Stop automating the revenue cycle. Start redesigning how the work gets done.

Health systems have spent years trying to make the revenue cycle more efficient. We digitized workflows, added work queues, automated repetitive tasks and, more recently, applied artificial intelligence to coding, clinical documentation integrity (CDI), denials and other functions.

Those advances matter. But before making the next technology investment, healthcare leaders should ask a harder question:

For decades, the operating model has remained largely unchanged. Coders review charts. CDI specialists look for documentation gaps. Auditors identify risk. Revenue integrity teams search for leakage. Denials teams investigate what went wrong after claims come back.

Technology has made each function more productive, but people still find the problems and technology helps them resolve those problems. AI gives us an opportunity to reverse that model.

Hospital executives are trying to control costs, protect reimbursement, navigate workforce constraints and manage growing complexity. More automation will help, but automation alone is not the answer. 

If AI is deployed separately across coding, CDI, revenue integrity, utilization management and denials, health systems may automate more tasks without addressing the fragmentation between them. The result is faster silos, and faster silos are still silos. 

The better question is not, how can we automate more of the work people do today? It is, which work should require a person in the first place?

In a traditional workflow, different teams evaluate different aspects of an encounter at different points in time. Problems are identified, work is routed, people investigate and errors are corrected, sometimes while there is still time to influence the encounter and sometimes much later. 

Now imagine reversing that process. Instead of waiting for a person to initiate every review, technology continuously evaluates the encounter. It applies the appropriate intelligence to the information available, advances high-confidence work and identifies potential issues as they emerge. When expertise is needed, the exception reaches the right person with the context required to make a decision.

The goal is not to remove expertise from the revenue cycle. It is to stop wasting expertise on work that does not require it. 

Health systems are organized around functions for good reason. Coders, CDI specialists, auditors, revenue integrity professionals and utilization management teams each bring distinct expertise. But technology does not have to reproduce those boundaries.

An encounter does not know that coding and CDI use different work queues. A documentation gap can affect coding. Coding can affect reimbursement. Medical necessity can influence denial risk. Quality, compliance and financial outcomes may all connect to the same clinical information.

Yet healthcare technology often treats them as separate problems. Designing intelligence around the entire visit — not just an isolated task — creates the opportunity to identify risk earlier, coordinate action and reduce downstream rework.

AI is often measured by how quickly it performs a task. That is useful, but it may be the wrong level of measurement.

Consider a highly trained employee with 100 possible cases to review. Making that person 20 percent faster is valuable. But what if intelligence could determine that only 15 cases require attention?

That is a fundamentally different productivity model. Technology completes what it can confidently perform and routes the exceptions that require expertise. The next major efficiency opportunity may not come from helping people process queues faster, but from changing what reaches the queue at all.

People do not disappear from the revenue cycle. Their expertise becomes more valuable because it is concentrated on ambiguity, complexity, quality, compliance and judgment.

If the objective is a more autonomous revenue cycle, leaders should ask whether their technology strategy is designed to get them there or whether they are adding AI to the same roles, applications, queues and handoffs they have had for years.

Incremental automation has value, and no organization can transform everything overnight. But leaders should be clear about the destination.

It will continuously evaluate encounters rather than depend on people to initiate every review. It will identify risk before defects become downstream problems. It will automate appropriate work while preserving human judgment where it matters. And it will organize work around the integrity of the patient visit rather than the limitations of individual systems.

That is more than a technology migration. It is an operating model change.

AI is advancing quickly, but healthcare does not need to automate simply because it can. Executives have a more important job: deciding what the revenue cycle should look like when intelligence and automation are built into its foundation.

How can I ensure my staff is operating at the top of their license and not wasting time on administrative tasks?  How can I provide career pathways and upskilling opportunities for my staff in the new world of AI? How should information move across coding, CDI, quality, revenue integrity and other functions? And what governance is needed to support trust, transparency and accountability?

The destination is not autonomy for autonomy’s sake. It is a revenue cycle in which people and technology each do the work they are best suited to do.

So perhaps the question is not, how much of your revenue cycle can you automate?

Ask instead:

If you could design the revenue cycle today, knowing what AI can make possible, would you design it the way it currently works?

If the answer is no, do not just automate the old model.

Build the next one.

Jason Burke is vice president of revenue cycle at Solventum.