I’ll be upfront: I am more comfortable talking about potholes and streetlights than operating rooms. My background is in the public sector, helping governments make sense of their data. Healthcare is newer ground for me, and I’m learning it the way I would want any partner to learn my world: by listening first.
Right now, nearly every leadership team I meet wants to talk about AI, which is understandable. The conversation I keep steering back to is the one that must come first: Is your data ready for it?
The deeper I go, the more I realize that public sector and healthcare data problems rhyme. Let me share what the unglamorous end of government taught me, and why I think it holds up in hospitals and health systems, too.
In government, the data that changes outcomes is almost never the data anyone gets excited about. It’s work orders, asset conditions, service requests, or a sensor on a streetlight or water main. This is data that helps public sector organizations know which light is about to fail, which stretch of road is about to crack, and which route to plow first when the snow lands. That’s AI at its most practical: models predicting what breaks next. Each one is only as good as the work orders and sensor readings feeding it.
None of these use cases makes headlines. Prevention beats reaction, and prevention runs on the data nobody finds interesting.
Here’s the pattern I keep noticing in hospitals. The clinical and research side of healthcare already draws enormous talent and attention: bright people, academic partnerships, and real investment. That work matters, but it’s not the place a partner like us makes the greatest impact.
The starved side is everything else: the data the hospital runs on as an organization. Finance, HR, supply, scheduling, the building itself, and the dozens of systems underneath them. That’s often neglected ground, and it’s where I think the fastest operational wins are hiding. Staffing forecasts, supply chain planning, and equipment maintenance are among the most practical AI use cases in a hospital. They don’t require a research lab, but they do need finance, HR, and asset data someone can vouch for.
Whether with public sector or healthcare leadership teams, I like to ask a simple question: The dashboard you presented last week, which is driving a staffing or budget call, how many of those numbers do you really trust?
Nearly every hospital runs on dashboards now, and most of them sit on data that’s incomplete, inconsistent, or outdated. In my experience, you can put a real number on that gap, then a dollar figure on what it’s costing, and that second number tends to be the one that gets a CIO’s attention.
Unfortunately, poor data quality doesn’t announce itself. It shows up as a report someone questions, a decision made on a shaky figure, and hours spent manually reconciling numbers.
Here’s why this matters more now. A skeptical analyst will question a strange number on a dashboard. An AI model won’t. It will learn from that number and hand you a confident answer based on it. Don’t trust the dashboard? Then you shouldn’t trust the AI trained on the same data.
Data problems are commonly rooted in ownership and responsibility (or a lack thereof). In government, I’ve worked with organizations sitting on a dozen systems, and whenever I’ve asked who owns a given dataset, or who has decided what “good” looks like for it, the room has gone quiet.
Hospitals feel the same to me. Yes, someone owns the health record, but who owns the quality of the finance data, the HR data, and the asset records that reveal whether a critical piece of medical equipment is one missed service from failing?
That’s data governance, and it’s both less exciting than it sounds and more valuable than it looks. Naming your data owners and stewards. Agreeing what “good” means. Standing up a group that meets and tracks whether things improve.
It’s also where AI governance starts. When an AI tool makes a bad recommendation, the first question usually focuses on where it came from. If no one owns the data, no one can answer.
The good news is you don’t fix this all at once, and you shouldn’t try. In government we usually start with something akin to a physical for your data: Score where an organization sits across a set of capabilities, find the gaps, and hand back a plain-language roadmap the senior team can act on, including what needs to be true before AI can deliver on its promise.
People generally ask the same follow-up question: How do we compare to our peers? That benchmarking instinct is healthy. In a moment when Canadian health systems are standardizing and bringing data home for sovereignty reasons, knowing where you’re starting from matters more than usual, including when it comes to which AI tools and platforms you can realistically use.
One of my OnX colleagues has described hospitals as data-rich and knowledge-poor. To use AI to pull knowledge from your data, you must be able to trust that data. That’s the layer we would love to dig into with you. If you want to make the most of AI, start with a check-up. Reach out today to start yours.
Heading to the Canada East Healthcare Innovation Summit 2026, hosted by Bamberg Health, on November 24? My colleague Jenya Doudareva, PhD, will be moderating a panel on Digital, AI, and Automation That Actually Deliver: What Has Moved the Needle.