The Hard Part Was Already Built

When I joined, the hardest part of the platform was already built. It turned out not to be the hard part.

In 2015, an industry forecast projected that 105 million Americans would be covered by Accountable Care Organizations by 2020.¹ By mid-2019 the real number was about 44 million.² That’s real growth, but less than half of what the industry was betting on. I was one of the people responsible for making that bet pay off.

The bet

In 2014, healthcare went all-in on “population health.” The Affordable Care Act gave rise to Accountable Care Organizations (ACOs), and the logic was simple: find the people who are about to get sick, intervene early, and keep them healthy. Prevention is cheaper than treatment. Those savings were supposed to let health systems stop getting paid per visit and start taking financial responsibility for the health of a whole population.

Every major health IT vendor had a population health platform. I was brought in to deliver one.

My seat

I ran customer success for one of the largest private health information exchanges in the country: a data network connecting more than 300 care settings across a major US health system. When our platform won their population health business, my job was to build the strategy to implement and operationalize it, sell that plan, and then deliver it. In other words: turn what had been promised into the ROI it was sold on. You could argue we won because the data connections were already in place. I can’t say for sure.

Three pillars

The platform rested on three pillars: data, registries, and interventions.

Registries were the differentiator: the logic that scans a population and flags who needs attention. They were built before I arrived, on years of research by leading academics and clinicians. Everyone believed that was the hard part, and it was done.

Where it broke

Two problems showed up. Ten years later, they still do.

First, the data. Registries need clean, complete, consistent inputs. Healthcare data in 2014 was none of those things. Systems didn’t talk to each other, the same patient looked different in every source, and fixing it was bigger and more expensive than anyone had budgeted. It was solvable, but costly.

Second, the intervention. Even with a perfect list, someone has to get that insight to the right clinician, at the right moment, inside a workflow they already use. Then that clinician has to reach a patient who has to change their behavior. Every handoff is friction, and every handoff is a person. That turned out to be the real product, and it’s the part software couldn’t deliver on its own.

AI can widen the funnel. Getting someone to act is still change management.

Ten years later

Here’s my provocation: data access and normalization is solved. Not today, but I’d bet within 5–10 years. AI is compressing years of interface and cleanup work into something much closer to months.

The intervention problem is different. I don’t believe AI closes that loop. Better lists don’t create trust, time, or a relationship between a care team and a patient who, I believe, wants to be healthier. AI can widen the funnel. Getting someone to act is still change management. I’d love to be proven wrong.

Same play, different industry

I work in permitting and development now, and I see the same three pillars. The data is application, parcel, and plan data. The registry is the code logic that decides what a project needs. The intervention is an applicant, reviewer, or inspector doing the next right thing at the right moment.

The way I frame it: technology works when the right data reaches the right hands, in the right place, at the right time. When modernization stalls, it’s rarely because the tool can’t do the job. It’s because the workflow never changed around it.

Where have you seen the last mile, not the technology, decide whether a modernization worked?