Blog

<h3>What Healthcare Gained. What It Lost. Where We Are.</h3>I've spent thirty years inside healthcare — as a Chief Experience Officer at Yale New Haven Health, NYU Langone, and Humana, and now as a founder building AI workflow tools for the industry I know best.But healthcare found me long before I found it. My father's illness made me a patient family member first — a child watching a system do its best and still fall short.In all that time I've watched healthcare reinvent itself repeatedly. Each reinvention solved a real problem. Each one created a new one.What I’ve come to understand is that healthcare hasn’t evolved randomly. It has moved through distinct eras — each with a clear promise, a hidden cost, and a defining human experience.I call this The Evolution of Care Framework.<figure id="figure-1rspe1e5e8mmwuxdta" class=""><img id="1rspe1e5e8mmwuxdta" src="https://cdn.durable.co/blocks/20qgIDOW3qtrQSBgbyDnkpZDtQ6rldIImHWZM9pNsdHioVQE99vMRVk8kzDY8Wwi.png" alt="A framework for understanding how healthcare has evolved through six eras — and why the age of AI may be the most consequential shift yet." title="A framework for understanding how healthcare has evolved through six eras — and why the age of AI may be the most consequential shift yet." class="object-cover bg-gray-100 rounded-2xl md:rounded-3xl lg:rounded-4xl" style="aspect-ratio: 16/9; object-position: 10% 5%;" data-corners="default" data-aspect="16:9" data-caption="" data-positionx="10%" data-positiony="5%" data-fullwidth="false" data-mce-src="https://cdn.durable.co/blocks/20qgIDOW3qtrQSBgbyDnkpZDtQ6rldIImHWZM9pNsdHioVQE99vMRVk8kzDY8Wwi.png" data-mce-style="aspect-ratio: 16/9; object-position: 10% 5%;"></figure>Relational Care The era of the known patient &nbsp;— &nbsp;Pre-1900sOnce, healthcare was personal because it had to be. The doctor knew your name, your family, your story — not because of a philosophy, but because there was no other way. House calls. Bedside manner. A relationship built over years.But its virtue was also its limit. Relational care could not scale. It reached those within walking distance and left everyone else beyond the hill. The villain: No scale. &nbsp; Institutional Care The era of the treated body &nbsp;— &nbsp;1900s – 1960s The hospital became the cathedral of medicine. Expertise concentrated. Specialization multiplied — the heart here, the lungs there, the brain in its own wing. This era solved for scale. But the cost was fragmentation. The patient became a collection of organs, each assigned to a different expert. No one held the whole story. The body was treated. The person was lost. The villain: Fragmentation. &nbsp; Systematized Care The era of the process unit &nbsp;— &nbsp;1970s – 1990s If fragmentation was the problem, process was the answer. Protocols. Pathways. Managed care. The system learned to standardize, to measure, to optimize. Efficiency soared. Variation dropped. But somewhere in the machinery, the human moment got scheduled out. The patient became a process unit — a set of tasks to complete, boxes to check. The villain: Bureaucracy. &nbsp; Consumer Care The era of the empowered shopper &nbsp;— &nbsp;1990s – 2010s The system had become efficient but impersonal. So the pendulum swung. The patient would now be a consumer. Choice. Satisfaction scores. Hospital ratings. You could shop for your care like you shopped for a car. This era solved for access — for those with resources. But it turned health into a transaction. The relationship became exchange. And those who couldn’t shop were left at the curb. The villain: Transaction. &nbsp; Quantified Care The era of the data self &nbsp;— &nbsp;2010s – 2020s Before the next era could arrive, something had to be built: the data infrastructure. Wearables. Portals. Genomics. The patient began generating data constantly — steps, sleep, labs, clicks. This wasn’t yet a new model of care. It was the bridge. It solved for measurement. For the first time, we could see ourselves in numbers. But the numbers multiplied faster than sense could make of them. The villain: Data overload. &nbsp; Orchestrated Care The era of the passenger &nbsp;— &nbsp;2025 – present And so we arrive here. The Orchestrated Era solves for intelligence. Agentic AI coordinates across every surface — scheduling, triage, documentation, diagnosis. Ambient intelligence listens in the exam room. The system acts on your behalf, seamlessly, continuously, invisibly. It knows you better than you know yourself. It predicts what you need before you ask. It orchestrates everything. And that is precisely the danger. The villain of the Orchestrated Era is Displacement — the slow, seamless experience of being moved out of your own health story. The system doesn’t replace you. It just doesn’t need you. Your judgment, your preferences, your presence become optional. You are still in the vehicle. You are still present. You are still, technically, the patient. But you are no longer the driver. You are the Passenger. &nbsp; The Question That Remains Every era solved the problem before it. And every era created a new one. In the Relational Era, we had no scale — but we had a soul. In the Orchestrated Era, we have infinite intelligence — and no one at the wheel. The challenge of the next decade isn’t building a better AI. It’s building a seat for the human that the AI can’t displace. &nbsp;

