What Healthcare AI Can Do When It Knows How You Work

BY ANDY FANNING, CO-FOUNDER & CEO

The first AI initiative you deploy is the hardest. The second should be faster. The fifth should outperform the first four by a margin that makes the early ones look like proof of concept.

That’s how it works when the foundation is right. And the foundation is context: a maintained, organization-specific understanding of how your company operates, built into every initiative you run and compounding every time it’s used.

Let me give you an example of what context looks like. I was recently in a workshop with the C-suite of a Blue Cross Blue Shield plan. As an icebreaker, I had the team describe a picture they’d like to see, with several people calling out ideas, and then Google’s Gemini (“nano banana”) generated it on the spot. It’s a fun, low-stakes way to get a room comfortable with AI.

This group had a few die-hard Knicks fans still riding high on the team’s recent championship, so that’s what they picked. The prompt was something like: “Madison Square Garden with the Knicks winning the championship and the Spurs looking dejected in the background. Put our logo on the Knicks jerseys.”

Here’s where context matters. The image Gemini produced wasn’t the recent championship the room expected. It was the 1973 team, the one Google had far more context on. Harmless in an icebreaker. Not harmless if you’re counting on AI to read a medical image or adjudicate a claim. Without the right context, the output is confidently wrong.

Most healthcare organizations are building AI initiatives. They’re not building that foundation. Those are not the same thing, and the difference determines whether your AI program grows or plateaus.

When AI has real organizational context, it doesn’t just perform. It compounds. And that changes what’s possible. Not in the abstract, but in the next initiative you stand up.

There’s a difference between AI that can do something and AI that can do it for your organization.

Most deployments are running the first version. The capability is real. The integrations are complete. The use case is valid. And the output keeps running into organizational reality at every edge, because AI is working from a generic model of how healthcare organizations operate, not a precise model of how this one does.

Context closes that gap. And context isn’t data. Every health plan and health system already has more data than it can use. Context is the organizational knowledge that doesn’t live in any system: how contracts get fulfilled in practice, what the regulatory environment looks like for this plan in this market, where decisions actually get made, the exception-handling logic that experienced people apply automatically and that nobody ever wrote down because everyone who needed to know it, already knew it.

When AI has that knowledge, the recommendation that’s technically correct stops being operationally impractical. AI knows which constraints are fixed and which are negotiable. It knows how this organization makes decisions, not how health plans generally make decisions. It can navigate the edge cases that aren’t documented, because they’re part of the context layer now.

Medical expense management is a useful example of what that enables. Managing medical expense at a health plan isn’t a discrete problem. It spans actuarial assumptions, clinical programs, network strategy, and regulatory compliance simultaneously. AI working from general capability can handle pieces of it. AI working from real organizational context can work on the whole thing — the organization’s structures, programs, and decision patterns included.

Context doesn’t improve performance at the margins. It changes what’s tractable. The recommendation that’s technically correct but operationally impractical isn’t an AI problem. It’s a context problem.

This is the part that changes the economics of healthcare AI.

Every initiative that runs through a real organizational context layer adds to what it knows. Every workflow mapped, every exception documented, every decision recorded makes the next initiative smarter from the start. AI’s understanding of how the organization works deepens with use. It doesn’t reset between initiatives. It builds.

The timeline effect is direct. An initiative that took six months to stand up the first time takes six weeks the third time, not because the organization got better at deploying AI, but because the foundational work is already done. The discovery work — mapping how the organization operates, identifying who holds the institutional knowledge, understanding where the constraints are — happened in earlier initiatives and accumulated into the context layer. Each new initiative inherits what all the previous ones learned.

This is what ROAI™ (Return on AI Investment) looks like when the foundation is right. Not one strong result from one initiative, but a trajectory. Each initiative is faster to stand up. Each one is more precisely calibrated to how the organization works. Each one generates a return that compounds into the next.

Most organizations measure AI initiative by initiative. The ones building context as infrastructure measure a curve. The initiatives get more valuable as the context layer grows, because the context layer is growing.

AI that doesn’t compound isn’t a foundation. It’s a project.

Most healthcare organizations are in the initiative-management phase right now: figuring out which use cases to fund, how to prioritize them, how to measure what comes back. That’s the right place to start.

It isn’t the destination.

The organizations that run differently because of AI are the ones that stop managing AI initiatives and start running the company on the intelligence layer they’ve built. That requires a context layer that is real, current, and deep enough to inform not just individual projects but how the organization makes decisions at scale.

What changes when that layer exists? Medical expense management stops being a siloed initiative and becomes a cross-functional capability that improves as it runs. Strategic planning draws on organizational intelligence that reflects how the company actually operates. Not how it was described in last quarter’s offsite. The decisions that used to require months of alignment run on a foundation that already knows the constraints, the history, and what’s already been tried.

The leaders getting there first aren’t waiting for better AI. They’re building the organizational context layer that makes AI worth deploying at scale. The technology is ready. The question is what you’re giving it to work with.

The organizations that win on AI won’t win because of better models. They’ll win because of what their models know about them.

The first initiative is the hardest because the foundation doesn’t exist yet. Every one after it should benefit from what came before. That’s what building context as infrastructure produces: an AI program that gets faster and more precise over time, not one that starts from scratch every time.

The goal isn’t a collection of successful initiatives. It’s a company that runs differently because of AI. Context is what gets you there.

Talk to Optura about building the organizational context layer your AI program needs to compound.