Joshua Liu

I was one of the 50+ people interviewed for this series on Epic. I DON’T agree with many Health Tech startups who shared their views for this piece. Partly because we all have our own unique n of 1 experiences - but also because I’ve evolved my views on markets and competition after building @SeamlessMD for almost 14 years now. We started working with Epic in 2018, initially integrating Epic and MyChart with our digital care journey platform. I didn’t know it but that same year Epic also released a similar tool in MyChart called Care Companion. Before folks take this out of context, I’m 100% sure timing was coincidental. But here’s the point: despite that, they integrated with us that same year. And that was great for our company. A few years ago, we ran into several health systems who wanted to be Epic-first, but they loved our care plan content: could they get our content in Epic’s Care Companion? For a while I had similar frustrations expressed by the other startups in this piece. But as SeamlessMD continued to grow, I became immersed in the principle of Customer-Centricity. Did we want to listen to health systems or just complain? What would add value to health systems? For those who preferred to be Epic-first, this meant collaborating with Epic to make our content available in Care Companion. So we did it and became Epic’s 1st and so far only Toolbox partner for Care Companion. Throughout all this, nothing has changed for our core product when it comes to Epic and MyChart integration. In fact, over the years, we have INCREASED our integrations with Epic, and their team has often been involved in helping to make this possible for our health system customers. And adoption of our core digital care journey platform? Doing great. It continues to be used by health systems on Epic. Now we also have the bonus of working with Epic-first health systems who otherwise would’ve ignored us. Think about it: are health systems only telling startups about new problems to be solved and thus new products to build? If anything, they are far more likely to tell their main vendor partners like EHRs. The difference is that Epic, unlike many incumbents in most industries, is far more likely to listen and build new stuff. Is that customer-centric or slowing innovation? Depends on your perspective I guess. I believe it’s a company’s job to earn their health system partnerships year after year. I also believe that it’s common industry dynamics that a startup has to be 10x better on a dimension the customer cares about to overcome the incumbent’s gravitational pull. As an innovator, you can choose to vent/complain or you can spend your time listening to customers and build. I’m not saying startups shouldn’t voice their disagreement or frustration - they absolutely should. But for me and my team at @SeamlessMD, we have a finite amount of time and energy, and we’re putting that towards building for patients and the health systems that serve them.

10 years ago, we filmed physiotherapy exercise videos - then gave up. The cost was unsustainable... BUT AI changes that. Today, I’m excited to announce @SeamlessMD’s AI-generated patient education videos - making visual education scalable. First, let me take you back… We hired a film crew, a physiotherapist and a standardized patient to produce 10 physiotherapy exercise videos for recovery after surgery. We got a fantastic film editor to turn that footage into beautiful videos that met evidence-based guidelines for patient education. We were proud of the outcome. The trouble started when we began making them available to our health system partners. We kept hearing stuff like: “Yes, we ask patients to do that exercise after surgery… BUT our physiotherapists ask patients to do it slightly differently - so we can’t show this video to them. They would get confused.” Multiple health systems. Same exercise in theory… different ways of teaching. Given the time and cost of making a single high quality exercise video, creating custom videos for every health system was unsustainable. So we gave up. We provided evidence-based static illustrations instead - not ideal but still better than nothing and scalable too. A few health systems filmed their way of doing exercises and we embedded those videos into their care plans. But then comes 2026, when AI video generators make it scalable to not just make video - but to customize exercises the way each health system wants. Which is why @SeamlessMD now provides AI-generated patient education videos: a growing library of ready-to-use visual education designed to help patients better understand and manage their care - plus the ability to customize for health systems as needed. Our goal: high-quality visual guidance at scale, aligned with a health system’s existing education, without expensive traditional video production. What’s critical is HOW we’re doing this. Rather than generating health information from scratch, @SeamlessMD uses AI to transform our established, evidence-based patient education content into video demonstrations and explanations. Every AI-generated video undergoes human editing and clinical review before reaching patients, combining the scalability of AI with human oversight. So where are we starting? The initial rollout focuses on physiotherapy exercises for patients undergoing hip and knee replacement surgery - now available to our health system partners. And if you need it tweaked? We can do that too. Coming soon: → Exercise videos for spine, cardiac and thoracic surgery. → After that, we’re working on chronic care, oncology and more - and well beyond exercises (e.g. videos on understanding disease, nutrition, etc.) → Videos in multiple languages Huge credit to Denice Koo, our VP Patient Education, on leading this innovation - and ensuring we use AI in a way that is evidence-based and patient-centric. More to come!