The Evolution of Care Framework

Artificial intelligence is advancing at extraordinary speed. In diagnostics, drug discovery, knowledge synthesis, and customer support, AI systems are already outperforming human cognition in defined domains. But here is the uncomfortable truth: Cognitive superiority does not automatically translate into better outcomes. Healthcare has already taught us that. Intelligence ≠ Impact Physicians are, by training, more clinically knowledgeable than their patients. They diagnose faster, understand physiology at a systems level, and remain current on evolving guidelines. Yet decades of research reveal a paradox: superior knowledge alone does not guarantee superior results. What patients need is not just information. They need connection, trust, and emotional alignment. A landmark review in the British Journal of General Practice found that physician empathy is significantly associated with improved clinical outcomes across conditions ranging from diabetes to the common cold. Another study demonstrated that patients treated by more empathic physicians had measurably better control of chronic disease and higher medication adherence. Clinical accuracy matters. But relational resonance changes behavior. Being right does not change lives. Being understood does. If this is true in medicine—one of the most knowledge-intensive fields in the world—it should give us pause in AI development. If AI Learns Only to Be Smart, It Will Miss the Point Most AI systems today are optimized for precision, speed, and breadth of knowledge. These are necessary capabilities, particularly in high-stakes industries. But they are not sufficient. If we build AI that outperforms humans cognitively yet underperforms emotionally, we risk repeating one of healthcare’s deepest systemic failures: mistaking information for transformation. A system can generate flawless discharge instructions in seconds. But if the tone feels clinical and detached, if the language ignores literacy barriers, cultural context, or emotional overwhelm, the information may never convert into action. We have already seen this dynamic in human clinicians. Why would AI be immune to it? When intelligence scales without relational calibration, trust erodes at scale. And without trust, adoption stalls. Empathy Is Not Soft. It Is Structural. Empathy in AI is often dismissed as cosmetic—something that makes systems more pleasant but not more powerful. That framing is outdated. Empathy is not a personality trait. It is a design variable. It can be intentionally engineered through: · &nbsp; &nbsp; &nbsp; Context-aware tone modeling · &nbsp; &nbsp; &nbsp; Transparent acknowledgment of uncertainty · &nbsp; &nbsp; &nbsp; Adaptive pacing based on cognitive load · &nbsp; &nbsp; &nbsp; Culturally attuned language framing · &nbsp; &nbsp; &nbsp; Follow-up prompts that reinforce understanding These are not superficial enhancements. They are performance multipliers in any domain that depends on human behavior change—healthcare, finance, education, employee experience, consumer engagement. In these environments, emotional miscalibration does not merely feel awkward. It suppresses trust, adherence, and long-term engagement. The competitive edge will not belong to the system that answers fastest. It will belong to the system that understands best. Designing for Collaborative Intelligence The future of AI is not human versus machine. It is human and machine—intentionally designed to amplify one another. This requires moving beyond raw computational strength toward what can be called Collaborative Intelligence: the deliberate integration of cognitive performance and relational intelligence. Collaborative Intelligence operates across three dimensions: 1. &nbsp; &nbsp; Cognitive Strength — accuracy, reasoning, synthesis, predictive capability 2. &nbsp; &nbsp; Emotional Calibration — tone, empathy, contextual awareness, transparency 3. &nbsp; &nbsp; Behavioral Design — nudging, reinforcement, adaptive engagement, sustained follow-through When these three dimensions align, AI does more than inform. It influences. It supports. It builds trust. And trust is the currency of every human system. Lessons Healthcare Already Learned Patient experience research offers critical guidance for AI architects: Trust is built on transparency, not perfection. Clinicians who openly discuss uncertainty and invite questions often build stronger trust than those who project absolute certainty. AI systems that obscure limitations or overstate confidence risk alienating users. Context matters more than content. A discharge plan delivered with presence and reassurance often outperforms a perfectly formatted document. Similarly, AI that senses when to pause, clarify, or check understanding creates a fundamentally different experience than one that simply outputs an answer. Efficiency is not always the primary need. In moments of vulnerability, people do not necessarily want speed. They want acknowledgment. AI systems optimized exclusively for throughput may unintentionally undermine engagement. These insights extend far beyond medicine. They apply wherever technology interfaces with human uncertainty. Beyond Healthcare: The Broader Implication Whether designing AI for finance, education, retail, biotech, or enterprise operations, the principle holds: The most effective systems will not simply be intelligent. They will be relational. Experience design in the AI era is no longer just about usability or interface clarity. It is about anticipating emotional state, adapting to cognitive load, and responding to unspoken needs. As AI becomes more capable, its success will depend less on whether it can perform—and more on whether people feel safe engaging with it. The Choice Ahead We have an opportunity to define the next phase of AI development intentionally. Rather than focusing exclusively on replicating human expertise, we can design systems that amplify human understanding and strengthen trust. If intelligence alone were sufficient, healthcare would already deliver consistently optimal outcomes. The evidence suggests otherwise. For AI to move beyond performance metrics and genuinely transform results, empathy must be embedded as a structural element of design. In the coming decade, the systems that succeed will not simply be the most intelligent. They will be the ones that people trust enough to use. 1. &nbsp; &nbsp; Derksen, F., Bensing, J., &amp; Lagro-Janssen, A. (2013). Effectiveness of empathy in general practice: a systematic review. British Journal of General Practice, 63(606), e76–e84. 2. &nbsp; &nbsp; Hojat, M., Louis, D.Z., Markham, F.W., et al. (2011). Physicians’ empathy and clinical outcomes for diabetic patients. Academic Medicine, 86(3), 359–364. 3. &nbsp; &nbsp; Mazor, K.M., et al. (2004). Disclosure of medical errors: what factors influence how patients respond? Journal of General Internal Medicine, 19(8), 716–721.

