is trying to sell me something”), the approach transformed adoption.

The principle applies beyond ordering. Any workflow requiring clinicians to remember additional steps during time-pressured encounters will fail. Successful implementations engineer proactive systems that present decision points when clinicians have time to evaluate, not in the middle of patient visits.

Executive takeaway: Workflow design matters more than technology selection. Identify friction points early and engineer solutions before scaling. “Fit the workflow” is not vendor marketing speak. It is the difference between 10% and 90% utilization. Invest in workflow analysis before deployment, not after utilization disappoints.

Governance That Works

Organizations described governance models evolving in real-time. Most integrated AI evaluation into existing technology, compliance, and security structures rather than standing up dedicated AI committees. This worked “because we’ve stuck with a lot of our vendor solutions” rather than building internally, one executive explained.

The evaluation framework most cited involves three to four stages, each requiring six to eight weeks. “Just add to the math, it will take time, just like many other good things,” noted one executive. The first stage alone determines if a solution is “not even worth considering because it’s just inadequate from a data standpoint.”

But speed pressure is intense. “We’re still in a fast-moving cycle,” another leader explained. “People want us to move fast.”

The balancing act requires risk stratification. Leaders asked, “How hard to reverse? What are the risk outcomes?” Low-risk, easily reversible decisions (like inbox response suggestions) can deploy faster. High-risk, difficult-to-reverse implementations (like diagnostic support) require extensive evaluation.

One framework gaining traction categorizes AI applications into three tiers. Tier 1 handles fast-track administrative automation with human review—easily reversible, low patient safety risk. Think inbox response drafts, discharge summary generation, appointment scheduling suggestions. Evaluation takes 4-6 weeks. Tier 2 covers standard clinical decision support with provider override required—moderate patient safety implications. Examples include medication recommendations, care gap alerts, remote monitoring protocols. Evaluation takes 8-12 weeks. Tier 3 demands rigorous review for direct patient impact, difficult-to-reverse decisions with high safety stakes. Diagnostic imaging interpretation, treatment selection algorithms, automated medication dispensing. Evaluation takes 12-16 weeks minimum with ongoing monitoring requirements.

The governance structure must also address data issues that caught several organizations off guard. When virtual nursing implementations use ambient documentation, audio files are stored by vendors, often with time-limited access (90 days at some vendors). Research teams that historically used clinical notes for studies suddenly found data inaccessible. “In the old days, it was just the note that they wanted,” one executive explained. Now researchers want audio files, which may not be available.

This raises questions about data ownership, retention policies, and storage costs. One executive asked pointedly, “Do you really want to spend storage and your cost and your infrastructure on those things?” with 10 petabytes of audio potentially required. But another countered that audio represents verified clinical encounters and may have legal or quality improvement value justifying costs.

Executive takeaway: Establish clear governance now, before you need it. The framework should enable speed where appropriate while ensuring safety where necessary. Address data ownership, retention policies, and research access in vendor contracts upfront. Organizations improvising governance during crises make preventable mistakes that create liability exposure or operational disruption.

Training the Next Generation

A sobering discussion emerged around critical thinking and clinical skills. One organization discovered residents “couldn’t discharge patients” when systems failed because “none knew how to write a prescription” manually. They had trained exclusively on computers.

Dr. Nina Princeton, Medical Director for Ambulatory and Informatics at Grady, identified the core tension. “This is the danger that I see. We really have to train our younger students, our medical students, our young doctors: use your brain. You can use this as an augmented tool, but you cannot replace it.”

Examples of concerning behavior included physicians ordering inappropriate meningitis vaccines rather than critically evaluating Best Practice Alert recommendations. Some clinicians ask “Why do I have to give somebody a meningitis vaccine?” instead of checking the problem list themselves. In the most troubling example, providers actually ordered meningitis vaccines inappropriately rather than dismissing alerts. “Which is more mind boggling,” Princeton noted.