Wow, what a convo with Richard Milani, Sutter Health’s first Chief Clinical Innovation Officer, where we chatted about remote monitoring, health literacy, payment models and what real innovation looks like. Here are my 10 biggest takeaways from Dr. Mliani on this week’s episode of The Digital Patient podcast: 1/ Why he calls fee-for-service the most insane thing in healthcare, and the one-line thought experiment that makes his case. 2/ Why Sutter doesn’t bill the CMS RPM CPT codes AT ALL. Plus his take on why the new outcomes-based ACCESS model won’t work yet. 3/ How Sutter took patients with long-uncontrolled hypertension and got 90% in control within six months - compared to 23% under control nationally. 4/ Why no patient gets discharged from Sutter’s RPM program - and how between 20% and 30% of people in control fall out within a year, and with continuous data, his team can often see it coming 3 to 5 weeks ahead. 5/ Where most health systems go wrong on health literacy before they even start: they never measure it. 6/ How he builds the ROI case for innovations that don’t show a clean financial return, starting with an analogy about raising children. 7/ What the ROI looks like when you get blood pressure and diabetes under control and why Medicare Advantage changes who has the incentive to care. 8/ Where he draws the line between incremental improvement and innovation. 9/ How he built a “focused factory” for chronic disease: no brakes, no windshields, just transmissions. In particular, how physician-level guideline adherence is generally under 50%, which is why he took this out of the physician’s hands. 10/ How a sliced apple explains why SutterSync devices go straight into Epic with no extra app to download. Sometimes the whole strategy is one less step. … and much, much more!

Someone pinch me: New Brunswick is the 1st in Canada to do a province-wide rollout of Patient AI for healthcare journeys (e.g. surgery, chronic care.) - and it’s powered by @SeamlessMD! 9 takeaways from the stories that have come out of our amazing partners @HorizonHealthNB and @VitaliteNB leading this rollout: 1/ Horizon Health Network and Vitalité Health Network, the province’s two health authorities, have both launched Seamless Answers: a clinician-approved AI assistant that gives patients answers 24/7, inside the SeamlessMD app their patients already use for their care journey. 2/ Patients across heart failure, heart surgery, orthopedics, maternity, colorectal, gynecologic oncology, pancreas surgery, liver surgery,  and hepatopancreatobiliary journeys can now ask a Conversational AI questions like “when can I remove my dressing?” or “is this symptom normal?” and get an instant answer. 3/ Patients were already asking ChatGPT about their care journey- but ChatGPT knows nothing about the clinical pathway their care team put the patient on and could provide instructions that conflict - creating a safety risk. With Seamless Answers, patients receive immediate, appropriate answers. 4/ Already 650+ patients across New Brunswick have asked the AI nearly 2,000 questions! 5/ 65% of patient questions at Vitalité come outside regular business hours, mostly at night and on weekends. That is when clinics are closed and when patients are most likely to feel stuck. Seamless Answers is clearly addressing a need. 6/ For patient safety, the AI uses Retrieval-Augmented Generation (RAG) to answer only from each health authority's clinician-approved education library in SeamlessMD, not the open Internet. Every answer cites its source with a link to the original document.  7/ The AI understands and responds in multiple languages - eg perfectly bilingual English and French for citizens in New Brunswick. 8/ For anything out of scope where the AI does not have relevant content for, it appropriately directs patients back to the care team. 9/ 94% of patients at both Horizon and Vitalité who tried the AI found it helpful. Thank you to Horizon Health Network and Réseau de santé Vitalité for the courage and conviction to lead the way with clinically appropriate AI for patient care. Grateful for SeamlessMD to be your partner on this journey! It’s only by partnering with healthcare teams - not circumventing them - that AI can go beyond consumer health education and move the dial on outcomes like reducing ED visits and readmissions.

