TLRT: The Last Mile of AI in Healthcare: Actionable Insights at the Point of Care

Summary

Artificial intelligence (AI) is rapidly moving through healthcare, shifting from pilot projects and conceptual promises to a critical phase of execution. Healthcare systems today are under intense pressure, facing accelerating costs, thin operating margins, and a generational shift in the workforce combined with a looming shortage of human labor. Amid this environment, AI is no longer a futuristic vision; it is becoming a practical necessity to deliver care efficiently and sustainably.

The challenge for healthcare leaders is executing on this potential: ensuring that predictive intelligence turns into real-time interventions and that insights are embedded directly into daily workflows. This is often called the “last mile of AI,” where unified data is translated into proactive, intelligent workflows without adding undue burden on clinicians and staff.

To explore how leaders are defining and measuring success as they operationalize AI at scale, CHIME convened a Thought Leadership Roundtable (TLRT) in Atlanta, Georgia, sponsored by Innovaccer. The session brought together CIOs, CMIOs, and other digital health leaders to share candid insights on the pace of innovation, the imperative for automation, and the necessary cultural and governance shifts.

The ensuing discussion detailed four major strategic imperatives for leaders navigating this seismic shift: defining success for the board, facing the automation imperative, and building governance to manage unprecedented speed.

‍
The session was moderated by CHIME President & CEO Russ Branzell and sponsored by Innovaccer. CHIME member participants included:

• Bonnie Boles, MD, SVP & Chief Medical Information Officer, Tanner Health
• Geoffrey Brown, VP & CIO, Piedmont Healthcare
• Sepi Browning, Director of Technology, Piedmont Healthcare
• Karl Kiessling, Senior Director & Client Site Leader, Medical Center Health System
• Chris Paravate, EVP & CDIO, Northeast Georgia Health System
• Aalpen Patel, Chief Clinical and Innovation Officer, MedQuest Imaging
• Todd Rowland, CMO, Motiv Health
• Anthony Villanueva, CIO, Neighborhood Health
• A Director of Epic Applications at a large non-profit health system

‍

The Near-Term Mandate: AI Outcomes

The last mile of AI is about action, not algorithms, noted Ashish Singh, President of Innovaccer. “Real success would be when a clinician doesn’t even realize they’re ‘using AI,’ but they spend less time on administrative work, patients get care faster, and the health system sees measurable value.”

The definition of AI and its role in healthcare may change from day to day and week to week, as Branzell noted, but healthcare organizations can’t afford to take the slow road to action. AI is not new, even in healthcare, but the rate of progress and widespread demand around this technology is compressing timeframes for execution. As systems confront the “last mile” challenge — turning prediction into action — leaders are setting aggressive targets for what AI must accomplish in the next 12 to 24 months.

‍

Augmenting Scarce Resources In a Time of Challenging Changes Continued

Anthony Villanueva, CIO at Neighborhood Health, captured this shift by framing the technology not as artificial, but as an amplifier of human capability. “I prefer to consider it as ‘augmented’ intelligence,” he said. “It’s going to serve as an enabler to drive gains within our industry.”

Many participants pinpointed significant and rapid gains in operational and clinical efficiencies. The Director of Epic Applications at a large non-profit health system noted that success stories already exist, particularly around processes. “Ultimately, processes are going to improve,” she said, referencing the tangible impact of ambient documentation, including “how much it improves care, how patients perceive their interaction, how physicians feel about not having pajama time at night.”

This urgency is fueled by a palpable sense that the AI train has left the station.

“Clearly, AI is here to stay,” assured Geoffrey Brown, VP & CIO at Piedmont Healthcare. “I also believe we’re making inroads on it impacting labor markets, healthcare operations, and healthcare IT support overall.”

However, the adoption is not universal, and some organizations will struggle if their core strategy is unsound.

Todd Rowland, CMO at Motiv Health, stressed that fundamental organizational structure will determine who succeeds. “The question is, what are the organizations going to do to improve AI strategy, and how are they set up in terms of business models,” he posited, noting that a lack of financial alignment with current models “is going to distinguish organizations’ ability to use AI.”

This sentiment was echoed by Aalpen Patel, Chief Clinical and Innovation Officer at MedQuest Imaging, who cautioned against tunnel vision. “I think the biggest factor is that AI, specifically generative AI, has really taken over, right?” he reasoned. “When you say AI, nobody else talks about anything else but Gen AI.”