The Epic Best Practice Alerts that should prompt clinical thinking instead sometimes replace it. Younger physicians trained exclusively with decision support tools may lack practice in unaided diagnostic reasoning. The concern extends beyond individual practice to systemic quality risk.

The proposed solution combines early AI introduction with verification requirements. “Preceptors ask them, ‘Why did you take that recommendation?’” forcing learners to defend clinical reasoning. At one teaching hospital, radiology residents must explain why they accepted or rejected AI findings about incidental discoveries. “You can still make this a learning environment,” Princeton explained. “You can still use this tool and just ask them, ‘Why are you doing it?’ If they cannot tell you why, they may have to go back and study.”

Some organizations recommend maintaining paper documentation practice alongside digital tools. One session presenter suggested having trainees “practice with paper and really think about what information is needed before reaching for technology.” The goal is not rejecting technology but ensuring fundamental skills remain sharp when technology fails or is unavailable.

Executive takeaway: Training programs must evolve with technology. Organizations that allow AI to erode fundamental clinical skills are creating long-term liability and quality risks. The next generation needs to use AI effectively while maintaining clinical judgment. Invest in curriculum development that integrates technology as an augmented intelligence tool while preserving core competencies. Consider this a patient safety issue, not just a training preference.

Navigating the Vendor Landscape

Market Consolidation

An informal but revealing post-panel discussion exposed market dynamics C-suites should understand. An entrepreneur in the ambient listening space acknowledged competing “against companies that have hundreds of millions of dollars” while facing Epic’s in-house development of native solutions.

The economic challenge is stark. “The unit of economics aren’t there” for venture-backed companies built to serve large hospitals when forced to compete with integrated EHR vendor solutions. “We know we’ve been selling our software for a million dollars, but now we’re going to sell down into ambulatory” does not work economically. As one attendee put it, the model is “trying to put a Rolls Royce engine inside a Toyota Camry.”

Epic’s dominance was repeatedly cited as “the biggest challenge” for third-party vendors. The company’s “Community Connect” and “Garden Plot” products extend enterprise capabilities to smaller practices, making Epic a competitor to vendors who historically partnered with Epic customers. Epic’s massive patient database containing “tens of millions, hundreds of millions of patients” enables AI recommendations impossible for standalone vendors to match.

Epic’s “Signal” tool, leveraging the entire Epic user database, can match a local patient with similar patients across the entire Epic ecosystem. For a 55-year-old white male on multiple medications for diabetes and hypertension, the system identifies thousands of similar patients and suggests treatments based on what worked for comparable cases. This “real-world evidence database larger than any clinical trial database” gives Epic structural advantage over point solutions.

One ambient documentation company representative acknowledged that major health system customers present at the conference were “my customers” historically. But now Epic and other large vendors are moving into ambient documentation themselves, creating existential pressure on standalone vendors.

The venture capital that fueled innovation is drying up as interest rates rise and investors demand profitability over growth. Companies that raised capital at high valuations to capture hospital market share now face squeezed margins and uncertain exit paths. Several vendors mentioned at the conference may not exist independently in 24 months.

Selection Criteria

Organizations that succeeded emphasized specific selection criteria beyond technology capabilities.

First, single-platform solutions. Piedmont selected a vendor providing both chronic care management and remote patient monitoring in one platform. This enabled 50,000-life scale with consistent clinical support and simplified contracting.

Second, clinical support included. “Not just technology” but comprehensive clinical staffing. Piedmont’s vendor handles device shipping, patient training, monitoring, and escalation. “We don’t want to be in the business of device management at this scale,” Nolan explained. This distinguishes true solution providers from technology-only vendors.

Third, EHR integration depth. “Through integration [with] the workflow, it’s sustainable,” noted one panelist. Solutions requiring parallel workflows fail. Data must flow to Epic “almost immediately” without manual intervention. Organizations should demand demonstration of bidirectional integration, not just one-way data feeds.

Fourth, proven scale. Site visits to hospitals nationally that had implemented successfully provided validation and learning. Organizations should insist on references at similar scale, not just pilot deployments. Ask about peak concurrent users, daily transaction volumes, and uptime statistics over trailing 12 months.

