“I see a clear three-step pattern emerging for successful adoption of AI in healthcare diagnostics”
Due to their proximity to patients and data, medical devices are in a unique position to benefit from artificial intelligence. But they come with their own challenges. We spoke with John Tremblay, a serial entrepreneur, digital health consultant and experienced leader in large medtech companies, about what it takes to make AI innovation in healthcare successful – and the importance of the Innovation Architect role.
John spoke with Julian Beimes and Dr. Paul von Bünau on July 24th 2025, the transcript has been lightly edited for clarity.
John Tremblay
Digital Health Consultant and Angel Investor, Healthy Context
Can you give us a short background on your experiences in healthcare?
Sure. I jumped into healthcare when I took on a home therapy innovation project with NxStage in 2011 surrounding kidney care. I didn’t have a healthcare background at the time, but they valued my deep digital expertise that included key domains including IoT, mobile apps and data visualization from various other products I helped launch.
Before that, I had co‑founded or led seven startups, so I brought a startup mentality into this established medical device manufacturer. I had many roles including being a VP of Business Development and Marketing, a product manager numerous times, R&D project lead, and even at one point handled value‑based selling for a startup that had the 2nd largest IPO in NASDAQ history at the time. That diverse skill set let me bridge organizational silos and act as a change agent in what was a very waterfall‑driven environment.
During our first pilot at NxStage, a patient approached me and said, “John, you helped save my life.” This patient had decided he would stop going to traditional in‑center dialysis because it can be such a terrible quality of life. But when we piloted an alternative home therapy option, we gave him an iPad-based self-care therapy app and he got his life back and felt in the driver’s seat again. I had previously worked at Verizon Wireless helping launch innovative communications services across their 100 million smart phone subscriber base – cool tech at the time to do RCS messaging, video chat and multicasting. But, the impact on that one NxStage patient was so different.
After nearly two years commercializing that patient and clinician innovation for NxStage, I went on to projects with Medtronic, Bayer Healthcare, a digital health–focused life‑insurance policy project for Allstate, other MedTech startups and numerous roles with Fresenius. At Fresenius, it was great to later have a hand in evaluating the organizational impact of my previous NxStage solution that served as input for NxStage’s $2 billion acquisition by Fresenius.
Of all these interesting companies and roles, what was the most impactful thing that you ever did?
I think it was developing the two renal care therapy‑management products, Nx2me and Kinexus. There are over 34,000 patients in the U.S. that are actively treating at home with it right now. It’s a huge differentiator for NxStage, helping with the patient experience and retention, and keeping competitors at bay.
The NxStage home hemo-diaylsis machine relied upon LED visual indicators that were hard for many patients to handle. By contrast, our iPad‑based application was something patients and their caregivers truly craved because it automated much of the required documentation, interaction with their nurse, and provided a range of insights into their therapy. Just one example: one woman shared that when doing her chores she now would take her iPad out to her barn while feeding her chickens, with the comfort her loved one was safe in real time while they were performing treatments. So I’d say that’s pride.
So you were working on digitally enabled innovation early on. When was your first contact with AI? I assume pretty early?
Prior to focusing on healthcare, I was VP of Product for a big data and machine learning startup processing more than 15 billion events per day with descriptive and predictive analytics for consumer behaviors on mobile devices. When I moved into healthcare, I immediately could see the opportunities and impact from digital.
At NxStage, they were mostly blind to the patient experience at home and we started remotely collecting raw machine data. We could see a range of patterns emerge to predict machine performance, whether one patient would succeed on therapy, which patients nurses should proactively reach out to, and many ways to automate manual workflow tasks with clinicians, customer service and technical support. At the time, the industry did not have ubiquitous access to mature AI tools and platforms, like is in broad use today.
Given the focus on an MVP, we never commercialised those advanced data insights, predictive models and automations in the two years I was at NxStage. But we could see different patient personas emerging with the help of the daily treatment, vitals and qualitative data we were receiving.
This feedback along with human factors studies helped us fine tune our roadmap to address the needs of both novice and more advanced patient users. We knew automation was needed to adapt to different patient personas and behaviors to ultimately address early signs of non-compliance and potential churn off the therapy that is quite costly on healthcare organizations. So, instead of driving manual follow‑ups with nurses, doctors, and the NxStage support staff, we were already thinking about automating both the risk‑stratification and intervention processes.
Adoption of AI has been slower in healthcare/MedTech than in other industries. In your experience, what has been critical to the success stories you have seen?
I see a clear three-step pattern emerging for successful adoption of AI in healthcare diagnostics, and I’ll share two examples.
First, I just started engaging a blood‑analyzer company doing point‑of‑care CBC tests. They have a chemistry analyzer, and they’ve now digitized the blood‑cell count, which is narrow in scope, using digital microscopes. I have encouraged them to not think about just a single point product with a point‑of‑care CBC analyzer. Instead they could leverage a broader platform play. Just like digital microscopes opened up an entire industry in digital pathology, they have an opportunity to become a digital hematology platform. From there, you can feed the images into a cloud platform to layer in more visualizations, build hematology AI models and operationalize a range of algorithms from the research community – and really own the platform and possibly leverage an ecosystem play rather than focus on just a one‑off test.
I also recently did diligence on a New York City startup called Remedy Logic that’s tackling spine MRI interpretation. A typical spine MRI can take a radiologist about 20 minutes to read – identifying and measuring anatomical structures, assessing stenosis and other issues. Remedy Logic’s AI can reduce that down to about a minute and a half review by a radiologist. It takes all measurements and uses a large language model to pre-populate a report, which the radiologist then reviews and finalizes. As another stress point in healthcare, radiology reads are outpacing available resources, so the impact can be significant and actually help the radiologist make more money per shift. While currently FDA approved for interpretation, they are not yet doing diagnoses, but they can and they will. And because they have one of the largest spine MRI databases, they will be able to correlate that with the interventions and therapies patients receive to see which have best outcomes. So eventually they are going to start providing recommendations to clinicians.