‍

Defining Success: ROI, Outcomes, and Organizational Value

When it comes to defining success to their governing boards — the true arbiters of resource allocation — the roundtable participant leaders prefer integrated financial results paired with human experience data.

Sepi Browning, Director of Technology at Piedmont Healthcare, cut straight to the core metric. “For me, it’s money coming back in from billing, from the bottom line — that’s what the board wants to see,” she stated, turning the spotlight directly on revenue. “If we can positively get past this step, then we can talk about gains like quality care. It’s a clear case of no money, no mission.”

However, leaders consistently paired financial success with patient and clinician satisfaction, arguing that one cannot exist without the other.

The Director of Epic Applications for a large non-profit health system suggested satisfaction is an interconnected ROI achieved when improvements to various areas create a positive loop. “Happy patients make happy providers,” she reasoned. “At the same time, if you have happy physicians you are going to have happier patients, right?”

Rowland extended this to include the entire workforce, asserting the goal is to elevate confidence across the board. “The goal is to take the level of confidence of everybody, no matter where they are,” he said. “We want to help them be as relaxed and confident and capable as they can be.”

Success ultimately hinges on solving a known problem, demanding organizational clarity.

Patel advised leaders to focus inward first. “I think we have to ask ourselves (and it’s going to depend on the organization): What are your strategic priorities?” he advised. “Because you can’t define success until we know what the problem is.”

Perhaps the most aspirational definition of success related to full integration emerged when leaders discussed the user experience.

Karl Kiessling, Senior Director at Medical Center Health System, proposed that AI succeeds when it ceases to be a distinct technology. “The only way I would know how to verbalize this idea is to say: When you don’t have to think about it, it’s so ingrained, so intuitive, and the confidence level is high that you no longer have to think about it,” he explained. This seamless experience allows technology to recede, restoring the focus to care.

AI’s best outcome is enabling human connection, added Chris Paravate, CDIO at Northeast Georgia Health System, further defining where this technology should reside or, more appropriately, integrate. “I think we have an opportunity to bring focus back into what we are passionate about, which is connecting people and the technology to be disruptive in the way that we orchestrate all of that workflow.”

‍

The Automation Imperative Reflects AI as a Necessity For Survival

The discussion moved from aspirational success to stark reality on the topic of what percentage of the current workload must be automated in the next few years to ensure survival. The consensus was that massive, unavoidable automation is now an economic and demographic necessity, with estimates ranging from 25% to 40%.

Villanueva said the percentage of workload that needs to be automated will vary depending on the size of the organization. “For larger systems, the number may land in the 25 to 35% range,” he explained. “But for smaller organizations, the cost of not automating is much higher, and they may need to automate as much as 40% of their workload in the next few years just to remain viable in this rapidly evolving industry.”

Rowland pushed the estimate even higher, specifically “50% of the mundane activities people don’t want to spend.”

These projections are supported by clear, quantifiable early wins.

Paravate shared a compelling example of AI in administrative areas. “We improved accuracy and efficiency by 46%, with coders going from 15 to 22 charts per hour,” he reported.

“We are all spending significant resources to review, audit, and ensure compliance,” Brown added, focusing on impact to overhead. “AI can now handle much of that process far more quickly and efficiently than traditional coding teams.”

Increased automation adoption doesn’t automatically mean lost jobs and laid off workers.

“Baby Boomer retirements and falling birth rates will lead to a reduction of somewhere between 30% and 40% of the productive capabilities of almost every domesticated society in the world,” Branzell reported, arguing that it doesn’t make sense to get rid of people at a time when we’re probably moving to a new economic revolution. “As we move from mass production to the information age, about 80% of existing jobs disappear, but a new 80% of jobs are created.”

Patel probed further into the idea that this technological inevitability is an opportunity for the future workforce, with radiology being a prime example. “Right now, radiologists are scared (of AI), but it is important that they see AI as good, not bad,” he said, referencing the expected workforce cliff coming in the next 10 years that could see the field 40% short of the personnel needed to handle demand. “In five years, they will be begging for AI.”

‍

Managing Speed and Building Trust via Governance and Culture

The greatest hurdle for AI adoption may not be the technology itself, but the organizational capacity to govern and integrate it at an unprecedented velocity.