Fifth, flexible implementation. Vendors willing to customize protocols by specialty and practice type enabled broader adoption. Rigid one-size-fits-all solutions hit adoption ceilings. The winning vendors treated implementations as partnerships, not product deployments.

Sixth, financial stability. In a consolidating market, vendor financial health matters. Organizations should request financial statements, understand burn rate for venture-backed companies, and include contractual protections for wind-down scenarios. One health system discovered their virtual nursing vendor had been acquired mid-implementation, disrupting support and roadmap commitments.

Build vs. Buy

Notably, every presenting organization partnered with vendors rather than building AI internally. The pragmatic assessment suggests that developing AI in-house requires resources and timelines most organizations cannot sustain.

“Small system, large system, extra large system,” one executive noted. “Our sweet spot is that we’re not tiny, but we’re not huge and multi-state. So our layers are a little bit shorter” between idea and execution. But even nimble organizations chose vendors for core technology.

The exceptions are academic medical centers with substantial research infrastructure and AI expertise. Emory, for example, conducts internal development for some applications while partnering for others. But even Emory relies primarily on vendors for production systems supporting patient care.

The competitive advantage comes from implementation excellence, not technology development. Organizations that obsess over clinical workflow integration, change management, outcome measurement, and continuous improvement will outperform those with superior technology but poor execution.

Executive takeaway: Unless you are an academic medical center with substantial research infrastructure and proven AI development capability, partner with established vendors. Choose vendors based on comprehensive support, financial stability, and commitment to your EHR platform, not just technology features. Negotiate contracts that include customization flexibility, performance guarantees, and exit rights. Build internal capabilities in implementation excellence, not technology development.

The Reimbursement Problem

Coverage Gaps

The most significant implementation barrier is not technical but political. Georgia Medicaid does not cover remote patient monitoring. This creates what panelists called “a critical gap for highest-need population.”

“This is a policy change that we would love everyone to support and get on board,” urged Angie Ross, Population Health Director at Grady. “Because it’s just what’s needed to improve outcomes.”

The cost-effectiveness argument is straightforward. Remote monitoring “can help prevent unnecessary admissions” and “improve care in regular primary care appointments,” resulting in net savings. Yet reimbursement does not reflect this value.

The policy failure is particularly concerning for maternal health. Grady’s postpartum hypertension monitoring program targets women at high risk for mortality and readmission. These are disproportionately Medicaid patients. The program uses remote blood pressure monitoring to detect dangerous hypertension before stroke or seizure. Clinical outcomes are measurably better. But Medicaid reimbursement limitations create sustainability challenges, requiring Grady to seek grant funding.

Coverage varies not just by payer but by specific plans within health plans. “It depends on payer, it depends on what plan that the patient takes with the health plan,” creating confusion and administrative burden. Organizations must verify coverage patient by patient, adding friction to enrollment.

Grady relies on grant funding for uninsured populations, which is not a sustainable model. Organizations serving vulnerable populations face stark choices: absorb costs, limit services, or decline to serve patients who need monitoring most. Each option creates strategic risk.

Telehealth Uncertainty

COVID-era telehealth expansions face uncertain futures. “Telehealth funding is not going to be viable,” one panelist warned. “Hospital-at-home programs are losing reimbursement support.”

The philosophical resignation was palpable. “Everybody will be screaming and talking at the same time. They’re going to push what they want to push anyway. So we just sit down and watch.”

If telehealth becomes unviable financially, alternatives are costly. Organizations must “send buses to go pick up patients” or “provide bus passes, the old way.” For rural populations, reduced virtual care access means reduced care, period. “Georgia is one of those metro areas where we’re just so spread out, and people just don’t have a transportation system,” one executive explained. Virtual care should solve this, but without reimbursement, economic reality constrains access.