We sat down with Dr. Keith Morse CMIO of Stanford Children’s, and I didn’t expect how candid he’d be about AI governance, build vs buy and pediatric informatics. Here are 8 things he said that will make you rethink your health system’s AI strategy: 1/ Where health systems need to get honest about AI governance. The pre-LLM playbook of auditing training data and retraining the model doesn’t work anymore, and most tools offer little more than an on/off switch. His answer on where the responsibility lands instead is thought provoking. 2/ Why he does NOT want clinicians to trust their AI tools. He shares what happens when an AI scribe turns an eight-year-old’s visit into a note about an adult woman, and why his answer to the clinician who complains is “that’s just what happens.” 3/ Why many AI projects succeed for reasons that have nothing to do with AI: people show up, care and change their workflows in a way they never would for an EHR optimization. 4/ How he picks use cases now that one model can serve a thousand of them. His example: an LLM to support ADHD medication prescribing, and why designing the project around that one use case may undershoot the impact. 5/ Where Stanford Children’s draws the line in its OpenAI rollout, as one of eight institutions in the program. Hint: almost all use is administrative so far, by design. 6/ What share of DIY AI solutions end up fitting into Stanford Children’s. Hint: his answer is 95% “sorry, this isn’t gonna fit,” and he makes the case with an EHR analogy and a memorable line about bubblegum and duct tape code. 7/ Why he doesn’t block ChatGPT, Claude or Gemini on the network. He explains the “pop-off valve” he built instead. 8/ Why a pediatric-specific LLM “feels silly” to him. 90% of the care process is the same as adults, and he says he’ll happily take a tool that’s 10% less valuable but scalable. … and much, much more!

Today Epic announced Connectors: a way for its own AI agents to interact with 3rd party apps. Gamechanger or nothingburger? My 4 thoughts: First, what we know: → “Connectors” let its AI agents (Emmie for patients, Art for clinicians, Penny for operations, custom agents built in Agent Factory) pull info from and complete tasks in 3rd party apps - e.g. an ED doc asking Art to check an ambulance app for arrival times or a patient asking Emmie to order a meal directly from a hospital’s dietary management app → Developers will build Connectors using the Model Context Protocol (MCP), an open standard → Health systems stay in control, with their security, legal, and compliance teams reviewing what each Connector can read and change before any data flows → Connectors will be available to health systems starting in March 2027 Alright my 4 thoughts: 1/ The overly simplistic way to read this is that it’s a way for Epic to keep users inside its own interface. I think this makes 3rd party apps stickier and more useful because it reduces friction to make 3rd party apps useful. Why? Because agents with access to other apps allow us to take more actions we otherwise wouldn’t have time or resources for. For example - my company pays for Customer Success software (a post-sale CRM in many ways), and my CS team uses it regularly, but as the CEO I’m rarely inside it - but I often use Claude to query the CS software for insights. That makes it way stickier because I’m getting more value from the data. You could imagine clinicians and patients using 3rd party apps more because of new tasks unlocked by AI agents. 2/ This might be a neat way to help 3rd parties learn what use cases are most valuable to health systems and then build what’s truly needed. E.g. imagine a digital care journey company like mine (@SeamlessMD) didn’t yet provide remote monitoring dashboards to a care team. But we notice clinicians are asking Art “how did a patient’s pain level trend post-op via the data they reported in SeamlessMD?” We could turn that into specific dashboards/reports to display in Epic. 3/ The one big open question is if and when this becomes bi-directional. Right now, this only allows Epic’s AI agents to pull data in and take action in 3rd party apps - it doesn’t allow 3rd party apps to do the same. Bi-directional access would make this even better for the ecosystem. What if a patient prefers to ask “when’s my next appointment” in ChatGPT? Will Epic make it easy to do that or try to keep everyone in MyChart and Emmie? Will be interesting to see how that plays out. 4/ From a business strategy POV, it’s presumably a way for Epic to grow revenue as the ecosystem around it grows. You’d imagine they’ll charge 3rd parties fees to capture some of the value this creates. Hard to imagine this revenue will be material to Epic’s business but who knows? I think we’ve all underestimated how big the market becomes as dollars spent on labour increasingly shift to AI agents.