Intelligence Isn’t Enough: What AI Must Learn from the Human Side of Healthcare

&nbsp; For years, reviews have acted as a kind of shortcut for trust. Stars, badges, "top-rated" labels...enough volume and the market assumed credibility. Most of us knew the system wasn't perfect, but it was widely tolerated.&nbsp; &nbsp; The Federal Trade Commission (FTC) just made something clear: that tolerance is over.&nbsp; &nbsp; Last month, the FTC sent warning letters to 10 companies for potential violations of its new Consumer Review Rule. The penalties? Up to $53,088 per violation. The prohibitions are specific: &nbsp; 1️⃣ No misrepresenting whether a reviewer actually used the product. 2️⃣ No conditioning compensation on positive sentiment. 3️⃣ No undisclosed insider reviews. &nbsp; While the headlines focus on fake reviews, the deeper implication is much bigger...especially for AI. In an AI-driven world, reviews are no longer just persuasive signals. They're inputs. &nbsp; When Reviews Become Data, the Stakes Change AI doesn't read reviews the way humans do. It doesn't sense tone, context, or incentive. It aggregates. It summarizes. It ranks.&nbsp; &nbsp; Which means reviews - and anything that looks like a trust signal - are quietly becoming training data, weighting factors, and decision accelerators. What used to influence a buyer one decision at a time, can now influence entire systems at once.&nbsp; &nbsp; That's why the FTC's move matters beyond consumer protection. It's an early form of AI governance, whether anyone calls it that yet or not.&nbsp; &nbsp; The Problem Isn't Just Fake Reviews - It's Manufactured Context &nbsp; Much of what's been normalized over the last decade sat in a gray zone:&nbsp; &nbsp; ▪️ Pay-to-play awards that look objective but aren't &nbsp; ▪️ Incentives tied to "sharing feedback" without clarity &nbsp; ▪️ Reputation systems that reward participation over truth &nbsp; I've seen it firsthand. Last week, someone reached out on LinkedIn trying to win my business. I looked up their company, and found the same person pitching had left a glowing 5-star Google review. A former colleague celebrated their nonprofit being named "top in NYC" - by an organization where you pay to participate in the rankings.&nbsp; &nbsp; Individually, these practices felt survivable. At scale, and inside AI systems, they become toxic.&nbsp; &nbsp; The Quiet Advantage of Being Conservative&nbsp; There's an irony here. &nbsp; Organizations that have historically been more conservative, more rigorous about methodology, more careful about separating measurement from marketing, are suddenly better positions.&nbsp; &nbsp; The old trust model was about volume and optics. The AI-era trust model is about evidence and provenance.&nbsp; &nbsp; Case studies matter more than star counts. Pilot narratives matter more than praise. Transparency matters more than hype.&nbsp; &nbsp; #AIGovernance #FTC #DataQuality #CX #ResponsibleAI