Branzell noted the pace of innovation is likely now moving faster than governmental or bureaucratic cycles. “For the first time, I truly believe the government is totally incapable of leading the way, of keeping up,” he said. “By the time you go through a bureaucratic process, what you were trying to regulate, manage, and put policy into will have long evolved into something else.”

To manage this, organizations are elevating AI governance from an IT issue to a C-suite mandate.

Patel detailed a model of shared ownership. “I co-chair our AI steering committee, which brings together every C-suite leader except the CEO,” he said. “The goal is to make sure leaders are educated and directly involved in the decisions — that’s how you build real buy-in.”

Paravate reinforced the need for dedicated executive focus. “I meet with the CEO every month for an hour just to talk about AI,” he reported. “He wants us to be 10% better than everybody else, which is a big, broad statement. How do you become 10% better at everything? IT needs to be 10% better at adoption, 10% better at speed — six-month installs are no longer good enough.” This top-down engagement is driving faster decision-making.

Brown noted that the AI topic’s elevation at the board level is now accelerating the adoption curve: “ Now your board chair is talking about AI and how it’s impacting the business, and validation is coming from physicians and nurses as much as executives,” he said. “In just two years, virtual nursing and AI-driven staffing analytics have gone from rare experiments to widespread adoption. The health systems doing the work now will be the ones that come out ahead.”

Bonnie Boles, MD, SVP and Chief Medical Information Officer at Tanner Health, noted safety is the primary mandate for clinical leaders. “AI is becoming our most powerful tool to enforce that at scale,” she assured. “The focus should be on integrating predictive models that flag high-risk patients — not just alerting clinicians, but giving them an immediate, actionable next step embedded directly into the EHR.” Singh summed up a key takeaway: “For AI to scale in healthcare, it must be trusted. That means strong governance, transparency, and clear ROI because, without trust from clinicians and executives, even the best technology won’t see adoption.”

Culturally, building trust hinges on transparency and education.

Patel argued that AI may seem mysterious and magical, but it’s just often poorly explained. He described how he uses simple examples to connect basic concepts like linear regression to more advanced models such as deep learning, showing that even complex algorithms can be broken down into understandable parts. “I push back on the idea of AI as a black box,” he said. “The challenge is complexity and knowledge — and our job is to make that understandable.”

Browning highlighted the need to ensure that current curriculum and academic approaches embrace these big technological changes. “In many schools, the curriculum is old and lacking modern AI and even cybersecurity components,” she cautioned. “So, not only are many people already in the workforce not properly trained on these technological changes, but the new workforce is not being trained on this and will not be mentally ready for what awaits them in healthcare relative to AI and automation.”

Finally, to sustain momentum, AI champions must master storytelling.

Rowland emphasized the importance of translating governance into visible outcomes. “We need to highlight success stories in ways that resonate,” he said, stressing the need to go beyond percentages or statistics to show tangible impact such as fewer falls or reduced skin breakdowns, in ways anyone can understand. “We can even use AI itself to help paint those pictures, making the results more meaningful for each audience.”

Selling the “why” of AI is the essential work of governance in a time of unprecedented change.

‍

Conclusion: Leading Deliberately in a Time of Disruption

The central conclusion of the Innovaccer TLRT was that AI is not a technological trend; it is a fundamental economic and strategic force reshaping the healthcare workforce and its business models. The speed of innovation—measured in weeks or months—is now outpacing traditional governance structures, forcing a massive, accelerated evolution in how healthcare is delivered.

Participants agreed that survival and success will depend on proactive leadership and a cultural shift across the entire organization. This leadership must achieve three key outcomes:

• Elevating AI to the C-Suite: Governance can no longer be relegated to IT; it must be a C-suite mandate with a clear focus on value measurement, education, and strategic alignment.
• Embracing Automation for Survival: Given the critical labor shortage, automating mundane tasks is no longer optional. It is the only way to meet demands and must be embraced as an opportunity to allow staff to work at the top of their licenses.
• Prioritizing Trust and Experience: While financial ROI is critical to the board, AI must ultimately deliver happy patients and happy physicians. Building trust requires transparency, continuous education for clinicians, and the patience to weather early friction and pushback.

As the healthcare industry steps fully into this new era, the focus moves from questioning if AI will work to ensuring that systems and culture are prepared to execute, measure, and sustain the gains. The goal, as one leader noted, is to make AI so “ingrained” and “intuitive” that it simply disappears into the background, allowing humans to focus on the care and connection that defines high-quality healthcare.