The policy uncertainty creates strategic ambiguity. Organizations making multi-million dollar investments in virtual care infrastructure face risk that reimbursement will evaporate before ROI is achieved. Executives must make capital allocation decisions without clarity on revenue models.

Value-Based Solutions

Organizations navigating this landscape are pivoting toward value-based care contracts where outcomes matter more than fee-for-service billing. Remote monitoring that prevents expensive readmissions generates shared savings regardless of whether the monitoring itself is reimbursed line-item.

“Value-based care is a revenue stream for health systems so that we can pay for this monitoring,” explained Nolan. Organizations with substantial value-based contracts can justify technology investments that fee-for-service economics will not support.

The strategy requires scale. A health system with 5,000 patients in value-based contracts lacks statistical power to demonstrate ROI from technology preventing small numbers of adverse events. But a system with 50,000 lives can show meaningful savings from even modest outcome improvements.

This creates strategic bifurcation. Large integrated delivery systems with substantial value-based contracts have economic foundation for virtual care investments. Smaller organizations dependent on fee-for-service revenue face much more challenging business cases, potentially accelerating market consolidation.

Executive takeaway: Reimbursement uncertainty is a strategic risk requiring board-level attention. Organizations should pursue a four-part strategy: (1) Advocate actively for policy changes through industry associations and direct engagement with regulators. (2) Accelerate value-based contracting to create alternative revenue models supporting technology investment. (3) Develop grant funding capabilities for coverage gaps, particularly for vulnerable populations. (4) Build business cases that do not depend on optimistic reimbursement assumptions. Model scenarios where fee-for-service revenue remains constrained and ensure ROI through value-based contracts and operational savings.

Equity and Access

The Digital Divide

A pointed exchange during Q&A illustrated patient engagement challenges. When asked about 80-year-old patients accessing digital health tools, the response was blunt. Most do not.

The problem cascades. “Do I go home first to get those instructions before I go to the pharmacy to get my life-saving medication? Once I’m in the portal, do I have to print it out? But how do I print it out if I don’t have a printer? Am I supposed to go to the portal every time now to figure out what I need to do for medication?”

One audience member counted “eight different ways to manage data once it’s time for the patient to take care of their own care” following discharge. This includes patient portals, text messages, phone calls, paper printouts, pharmacy direct communication, email notifications, MyChart bedside education during stay, and care manager follow-up calls. The complexity itself becomes a barrier.

The assessment was stark. “If you’re 70 and up and you have two chronic health conditions,” digital-first approaches create structural barriers. Without intensive case management (which not all organizations provide consistently), technology widens rather than narrows disparities. The panel acknowledged this reality. “It’s still the wild wild west,” one executive admitted.

Grady’s postpartum hypertension monitoring program illustrated population-specific challenges. Despite being younger than typical remote monitoring patients, new mothers had significantly lower device transmission rates than heart failure or COPD patients. “Patients think if you’re having abnormal symptoms, it’s normal. You just had a baby,” explained Ross. Between newborn care demands and dismissing dangerous symptoms as normal postpartum recovery, technology alone proved insufficient.

The response required “community health workers” and enhanced support “meeting patients where they’re at.” This is resource-intensive but necessary. The question for health system executives is whether equity is a constraint to be managed or a market opportunity to be captured.

Rural Connectivity

“Georgia is one of those metro areas where we’re just so spread out, and people just don’t have a transportation system,” one executive explained. Virtual care should solve this, but “connectivity in rural areas” creates exclusion criteria.

The emerging solution involves universal SIM cards in monitoring devices, eliminating dependence on patient home internet or cellular plans. Organizations planning hospital-at-home programs are prioritizing this technology. Devices with built-in cellular connectivity cost roughly $50 more per unit but eliminate the single biggest barrier to rural deployment.

But even with connectivity solved, reach remains limited to existing patient populations. “Huge gaps in the middle of the state” persist where organizations have no facilities or provider relationships. Virtual care can improve access for existing patients but does not automatically extend access to unserved populations.

