Volume 4: An Exclusive Feature on Why AI Needs a New Operating Model

Healthcare's real crisis isn't AI readiness; it's the broken legacy architecture beneath it. Why true transformation means rebuilding for the patient first.

Healthcare's real crisis isn't AI readiness; it's the broken legacy architecture beneath it. Why true transformation means rebuilding for the patient first.
May 20, 2026
Paul Sanford
Advisory Board Member, Firstsource
Download now
Healthcare's real crisis isn't AI readiness; it's the broken legacy architecture beneath it. Why true transformation means rebuilding for the patient first.

My parents are 81 years old and living on a fixed income. Like millions of Americans their age, they rely on health insurance and Social Security to keep going. This year, the Medicare Advantage plan was discontinued in their state. What followed was not a medical crisis; it was paperwork, phone calls, and anxiety that showed up every single day.

They had to select a new insurer, a new primary care physician, and a new drug provider, hoping their specialists would still be covered, recalculating deductibles and out-of-pocket costs, with no one to guide them through any of it.

I have spent decades in this industry. I have been in the rooms where these systems were designed, in the committees where they were approved, and in the integrations that were supposed to simplify them. And watching my parents navigate what we built, I kept arriving at the same thought. We never actually designed this for them. That is the problem. Not the technology. Everything that came before it.

Every phase created value. None of it solved the problem.

I came up through every phase of this industry's evolution. In the 1990s, managed care promised to rein in costs. In the early 2000s, I was on the other side of the table, moving work offshore to capture the labor differential. In the 2010s, we built shared services, consolidating functions to squeeze out redundancy. I did not just observe these phases from the outside. I designed and implemented them, and I believed in each one at the time.

Hospital margins sat at a median of 1.1% in 2025,¹ and McKinsey projects margin pressure of up to 13% over the next five years.² The organizations facing this are not facing just a bad year. It is a structural break, and most organizations are responding to it as they have responded to everything for the past two decades.

Here is what I know from having run these plays myself. Every one of those levers relocates cost elsewhere rather than reducing it. Labor arbitrage is priced into contracts before the conversation starts. Shared services consolidation works until the complexity grows back around it, and it always does.

The real cost driver is the internal architecture: fragmented business units running parallel processes, carrying redundant technology, negotiating separately with the same vendors. It accumulates as a complexity tax that never shows up cleanly in a budget line, just always there in the margins, in the delays, and in the decisions that never get made.

When I oversaw the shared services operation at Cigna, spanning health plan, pharmacy, and pharmacy benefit management (PBM) across 34,000 people, organizational complexity was as significant a cost driver as anything in the external market. The $60 billion Express Scripts acquisition doubled the scale of that challenge overnight. What I learned is that scale does not solve complexity. It exposes it.

AI on a broken process gives you a faster broken process

Healthcare has adopted AI at 2.2 times the rate of the broader economy,³ and the pilot era is over. But when most organizations encountered this capability, they did what they always do. They added it to what already existed.

There is a critical difference between deploying AI and redesigning for AI, and most health systems are doing the former, layering intelligent tools onto legacy workflows built around the assumption that a person reviews, a person routes, and a person approves. That assumption is still intact underneath all the new technology. AI is making those processes faster. It is not making them right.

Most organizations ask how to improve existing processes by 20%. That is the wrong question. The right one is whether, if you were building this today with AI available from day one, you would build anything like this at all. Ask that honestly about almost any established healthcare workflow, and the answer is uncomfortable.

The barrier is not technology. It is governance. Large healthcare organizations are not monoliths but collections of business units, each with its own P&L, vendor relationships, and a sense of who owns what. Redesigning a process end-to-end when ownership is split four ways is close to impossible, which is why AI initiatives stay in pilot mode well past the point of proven value. Nobody has the mandate to rebuild end-to-end. So, the AI gets bolted on, the old structure stays intact, and the problem keeps compounding.

The experience was never designed. It was inherited.

Healthcare is the only industry where someone else pays your bill, and you find out what's covered when you're already in the system. Employer-sponsored insurance put distance between the consumer and the cost. Fee-for-service rewarded volume over outcomes. Prior authorization grew a utilization tool with no investment in making it navigable for the person on the other end. None of these decisions were wrong in isolation. Together, they built a system where the patient has the least information and the least power.

One in three adults in the US skipped or postponed care in the past 12 months because of cost.⁴ Prior authorization, in theory, manages utilization, but in practice it often manages it by creating enough friction that people give up. That experience is invisible to the organizations running the process and very visible to the person trying to get care.

The system was designed, but not for the patient. It was built around payers, providers, and employers, and the patient was assumed to fit around it. Until organizations treat the experience crisis and the cost crisis as the same problem, they will keep solving for the wrong thing.

Technology can't save a system that was never designed for the patient

I am not pessimistic about what AI can do in healthcare. I have seen too much genuine progress to be dismissive. But the organizations that come out ahead in this decade will not be the ones that deployed AI fastest or cut costs hardest. They will be the ones that used this moment to ask whether the model they have been running is worth keeping, and then rebuilt from the answer.

Someone, be it a CFO, a COO, or a board, must decide that fragmentation ends. One team owns the outcome, not just the process. The measure of success is whether the patient stayed in care, whether the claim got resolved, whether the person on the other end got what they needed. Not whether the SLA was met.

It also means being honest about what rebuilding requires. The teams being asked to redesign their workflows are not resisting because they are complacent. They are resisting because nobody has acknowledged what they are being asked to let go of.

The answer is not technical; it's about leadership. The will and clarity to reject the status quo, and then realign everything around the future: strategy, rewards, incentives, organizational design, and the processes that govern how work gets done. The STAR model taught me that if any part of that system is out of alignment, transformation does not slow down; it stops. The organizations that moved were the ones where leadership showed up having already changed something themselves.

The demographic math makes this urgent. More than 61 million Americans are already 65 or older, and by 2030, that number will hit 71.6 million. That makes it one in five.⁵ The most complex to serve, the least equipped to navigate what we built: all arriving at once. My parents are not an edge case. They are the beginning of the wave.

A note to close

I think about my parents when I sit in rooms where executives talk about transformation. The presentations are polished, the technology is real, and the pilots are promising. But I keep asking the same question: Would the person at the center of this system recognize it as something designed for them? In most cases, the answer is still no.

The organizations that get this right will not just be more efficient, but will have finally built something worth using. And for the first time, the person at the center will know it.

References:

Download UnBPOTM Quarterly