Fake Reviews Were a Marketing Problem. Now They're an AI Problem.

The Power of “Stopping”&nbsp;They say AI is moving at breakneck pace. But sometimes, the most strategic thing a leader can do is stop.&nbsp;A photo I took during this morning’s snowstorm because it reminded me of a recurring friction point in AI implementation: we are so focused on the acceleration that we forget the anchor.&nbsp;Rather than only considering the “human-in-the-loop” during a review phase, we must put The Human Middle at the center of the entire lifecycle.&nbsp;1. &nbsp;The Discovery Gap: This is where I am most often called in to “fix” things. Organizations don’t fully map the actual experience of their staff, patients, or customers. When they “plow ahead” to AI-ify a broken process, they inevitably have to back track.&nbsp;2. &nbsp;The Refinement Reality: During the design phase, humans are needed to evaluate and refine the AI output. No AI project has a pristine first implementation. In my world refinement is not failure – it is the essence of success.&nbsp;3. &nbsp;The Evolution Loop: Even after implementation when metrics look “green,” &nbsp;you must re-evaluate against the changing human needs and new AI advancements.&nbsp;Without involving humans at every step, you aren’t building a breakthrough – you’re building a barrier.&nbsp;Don’t let the “exhilarating speed” of AI make you skip the foundation.&nbsp; Stop. Engage people. Then build.&nbsp;hashtag#HumanMiddle hashtag#AIStrategy hashtag#HealthcareInnovation hashtag#RoarHX hashtag#Leadership hashtag#aicxdesign

The Power of "Stopping"