Some organizations are exploring telemedicine partnerships with rural hospitals and clinics to extend virtual specialty access. Others are developing mobile health units with connectivity solutions. But these require sustained investment and operational commitment beyond technology deployment.

The Business Case

For safety-net systems like Grady, which serves 33% uninsured patients, equity is mission, not option. But market-driven organizations face business imperatives too.

Value-based contracts increasingly include quality metrics around health equity and disparity reduction. Organizations with widening gaps face financial penalties in these contracts. Technology that improves outcomes for commercially insured patients while failing Medicaid populations creates measurable financial risk.

Moreover, markets are changing demographically. Atlanta’s population is increasingly diverse. Organizations that cannot serve diverse populations effectively will lose market position. Technology that works only for affluent, tech-savvy patients will not scale to market needs.

Patient experience increasingly drives market share. Organizations that deliver seamless digital experiences for some patients while creating barriers for others generate negative word-of-mouth and reputation risk. In markets with choice, patients select providers who meet them where they are.

The talent imperative also drives equity focus. Clinical staff want to practice in environments that serve all patients effectively. Organizations with digital-divide problems that leave some populations underserved face staff dissatisfaction and recruitment challenges.

Executive takeaway: Build equity considerations into technology selection from day one. Test solutions with diverse populations before scaling. Budget for the enhanced support (case management, community health workers, alternative communication channels, language services) that makes technology accessible to all populations. Measure disparities in utilization and outcomes explicitly, and address gaps systematically. The organizations that crack inclusive deployment will have competitive advantage in increasingly diverse markets and value-based contracts that penalize inequity.

The Data Access Crisis

A critical issue emerged from the discussion that deserves board-level attention. As organizations deploy ambient documentation at scale, they are creating a data access crisis for clinical research and quality improvement.

Traditionally, research teams accessed clinical notes in EHR databases for retrospective studies. Notes contained complete documentation of clinical encounters, enabling everything from outcomes research to quality improvement to population health analytics.

Ambient documentation changes this fundamentally. AI listens to clinical encounters and generates notes, but the audio files are stored by vendors, often with time-limited access. One vendor provides 90-day access to encounter audio for the provider who conducted the visit, after which audio is purged or archived inaccessibly.

“In the old days, it was just the note that they wanted,” one executive explained. But researchers increasingly want audio for studies examining communication patterns, diagnostic reasoning, or patient-provider interaction. The audio may contain nuances lost in transcription. Yet the audio is not available, or available only at additional cost with complex contracting.

This creates distributed data problems. Research teams need audio files, DICOM images from radiology, AI-generated insights, alert data, structured EHR data, and unstructured clinical notes. “Most of us haven’t figured out how to put all these into one database” for unified analysis, one panelist admitted.

The storage question adds complexity. Audio files consume vastly more storage than text. One executive estimated “10 petabytes of storage” could be required for comprehensive audio retention at a large health system. The question becomes: “Do you really want to spend storage and your cost and your infrastructure on those things?”

The counterargument focuses on verification and legal defensibility. Audio represents verified clinical encounters, while AI-generated notes are derivative. In legal proceedings, having audio documentation might prove valuable. But nobody knows for certain because case law has not developed.

Some organizations are establishing retention policies that archive audio for high-risk encounters (complex surgeries, difficult diagnoses, contentious patient interactions) while purging routine audio after verification. Others are negotiating with vendors for extended retention or direct research access. Still others are building data lakes that consolidate disparate sources, audio included.

Executive takeaway: Address data ownership, retention, and research access in vendor contracts before deployment. Convene research, compliance, legal, and IT stakeholders to define requirements. Consider the long-term value of audio data for quality improvement, risk management, and research. Budget for storage infrastructure if retention is required. This is not a minor technical issue. It affects institutional research capabilities and potentially legal defensibility of clinical documentation.

