Why Healthcare's AI Moment Is Different — And Why Most Vendors Will Get It Wrong

Healthcare has never been short on technology promises. For the better part of three decades, every major wave of innovation arrived with the same claim: this will finally fix the fragmentation. EHRs would unify clinical and financial data. Patient portals would close the communication gap. Revenue cycle management platforms would eliminate the manual work. Telehealth would expand access. In every case, the technology delivered something — just not the transformation it promised. In nearly every case, it was deployed in a silo.

I've spent the better part of 25 years in and around healthcare revenue cycle, consulting on implementations and solving the kinds of problems that don't show up in a vendor's demo but surface six months after go-live. What I've watched, repeatedly, is an industry that adopts technology function by function, vendor by vendor, and then spends years trying to stitch those decisions into something coherent. The stitching never quite works. Now, at what should be the most consequential technology moment in healthcare history, I'm watching the industry prepare to make the same mistake again — this time with AI.

The Silo Problem Isn't New. It's Just Expensive.

Healthcare organizations didn't set out to build fragmented technology ecosystems. They built them one urgent problem at a time. A coding accuracy issue led to a coding tool. A denial rate spike led to a denial management platform. Patient collections lagging brought in a patient engagement vendor. Each decision was defensible in isolation. Each added a contract, an integration, a training program, and a support relationship.

By the time most health systems stepped back to look at the full picture, they were managing dozens of point solutions across the revenue cycle — each with its own data model, its own workflow, and its own definition of success. The EHR was positioned as a unifying platform but never fully delivered on that promise. What it did do was force a conversation about interoperability that the industry has been slowly, painfully winning.

Over the last decade, real progress has been made. Revenue cycle leaders have rationalized vendor footprints, consolidated platforms, and pushed for end-to-end workflow design rather than function-by-function fixes. Mid-cycle and front-end tools began talking to back-end systems. Patient access workflows started connecting to billing outcomes. Organizations began measuring revenue cycle performance as a continuum rather than a collection of discrete metrics. It has been hard, expensive work but it has moved the needle.

That progress is now at risk.

AI is Being Deployed the Same Way We Deployed Everything Else

Walk into any major healthcare conference today and the AI conversation is everywhere. Prior authorization automation. Clinical documentation assistance. Denial prediction. Payment integrity. Charge capture optimization. Patient communication. Each use case is real. Each has a vendor behind it. Each is being evaluated, piloted, and purchased the way healthcare has always bought technology — one function at a time.

The result is predictable to anyone who has lived through the last three decades of healthcare IT: a new generation of point solutions, this time powered by AI, layered on top of the integrated workflows that revenue cycle teams spent years building. Instead of eliminating vendor complexity, we are expanding it. Instead of designing intelligence into the end-to-end workflow, we are bolting AI agents onto individual functions and calling it transformation.

This matters more with AI than it did with prior technology waves, for a specific reason. The power of AI is not in what it can do in isolation. The power of AI is in what it can learn, adapt, and orchestrate across connected workflows. An AI agent that automates prior authorization but has no visibility into the downstream denial patterns it creates is not intelligent, it is automated. An AI tool that optimizes charge protocols but operates independent of the payer contract terms driving reimbursement is only solving half a problem. The return on AI investment in healthcare will not come from deploying smarter point solutions, it will come from deploying AI that understands the revenue cycle as a system.

Right now, most healthcare organizations are not doing that. Most vendors are not asking them to.

The Vendor Incentive Problem

It would be unfair to place all of the blame on health systems. The vendor ecosystem is not incentivized to sell end-to-end thinking. Vendors are incentivized to sell their product. A denial management company wants to own denial management. A prior authorization automation company wants to own prior authorization. The idea that their AI agent should be designed to operate within a broader intelligence architecture — one that might include a competitor's product — is not a conversation most vendors are eager to have.

This is the same dynamic that created the fragmented RCM technology landscape of the 2000s and 2010s. The difference is that AI amplifies the consequences. When disconnected AI agents are making autonomous decisions across the revenue cycle without shared context, the errors compound faster, the blind spots are harder to detect, and the cost of misalignment grows with every transaction.

Healthcare organizations that are evaluating AI vendors today need to be asking a question that most vendors are not prepared to answer: how does your AI agent communicate with the rest of my revenue cycle workflow? If the answer is a vague reference to APIs and interoperability, that is not an answer. That is a placeholder.

What Nirvana Actually Looks Like

The healthcare organizations that will extract real, durable value from AI are the ones that resist the pressure to deploy function by function. They will demand a verticalized AI architecture where intelligent agents are designed from the start to work together across the revenue cycle. Those agents share context, learn from shared outcomes, and escalate decisions that cross functional boundaries.

This does not mean waiting until a perfect end-to-end AI platform exists before acting. It means that every AI investment made today should be evaluated against a simple question: does this agent make the rest of our AI strategy smarter, or does it operate in isolation? If the answer is the latter, the long-term cost of that decision will eventually outweigh the short-term gain.

The organizations building AI strategies this way are not doing it because a vendor told them to. They are doing it because someone in the room has been around long enough to recognize the pattern. They see the silo forming before it hardens. They demand something different while there is still time to design it right.

That perspective comes from those who have consulted and solved these problems across the revenue cycle in hundreds of healthcare organizations, across multiple segments of the industry. It comes from seeing what fragmentation costs. Not in theory. In write-offs, in denials, in audit findings, in the quiet erosion of margin that no dashboard captures cleanly.

Healthcare deserves AI built by people who have lived that. Healthcare organizations deserve to demand it.