𝗧𝗵𝗲 𝗺𝗼𝘀𝘁 𝗰𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝘀𝗽𝗮𝗰𝗲 𝗶𝗻 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘁𝗼𝗱𝗮𝘆 𝗶𝘀𝗻'𝘁 𝗔𝗜. 𝗜𝘁'𝘀 𝘁𝗵𝗲 "𝗛𝘂𝗺𝗮𝗻 𝗠𝗶𝗱𝗱𝗹𝗲."That's the place where AI capability meets actual human adoption—and where most implementations stall.For the past year, I've stepped away from the global stage to go deep into the technical and ethical frontiers of AI.Between traveling 6 days a week to healthcare systems worldwide and my time as a CXO at Yale and NYU Langone, I saw a recurring pattern: 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗼𝗻𝗹𝘆 𝘀𝗰𝗮𝗹𝗲𝘀 𝗮𝘁 𝘁𝗵𝗲 𝘀𝗽𝗲𝗲𝗱 𝗼𝗳 𝗵𝘂𝗺𝗮𝗻 𝘁𝗿𝘂𝘀𝘁.AI is now moving faster than most corporate cycles can keep up with. I've spent the last year getting certified with NVIDIA, Google, and MIT to translate these rapid shifts into practical, human-first workflows—filtering out the noise so my clients can focus on what actually works.Today, I'm excited to share what's next with 𝗥𝗢𝗔𝗥 𝗛𝗫.I'm helping boards and executive teams turn AI's "exhilarating but alarming" speed into clear, human-first strategies.𝗖𝘂𝗿𝗿𝗲𝗻𝘁𝗹𝘆 𝗯𝗼𝗼𝗸𝗶𝗻𝗴 𝗤𝟭 𝗮𝗻𝗱 𝗤𝟮:→ 𝗘𝘅𝗲𝗰𝘂𝘁𝗶𝘃𝗲 𝗔𝗜 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 &amp; 𝗥𝗲𝗮𝗱𝗶𝗻𝗲𝘀𝘀: Moving from "tech-first" to "human-first" roadmaps.→ 𝗣𝗲𝗿𝘀𝗼𝗻𝗮-𝗗𝗿𝗶𝘃𝗲𝗻 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝗗𝗲𝘀𝗶𝗴𝗻: Using AI to unburden staff and deepen consumer engagement.→ 𝗕𝗼𝗮𝗿𝗱-𝗟𝗲𝘃𝗲𝗹 𝗔𝗱𝘃𝗶𝘀𝗶𝗻𝗴 &amp; 𝗞𝗲𝘆𝗻𝗼𝘁𝗲𝘀: Translating complex AI shifts into strategic clarity.And this April, I'm releasing a book about a pattern hiding in plain sight—one that's been quietly shaping decisions for decades. And now that AI is learning from this data, it matters more than ever. More on that soon.If your team is ready to move past the hype and into implementation that people actually love to use—DM me "READY" or drop a comment below.

The Human Middle

Dad as the DesignerBy Joan KellyThe first experience designer I ever met couldn’t move his body. But he taught me everything I needed to know about dignity, systems, and the unseen architecture of care. He was my father, a 100% Disabled American Veteran with Multiple Sclerosis (MS), and my first teacher in what it means to design human experiences that work. I was five years old when he became a quadriplegic. From that moment on, our family became an innovation lab. Not in the way that word is tossed around in pitch decks—but in the quiet, necessary, soul-deep way of figuring out how to live. We adapted everything: mealtimes, mobility, communication. The Veteran’s Administration would bring us early tools and systems to test, and we gave them feedback not just from theory, but from lived necessity. What this taught me, and what continues to guide every system, workflow, and AI tool I build, is that care isn’t a service. It’s a system. And if that system isn’t designed around real human needs, it fails the very people it’s meant to serve. People often ask me how I “see signals” so clearly or how I come up with experience innovations that feel intuitive and deeply human. The answer is simple: I was trained in a house where missing a signal could mean pain, fear, or loss of dignity. When someone you love can’t move, you learn quickly how to move systems for them. And so, I did. And I still do. In every healthcare journey I map, in every AI-powered experience I help build, I come back to that home. To my dad. To the wisdom of invisible labor, emotional alertness, and the discipline of care. Caregiving isn’t soft. It’s strategic. It’s complex. It’s a form of systems thinking that rarely gets the credit it deserves. So, when we talk about redesigning healthcare, or applying AI to human experiences, we need to ask: have we designed for dignity? Have we made space for the quiet needs? Have we understood that behavior change begins in spaces where people feel safe enough to be seen? Care taught me how to lead. How to build. How to see. Now, it’s time our systems do the same. Enter quiet. Leave with a roar. &nbsp; &nbsp; &nbsp;

Caregiving, Dignity, and the Design of Human Systems

RoarHX

RoarHX helps organizations design the future of human-AI experience.
Founded by former healthcare Chief Experience Officer Dr. Joan Kelly, RoarHX provides frameworks, tools, and advisory services that help leaders identify where AI can transform workflows, improve patient and consumer experience, and unlock new operational value.
Through persona modeling, workflow mapping, and AI readiness assessments, RoarHX helps organizations move from AI curiosity to real-world implementation.

Insights

The Evolution of Care Framework

Intelligence Isn’t Enough: What AI Must Learn from the Hu...

Fake Reviews Were a Marketing Problem. Now They're an AI ...

The Power of "Stopping"

The Human Middle

Caregiving, Dignity, and the Design of Human Systems