Your Action Plan

First 90 Days

Start with an honest infrastructure assessment. Don’t rely on vendor promises that existing equipment will suffice. Inventory existing cameras with specifications (resolution, connectivity, age). Assess patient room TVs for compatibility with virtual nursing platforms. Evaluate network capacity for simultaneous video streams and remote monitoring data transmission. Identify gaps before vendors promise easy implementation. Budget realistically for upgrades—assume 30-40% beyond vendor estimates. If vendors quote $2 million, budget $2.8 million and establish contingency for unexpected issues. Organizations that under-budget infrastructure face mid-implementation funding crises that delay deployments and damage credibility.

Establish or refine your governance framework. Don’t wait for crisis to define AI evaluation processes. Convene stakeholders from clinical leadership, IT, compliance, legal, risk management, and quality to establish or refine governance structure. Define risk stratification criteria determining evaluation rigor required for different AI applications. Establish clear accountability for decisions. Create processes that enable speed where appropriate (low-risk applications) while ensuring safety where necessary (high-risk applications). Address data issues proactively: ownership, retention, research access, storage costs. Define policies before vendors control the conversation.

Assess the vendor landscape strategically. Look beyond technology features to comprehensive support models. Evaluate vendor financial stability (request financial statements for private companies, analyze public company health). Prioritize EHR integration depth, not just claimed compatibility. Conduct site visits to similar organizations that have implemented successfully. Insist on references at your scale. Ask about peak concurrent users, daily transaction volumes, and uptime statistics over trailing 12 months. Interview frontline staff at reference sites, not just executives. Understand what worked and what didn’t. Negotiate contracts that include customization flexibility, performance guarantees, and exit rights. Address what happens if vendor is acquired or fails. Don’t accept standard contracts without modification.

Review workforce strategy with a new lens. Calculate clinician time currently spent on tasks AI could handle. Use time-motion studies or self-reported data to quantify documentation burden, administrative tasks, and non-clinical work consuming clinical staff time. Model talent retention impact of reducing administrative burden. Survey staff about burnout sources and technology wish-lists. Consider competitive positioning for clinical recruitment in markets where competitors deploy advanced technology.

Analyze payer mix and reimbursement exposure. Quantify revenue at risk from reimbursement uncertainty. Model scenarios where telehealth funding ends, where Medicaid continues non-coverage of remote monitoring, where fee-for-service rates decline. Identify value-based contracts that could support technology investment through shared savings. Prioritize expansion of value-based contracting to create sustainable economics for virtual care. Develop grant strategy for uninsured population coverage. Identify foundations, government programs, and corporate sponsors that fund health equity initiatives.

Next 12 Months

Pilot with intention to scale. Select initial use cases with clear ROI and high clinical need. Avoid “innovation theater” pilots that cannot scale. Define success metrics before launch (clinical outcomes, operational efficiency, financial impact, satisfaction). Establish baseline measurements. Build internal champions through early success. Identify respected clinicians who will test solutions and provide feedback. Invest in their success. Use their testimonials to accelerate adoption. Plan for rapid scale once validated. Competitors are moving fast. Organizations that succeed with pilots but delay scaling lose competitive advantage. Establish governance approval for accelerated rollout contingent on meeting pilot success criteria.

Redesign workflows proactively. Don’t assume technology will fit existing processes. Map current workflows in detail before technology selection. Identify friction points where technology could reduce burden. Engineer out bottlenecks like ordering requirements. Piedmont’s bulk ordering solution provides a model. For any workflow requiring clinicians to remember additional steps during time-pressured encounters, redesign to present decisions when time permits. Involve frontline clinicians in design from start. Co-design sessions where end users shape implementation produce better adoption than top-down deployments. Clinicians who help design solutions become advocates.

Build training programs that preserve judgment. Update curricula to include AI tool use while emphasizing critical thinking. Implement verification requirements for AI-generated recommendations. Make “Why did you accept this recommendation?” standard practice. Maintain fundamental skill practice alongside technology. Consider periodic “unplugged” days or shifts where clinicians practice without AI support, maintaining manual skills. Develop simulation-based training that presents technology failures requiring manual workarounds. Ensure staff can function when systems are unavailable.

