Why clinician-led product development matters more than ever
For clinicians such as Matt Sakumoto, MD, Chief Clinical Product Officer at Nabla, who have spent years evaluating tools from the outside, the shift from advising to building can feel like stepping onto an entirely different playing field. At Nabla, that means working directly with clinical customers, engineers, and product teams to shape how ambient and generative AI function within clinical workflows, rather than providing feedback through a health system's broader governance structures.
“Joining Nabla has opened up a whole new world where I have the opportunity to be part of a team that is fundamentally changing how we practice medicine,” he says. Another benefit: “I also use the product every week and get to experience what it’s like as a customer, which makes the feedback loops incredibly fast,” Sakumoto says.
Sakumoto’s views come from his time as a regional CMIO at Sutter Health, during which he evaluated and rolled out new technologies within a large health system. This gave him a clear sense of what works in clinical workflows and how to make them successful. In this interview, he explains why he moved from health system leadership to product development, what it means for clinicians to lead AI projects, and what health system leaders should look for when assessing ambient and generative AI tools beyond just their features.
Q: You’ve spent years inside health systems, often evaluating or integrating new tools. What specifically about Nabla made you decide this was a place you wanted to build from the inside, rather than advise from the outside?
Sakumoto: AI is changing medicine much faster than it did just five years ago, and that really motivated me. In a big health system, I could only make small changes to an AI tool from the EHR vendor, but it taught me a lot. Moving to Nabla felt like the natural next step.
I was impressed early on by Nabla’s speed of execution. I now work with a team of engineers and product managers who are at the forefront of technology. Tasks that used to take weeks or months in a large system now happen in just hours or days. Their ability to pull insights from a visit in real time, particularly compared to other ambient AI assistants, stands out. The team can really be present during the visit and anticipate what comes next, which helps other doctors and me every day. I’m also inspired to use technology in ways I hadn’t conceptualized before. Nabla also truly cares about making medicine more human again.
Q: In your first months at Nabla, what’s a concrete example of a clinical or operational constraint that forced a product trade-off? How did you decide what to prioritize and what not to build?
Sakumoto: Often, systems have to choose just one of a few options. But in healthcare, product trade-offs are really about balancing standardization with customization. It’s impossible to meet every single person’s needs. We work with many large academic medical centers, and a lot of their clinicians are highly specialized. The challenge is supporting the specialization while still rolling out solutions across an entire division or health system. I’m proud that we haven’t had to pick one over the other. We’ve helped specialized clinicians with custom instructions and workflows, while still scaling up across the whole organization.
Q: What does effective collaboration between clinical leaders and engineering teams actually look like when you’re building AI for regulated health system environments? What has worked for you, and what hasn’t?
Sakumoto: As a doctor, I’ve always believed in the idea of 'going to the bedside.' It’s one thing for customers to send in questions, but it’s completely different to see how things work in a real clinical environment. Nabla invests in bringing our engineers and product teams to the clinic to watch and learn. That’s where we find real insights and true teamwork. Building exactly what the customer asks for doesn’t always work, since there might be a simpler answer. Instead of adding just a new button, our team is good at figuring out the real problem and making things simpler. Good collaboration also means working closely with health system leaders, understanding their goals, and helping them manage change.
Q: Based on your early experience, what do you think CMIOs evaluating ambient or generative AI tools should be looking for beyond feature sets, especially when it comes to adoption, trust, and long-term sustainability?
Sakumoto: Sustainability is crucial. You need to ask who will be a strong partner, and who can grow with your organization over time. It’s not only about features. It’s also about finding someone who understands the whole picture of clinical and care team workflows, and how tools like ambient AI and clinical assistants fit into that process.
For example, one of our larger insurance clients has many care managers. We had to learn which insights were most important for them to document and make sure that information could be shared across different touchpoints. We’re not just capturing a single visit or conversation between a care manager and a patient—we’re helping tell the longitudinal story.
Another key part is staying vigilant as AI evolves. Technology has both inputs and outputs, and it’s important to know how the AI handles each and to watch them closely. We avoid black-box solutions. If we suggest a diagnosis or give a clinical nudge, we explain why and show our reasoning. That transparency builds trust and keeps patients safe.