What ties these two examples together is a three-step pattern: First, automate interpretation of highly repetitive quantitative and operational tasks. Second, once interpretation is reliable and helping address stress points in current workflows, layer on models that suggest or prepopulate diagnoses under a limited regulatory indication. Third, with real‑world outcome registries, you can then build engines on top that recommend therapies based on what has been proven to work. There are many examples of successful diagnostics following this 3 step course, including Paige AI [acquired by Tempus] that has successfully traversed all 3 in the pathology and oncology space.
From our collaboration, I know that UI/UX is clearly a topic close to your heart. What would be the ideal UI/UX paradigm for all the AI applications we will see in healthcare?
It is funny you ask. One of the other projects I’ve been working on is a next‐generation EHR for kidney care. I evaluated Epic, Cerner, and some more innovative platforms like Healthie and Canvas.
The problem with most EHRs today is they’re built around a very patient‑record–centric view of the world, which works okay for general medicine and acute care, but is terrible for many specialty care settings, all of which have unique workflows and way too many clinician clicks and eyeballs navigating the patient’s chart. If you need to build a single specialty dashboard in Epic, you’re literally breaking apart three different module pages and rebuilding them. It’s a major forklift and that's why Epic and Cerner charge hundreds of millions of dollars to customize it for each customer. Unfortunately for many smaller provider practice or specialty areas that level of spend and level of effort isn’t realistic.
Now with AI, it becomes crystal clear that the magic isn’t just in the models and algorithms – orchestration and user experience will be even more important. You need a seamless service‑orchestration layer at the core. Think of it like a plug‑and‑play UI: you drag in an enrollment engine here, slot in a monitoring module there, and swap in a triage workflow, all without rewriting the backend. AI suggestions, data visualizations, and user actions must flow naturally from one step to the next, with as little friction as possible. That’s the AI‑centric UX paradigm I’m most excited about.
We are starting to see the new breed of next gen EHR players starting to take off that have this inherent flexibility. The high profile IPO Hinge Health is using the Healthie EHR to power their virtual physical therapy and chronic pain management offering for musculoskeletal care.
So basically bringing lean UX into the clinic?
Yes. Look at AI scribes that everyone is now trying to leverage. It’s going to be commoditized really quickly, there are a lot of me‐too offerings. Epic and Cerner even packaged some AI scribe functionality, but focused narrowly on clinician notes. My question: why stop there? You should be able to invoke AI scribing for every function in the EHR. Need to place an order? AI can suggest the pharmacy and dosages. Onboarding a patient? Let AI draft the forms and consents. Triaging and assessments? AI can push automated surveys or interventions – like what Andor Health is doing – pre-processing patient responses before bothering the clinician.
To make that work across specialty care workflows, e.g. kidney dialysis, physical therapy, cardiology, remote patient monitoring, etc., you need truly flexible service orchestration. I saw this early at NxStage when they moved the customer‑service team off a rigid, custom built platform onto a 4GL drag‑and‑drop system. Suddenly, the product manager themselves could tweak order‑supplies flows and optimize the experience in real time, no IT ticket required.
That user‑driven orchestration is exactly what next‑gen medical software needs. With it, you can plug in new AI tools, experiment freely, begin to close end-end care loops and continually refine the end‑user experience without ripping everything apart. This is a key enabler to shift healthcare from slow waterfall to nimble lean UX/Agile methods like where the rest of digital innovation has gone.
What you describe requires bringing together a diverse set of skills. However, traditional organisational structures are often silo-ed by discipline. What are your tips to counter that?
One way to counter silos is to create true cross-functional teams led by a key change agent who can create a sense of urgency, build momentum, and build bridges with functions that are most risk averse and new to digital innovation. For example, when leading the commercialization of the NxStage’s digital health offering early on, I needed to be well versed on the UX, architecture details, embed myself in the development team, and distill down our plans into language our legal, medical and regulatory could most easily understand since most were new to digital. Other times, I had to help re-write certain R&D and risk management processes more optimized for digital. Staying in everyone’s own swim lane was not going to cut it. When I needed to give a label to my role that did not fit traditional company functions, the term Innovation Architect seemed to be the best fit, and I am currently writing a book about this concept.
What companies often do instead is hire project and programme managers. That’s fine in theory – but in practice, they often can’t speak the language of other functions and so they need to pull in the regulatory person, the development lead, a UI person, and so on. Which just means more meetings, more people involved, and slower progress. Plus, when embracing digital/AI technologies and more lean UX/Agile methods within traditional healthcare organizations, you must have someone who can more broadly drive an innovation and increased impact mandate.
An Innovation Architect is a crucial leader who enables teams to stay small, keep decision cycles short, ensure that targeted impact stays on track, and all stakeholders are covered throughout the process.
John, thank you for these insights into your work and AI in MedTech.
My pleasure – thanks for having me.
About John
John Tremblay has 25 years of experience in executive and product innovation roles in tech and healthcare. Trained initially as an electrical engineer and with a technology innovation MBA (both Northeastern University), he has a strong track record spanning seven early stage startups, large MedTech companies and digital health consulting. Recently, he was Vice President of Digital Strategy & Portfolio for the $5B MedTech renal care business at Fresenius Medical Care.
Within renal care, John helped launch 2 successful life-saving home therapies now serving 34k+ patients in the US with world-class rated clinician experiences.
John is also an angel investor with LaunchPad Venture Group and leads the digital health innovation consulting practice at Healthy Context, LLC.