Develop comprehensive change management. Plan for at-the-elbow support, not just online training. Budget for super-users at each facility who can provide real-time assistance during go-live and early adoption. Identify and develop champions at each facility. Invest in their success through dedicated support, early access to features, and recognition. Communicate transparently about challenges and learnings. Staff respect honesty. Organizations that pretend implementations are seamless lose credibility when problems emerge. Create feedback loops for continuous improvement. Establish mechanisms for frontline staff to report issues and suggest improvements. Act on feedback visibly.

Establish outcome measurement from day one. Define metrics before implementation. Clinical outcomes (falls, length of stay, chronic disease control, readmissions). Financial outcomes (cost per case, revenue capture, margin impact). Operational outcomes (throughput, wait times, capacity utilization). Satisfaction outcomes (patient, clinician, staff). Track metrics rigorously. Establish dashboards updated at least monthly. Share results transparently with stakeholders. Use data to refine approach. When outcomes disappoint, investigate root causes and adjust. When outcomes exceed expectations, understand why and replicate success factors. Measure and address disparities explicitly. Track utilization and outcomes by patient demographic groups. Identify and close gaps systematically.

Three-Year Horizon

Build toward value-based care infrastructure. Accelerate contract negotiations incorporating virtual care infrastructure. Position technology as population health enabler, not cost center. Develop sophisticated analytics linking technology deployment to outcomes improvement. Design shared savings models with payers explicitly incorporating technology-enabled care management. Negotiate upside opportunities tied to outcomes improvement enabled by virtual care and remote monitoring. Demonstrate value through data. Develop business intelligence capabilities that show clear causality between technology deployment and improved outcomes. Use this data in negotiations.

Create strategic technology partnerships. Move beyond transactional vendor relationships to strategic partnerships. Co-develop solutions for your specific populations. Some leading organizations establish joint development agreements with vendors, sharing development costs in exchange for customization and favorable economics. Secure favorable economics through volume commitments. Multi-year commitments with expansion provisions create basis for better pricing and dedicated support. Consider equity investments in strategic vendors. Some health systems take minority stakes in technology companies critical to strategy, aligning interests and securing long-term partnership.

Develop competitive differentiation. Market virtual care access as patient experience advantage. Develop consumer-facing communications highlighting technology-enabled convenience, accessibility, and quality. Position technology-enabled practice as clinician recruitment tool. Feature virtual nursing, AI documentation support, and reduced administrative burden in recruitment campaigns. Build reputation as innovation leader. Publish outcomes. Present at conferences. Attract media coverage. Reputation as technology leader attracts patients, clinicians, and partners.

Address equity systematically, not symbolically. Partner with community organizations for digital literacy and access. Collaborate with libraries, churches, community centers to provide technology training and access points. Develop alternative communication channels for diverse populations. Ensure voice calls remain option for patients who do not use patient portals. Provide language services for non-English speakers. Measure and close disparity gaps as quality imperative. Establish executive accountability for equity metrics. Link leadership compensation to progress on disparity reduction. Deploy community health workers, case managers, and navigators to support populations needing enhanced assistance with technology adoption.

Engage in policy advocacy. Join industry coalitions pushing for reimbursement reform. Organizations like American Hospital Association, American Medical Association, and state hospital associations coordinate advocacy. Document outcomes compelling to policymakers. Quantify lives saved, complications prevented, costs avoided. Translate clinical outcomes into policy arguments. Advocate at state and federal levels for coverage expansion. Meet with legislators and regulators. Provide testimony. Submit public comments on proposed rules. Build relationships with payer leaders to advocate for rational reimbursement policies that recognize value of virtual care and remote monitoring.

The Bottom Line

The healthcare technology transformation is not coming. It is here. Organizations deploying AI and virtual care at scale report double-digit improvements in chronic disease control, meaningful reductions in clinician burnout, and operational efficiencies that materially impact margins.

