Pager Health Innovation & Technology Culture

How does innovation show up in your company culture?

At Pager Health, innovation is a daily practice, not a one-off initiative. Working at the intersection of AI and healthcare means curiosity and responsibility have to coexist. Our culture encourages teams to ask hard questions like, “Should we build this, who does it help, and where do clinicians need to stay in the loop?” before jumping to solutions.

Innovation also shows up in how cross-functional we are. Product, engineering, data science and clinicians collaborate closely, ensuring we’re not just building technically impressive systems, but ones that actually work in real healthcare settings. One of our secret weapons in our products that's been here since day one at Pager is how much clinical thought is baked into them. Our goal is not replacing physicians. We’re giving them AI superpowers so they can enjoy practicing medicine again. And the patients who see them? They receive faster and elevated care. Wins all around. 

At Pager, we value people who think deeply, challenge assumptions, and stay grounded in the lived experiences of patients and clinicians. That balance is what allows us to innovate responsibly.

What’s one recent innovation that improved user or employee experience?

One of our most significant recent innovations was the development of proprietary safety and clinical guardrails that allow us to responsibly expose a LLM directly to members in a healthcare setting. Internally known as Project Helix, this work took over a year and brought together PhDs, medical doctors and an exceptional team of engineers. I’m incredibly proud of the team not just for what they built, but for how thoughtful and rigorous they were throughout.

The goal was to solve one of the hardest problems in healthcare AI: enabling real-time, member-facing AI in high-stakes clinical scenarios without compromising safety or trust. The system actively de-risks situations like crisis emergencies and behavioral health events with real-time mechanisms that detect and intervene when patterns of harm emerge and bring clinicians in when needed.

We also invested in building systems that can identify hallucinations, intervene in the moment, and continuously refine the model. The result is a better experience on both sides: Members get faster, clearer guidance, and clinicians gain confidence that AI is augmenting care responsibly, not introducing risk.

How do you balance experimentation with stability?

In healthcare AI, stability starts with accuracy. We’re willing to experiment when the benefits are clear, but only within very high performance standards. An accuracy mistake in healthcare isn’t like getting a recipe wrong; it can directly impact someone’s loved one.

Before any model leaves experimentation, we test extensively across edge cases, real-world scenarios and clinical contexts until we clearly understand its accuracy and behavior. We don’t consider a model stable just because it works most of the time; stability means we can measure its accuracy, trust its outputs, and stand behind it in high-stakes situations.

Once in production, we continuously monitor model performance. If accuracy drift changes, our data scientists are alerted immediately so we can intervene to refine the model. That closed loop is how we innovate while maintaining the standards healthcare demands.

If you’re evaluating a healthcare company that says it’s using AI, ask questions about testing, accuracy and ongoing monitoring. Responsible AI in healthcare requires real rigor and discipline — and teams who lose sleep over getting it right (I have the Slack messages to prove it).

What tools support your day-to-day work?

Aha, Jira, Claude Code, Figma, Jupyter Notebooks and Python.

How does your team experiment?

Discovery runs as a staged, evidence-driven process with explicit go/no-go decisions at each gate — not a continuous flow where ideas quietly graduate into roadmaps.

It starts with a draft product opportunity assessment. Before we commit any real cycles, we pressure-test whether the opportunity itself is worth pursuing. Is the problem real and painful for a specific market segment? Is it big enough to matter — does the TAM clear our bar? Does it fit our strategy, where we have the right to win, and what are the competitive alternatives? The draft POA forces crisp answers to those questions and drives the Stage-Gate 1 decision: go or no-go. If it can’t clear that bar, it doesn’t move forward, no matter how interesting it is technically.

If we go, we move immediately into rapid prototyping, not to build production software, but to generate enough evidence to either kill the idea or sharpen it. The prototype is a tool for market conversations, not an engineering commitment. We take it directly to customers and target-market buyers: Does the problem resonate? Does the proposed shape of a solution feel compelling? What would they pay for? What would they refuse? We conduct a week of prototyping.

How does Pager Health adapt to change? 

The sharpest example right now isn’t a market shift or an org change — it’s how fast AI is collapsing the traditional boundaries between product, design and engineering. The roles we played twelve months ago aren’t the roles we will play tomorrow, and pretending otherwise is the fastest way to become irrelevant.

A year ago, a PM wrote requirements, a designer mocked the flows and engineers built the thing. That assembly line is breaking down. PMs are now prototyping working software in an afternoon. Designers are shipping functional components, not just Figma files. Engineers are operating at three to five times throughput with AI pair-programming and spending more of their time on architecture and judgment than on boilerplate. The lines between who specifies, who designs and who builds are blurring fast, and the teams that win are the ones where everyone is a builder.

The honest part: None of us knows exactly what a product org will look like in 18 months. The specific skills, the ratios, the titles — all of it is in motion. The adaptation isn’t landing on a new steady state; it’s building an org that can keep re-learning faster than the technology keeps changing.