Why Computer Vision Belongs in Your Hospital's AI Infrastructure Stack

Why Computer Vision Belongs in Your Hospital's AI Infrastructure Stack
Why Computer Vision Belongs in Your Hospital's AI Infrastructure Stack
David Schummers
LinkedIn Logo
Chief Executive Officer
August 17, 2026

Every health system is building an AI strategy right now. Most of those strategies start — and often stop — with language models: ambient documentation, chatbots for patients, tools that summarize notes. That's a reasonable place to start. It's also an incomplete one.

Language models are excellent at reasoning over what's written down. But in a hospital, especially in the operating room, most of what matters never gets written down at all. Computer vision is how you close that gap — and it deserves a place in your infrastructure planning, not just your innovation pipeline.

The problem: your data describes intentions, not events

EHR data tells you what a clinician documented. It doesn't tell you when the patient actually entered the room, how long the team spent on setup, whether the timeout happened before or after the first incision, or how often a room sits idle between cases. That information exists — it just isn't captured anywhere structured.

This isn't a data quality problem you can fix with better documentation habits. Clinicians are busy, documentation is retrospective by nature, and asking staff to type more isn't a scalable answer. The gap is architectural: you don't have a sensing layer for the physical world, only for the paper trail that follows it.

Computer vision as the sensing layer

Think of computer vision the way you'd think about any core infrastructure investment — like your network backbone or your data warehouse. It's not a single application. It's a capability that other applications get built on top of.

Once cameras and models are deployed in a room, you have a persistent, structured record of what physically happened: room turnover times, team presence, workflow steps, delays. That record becomes the foundation for:

  • Operational tools — block utilization forecasting, turnover benchmarking, capacity planning
  • Safety tools — timeout verification, instrument count support, protocol adherence
  • Documentation tools — automatically populated case timestamps instead of manually logged ones
  • Predictive tools — models that need real workflow data, not just billing codes, to be accurate
Learn how Apella is associated with measurable case volume increases across hospitals.
Read the full analysis

The key point for a CIO: you install the infrastructure once, and the applications compound. Each new use case built on that same camera and model layer costs a fraction of what the first one did, because the hard part — reliable perception of the physical environment — is already solved.

Why this matters more now

Two forces make this the right moment to act. First, health systems are under real margin pressure, and OR throughput is one of the highest-leverage places to find it — but you can't optimize what you can't measure, and most systems are still estimating utilization from scheduling data rather than actual room activity. Second, AI investment is accelerating fast enough that systems risk building a stack of point solutions that don't share a foundation. Computer vision, deployed as infrastructure rather than a one-off pilot, prevents that fragmentation.

What to look for

If you're evaluating computer vision vendors, the infrastructure framing changes the questions you should be asking. Instead of "does this solve our turnover problem," ask:

  • Does this platform support multiple downstream applications, or is it purpose-built for one?
  • How is data structured and stored — can it feed other systems, or is it locked into a single dashboard?
  • What's the integration path with our EHR and scheduling systems?
  • How does the vendor think about expanding use cases over time without re-instrumenting rooms?

The right answer to all four should point toward a platform, not a product.

The bottom line

Text-based AI is necessary but not sufficient. The events that drive cost, safety, and throughput in a hospital happen in physical space, and computer vision is the only technology that observes that space directly rather than inferring it from what got typed up afterward. Treated as infrastructure, it becomes the foundation that every other AI investment in the OR gets to build on.

Learn how Apella is associated with measurable case volume increases across hospitals.
Read the full analysis

Why Computer Vision Belongs in Your Hospital's AI Infrastructure Stack

David is a co-founder and the chief executive officer of Apella. David started his healthcare career over 20 years ago in finance, covering emerging medical technology companies and large diversified healthcare companiies. Through this experience, he was drawn to the entrepreneur energy and decided to transition to operating companies focused on new technologies. David has been on leadership teams that created new standards of care for multiple disease states including spinal pathologies, gastrointestinal disorders and cancers. Most recently, David joined Auris Health in 2014 as the company's first commercial executive and helped transition the company from an early start up to the largest start-up transaction in medical technology history, a $5.7B sale to Johnson and Johnson in 2019.