The evidence from Georgia health systems is unambiguous. Piedmont’s 200,000 completed virtual nursing tasks and 20% reduction in uncontrolled blood pressure are not projections. They are achieved results. Emory’s dramatic fall reductions and improved patient satisfaction scores are measurable. Tanner’s zero falls in behavioral health units with below 1% false alert rates demonstrate technology maturity. Grady’s success managing complex, vulnerable populations through remote monitoring proves applicability beyond affluent populations.

But the window for competitive advantage is narrowing. As one panelist observed, there has been “a natural gravitation” toward technologies that genuinely help clinicians and patients. Once physicians experience 130 minutes returned to their shift, or patients achieve blood pressure control through remote monitoring, expectations reset permanently. Facilities without these capabilities will struggle to compete for talent and patients.

Organizations succeeding share common characteristics. They invest realistically in infrastructure (not accepting vendor promises that existing equipment will suffice). They engineer workflows proactively (not forcing technology into existing processes). They govern rigorously but not bureaucratically (distinguishing high-risk from low-risk applications). They partner strategically with vendors (seeking comprehensive support, not just technology). They commit to equitable implementation (budgeting for enhanced support enabling access for all populations).

Importantly, these organizations are not waiting for perfect solutions or complete reimbursement certainty. They recognize that reimbursement policy lags clinical capability by years, and that waiting for policy clarity means falling behind competitors. They are building business cases on value-based care economics and operational savings, not optimistic fee-for-service assumptions.

They are learning by doing, iterating rapidly, and building organizational capabilities that compound over time. The organizations that started 18 months ago are now achieving scale competitors cannot match quickly. The learning curves, workflow refinements, technical infrastructure, and cultural adaptation required for successful deployment take time. Organizations starting today face 18-month runway to reach maturity leaders have already achieved.

The question for healthcare C-suites is not whether to invest in AI and virtual care. The data from organizations already deployed proves the business case conclusively. The question is whether your organization will be among the leaders capturing competitive advantage, or among the laggards explaining to your board why patient satisfaction is declining, clinician turnover is accelerating, and operating margins are compressing while competitors with better technology pull ahead.

The technology is ready. The business case is proven. The vendors exist. The clinical evidence is compelling. What is required now is leadership commitment to execute with the sophistication these implementations demand.

The sophistication includes honest infrastructure assessment and adequate budgeting. It includes workflow engineering that eliminates friction rather than creating it. It includes governance that enables speed while ensuring safety. It includes training that preserves clinical judgment while leveraging AI capabilities. It includes vendor partnerships that provide comprehensive support. It includes reimbursement strategies that do not depend on policy changes you cannot control. It includes equity commitments backed by resources, not just rhetoric.

The next two years will separate healthcare’s technology leaders from those struggling to catch up. Leaders will be defined not by who has the most advanced technology but by who implements most effectively, achieves measurable outcomes, and creates sustainable competitive advantage.

Which will your organization be? The executives presenting at this conference have provided a roadmap. They have shared their successes and, importantly, their failures and lessons learned. They have demonstrated that scale is achievable with proper planning and execution. They have shown that the business case is real, not aspirational.

The only question remaining is whether healthcare leadership has the conviction to act decisively while the competitive window remains open, or whether another round of analysis and deliberation will leave your organization watching competitors pull ahead with technology, talent, and market position advantages that prove difficult to overcome.

The organizations that move now with sophistication will define the next decade of healthcare delivery. Those that hesitate will spend that decade trying to catch up.

Ready to Transform Your Healthcare Organization?

OrbDoc’s AI-powered clinical documentation platform helps health systems achieve the outcomes described in this analysis. Our technology returns time to clinicians, improves documentation quality, and enables the virtual care transformation reshaping hospital economics.

Request a demo to see how leading health systems are deploying AI at scale.

Explore our solutions for hospitals and health systems.

Learn more about how AI medical scribes work in practice.

This analysis is based on presentations and discussions at the 2025 Georgia Healthcare Conference featuring executives from Piedmont Healthcare, Emory Healthcare, Tanner Health, Grady Health System, and technology vendors.