Re-Engineering Revenue Cycle for the AI Era

‍

Re-engineering revenue cycle for the AI era.

Artificial intelligence is moving into coding, clinical documentation, authorization, analytics, and other parts of the revenue cycle. But in my work with healthcare leaders, I continue to see the same tension: technology can accelerate a process without correcting the design beneath it.

That tension framed “Back to First Principles: Re-Engineering the Revenue Cycle for the AI Era,” a Vee Healthtek webinar presented in partnership with Becker’s Healthcare. I had the opportunity to moderate a conversation with Dawn Castro of Surgery Partners, Shannan Bolton of Stanford Health Care, and Peter Thompson of UF Health. Together, we explored what revenue cycle leaders must reconsider before scaling AI.

My clearest takeaway was this: AI value begins with redesigning the work. Leaders need to understand why a process exists, where defects originate, how users experience the workflow, and which outcomes the technology is expected to change.

Start with the work, not the technology. | Start With the Work

Revenue cycle improvement often begins with a familiar question: how can we make the current process faster? Shannan challenged that starting point. Before optimizing or automating a workflow, leaders should ask whether the work should exist in its current form and whether the process was designed to prevent problems or manage them after they occur.

“Our first question is not how the work is done today, but why it should exist at all.”  
Shannan Bolton, Stanford Health Care

Denial management was a useful example. A workflow designed primarily to recover dollars after a claim fails may become more efficient without reducing the source of the denial. Shannan also noted that a result that remains consistent over time can point to the design, rather than an isolated execution issue. This shifts the conversation from who made an error to what the process repeatedly produces.

Fit AI to operational reality. | Fit AI to Operations

Dawn’s experience reinforced why a promising capability can still miss its expected value. An authorization use case may look attractive at an enterprise level, but performance can vary across service lines, payers, states, systems and user groups.

“What can you actually do today? Let’s focus on those service lines instead of the big, pie-in-the-sky opportunities.”
Dawn Castro, Surgery Partners

Her team’s work showed the importance of observing how people use technology in practice. Seat rides exposed integration gaps, repeated work and the trust issues that develop when users have experienced unreliable tools. The lesson for me is that AI adoption is not a communication task added after implementation. It must be designed into the workflow from the beginning.

Peter shared a related example from clinical documentation integrity. When an initial implementation did not sufficiently reflect clinicians’ day-to-day experience, the organization stepped back and involved clinician work groups in shaping the reimplementation. The technology had to fit the people and the work, not the other way around.

Move prevention upstream. | Move Prevention Upstream

The panel also returned to the need to move attention earlier in the revenue cycle. Recovery work will remain necessary, but leaders should not judge performance only by how effectively teams resolve downstream defects.

“There is a lot that we can do to prevent things from occurring in the first place.” 
Peter Thompson, UF Health

Peter described denial-prevention work groups that examine root causes and look for earlier intervention points. This idea connected with the measures discussed across the panel: revenue leakage, first-pass payment, yield, avoidable denials, speed to payment, and the percentage of accounts requiring human intervention. Each offers a different view, but all help leaders see whether redesign is reducing avoidable work or simply moving it faster.

A first-principles sequence for revenue cycle AI.

‍

Redesign roles and oversight. | Redesign Roles and Oversight

AI also changes the work people perform. Peter described a shift from completing tasks to auditing AI-generated output. That shift requires skills in exception management, quality review, and identifying when technology is not performing as expected.

“With AI, we’re moving people from being doers to being auditors.”
Peter Thompson, UF Health

Shannan discussed roles focused on automation leadership, workflow review, and post-launch monitoring. Dawn extended the point to product ownership, RFP discipline, and technology-vendor management. These are not supporting activities around AI. They are part of the operating model required to govern it with open accountability.

Retire low-value work. | Retire Low-Value Work

Re-engineering is also about deciding what to stop. Shannan described efforts to remove duplicate work queues, manual reconciliations, low-value reviews, and overlapping tools. The objective is not consolidation for its own sake. It is to create a less fragmented environment in which teams can act on insights instead of maintaining unnecessary activity.

What I would ask revenue cycle leaders to consider. | Leadership Takeaways

• Does the work still need to exist in its current form?
• Is the AI use case aligned with the users, service lines, payers and systems it will affect?
• Are we preventing avoidable defects or becoming faster at recovering from them?
• Who owns adoption, auditing and performance after the technology goes live?
• Which outcomes will show that the operating model has improved?

The organizations best positioned to scale AI will be those willing to question the work before automating it, involve the people closest to the process, retire unnecessary activity and make performance visible. That is the work behind an AI-ready revenue cycle.