Written by Andrew Crook, Senior Advisor, and John Hillery, Director of New Product Development, Tally, Inc.
It is becoming commonplace to talk about AI in breath-taking terms: revolution, transformation, labor market disruption, once-in-a-generation change.
Considering the revenue cycle, this may sound more like a warning than a promise. Financial leaders are right not to jump in blindly. There are unanswered questions about compliance and information security risks, concerns that the audit trail will become a “black box,” and how serious the impact of “illusions” on quality and collection processes will be.
Some leaders believe that seeing others act first helps improve governance and processes. However, waiting and watching has its consequences. UnitedHealth Group plans to invest $1.5 billion in AI.
We’ve looked at some practical ways that AI automation tools can be safely and productively applied in provider organizations. These basic principles will help move AI from difficult to doable.
- Don’t jump to layoffs. Enhance your existing people and processes with automation tools for instant success. This is a tandem bicycle concept, especially suitable for teams of “earning cyclists”. Long-term issues around staffing, roles and responsibilities, and the shape of the finance department require more careful planning and should not be rushed. At a medical device company, staff generated hundreds of reports on payer policy and rate changes, but were unable to consistently address or cover most of the payers. Now, instead of creating systems, they review, approve, and update systems, act on financial plans, and make projections that they previously didn’t have time for.
- Do your homework on the risks and the best starting point. The most advanced leaders consider the stages of the revenue cycle and diagnose where opportunities for improvement exist while mitigating irreversible risks.
At each step, they ask: How recoverable is a mistake before it causes permanent revenue loss? AI effectiveness is a moving target, so score it using current benchmarks or vendor data, as you did with your scorecard. Start with what you are ready for. This is an area where there is a huge opportunity for improvement and where AI can be effective, and the downside if something goes wrong is manageable.
- Rather than just expecting AI to “learn as it goes,” think carefully about providing data and expertise to your tools. There is a concept in the market that AI simply learns and “understands” and improves itself over time. While agents are powerful and models are constantly being improved, the reality is that giving the AI the right data context is more important than the AI itself. This is good news for RCM leaders. You have the data and the experience. You just need to take advantage of it. Billing history, past results with payers, tips and tricks on people’s minds, etc. all need to be systematically fed into the data sources used by AI.
- Deeply explore the meaning of “a person who gets caught up in something.” It is important to note that AI can be used to varying degrees. You don’t have to choose between 100% human manual and 100% AI. On the one hand, AI agents could become just a “power tool” for RCM staff to use as a productivity accelerator. On the other end of the spectrum, revenue cycle tasks can be built to run autonomously, and RCM can be linked into the chain to run end-to-end, achieving what is known as “touchless” RCM. Most organizations don’t move from manual processes to AI overnight. At Destination Smiles, one of the dental providers we work with, AI performs several RCM steps and provides tips for AR follow-up to staff. The CEO emphasizes speed, but “[otherwise]you can’t tell if it’s an AI or a human, because it’s both and the process is pretty much the same.”
“Human involvement” appears in almost every RCM AI pitch. This may mean a check-based approach, where humans review the AI’s output and flag errors. A production-based loop means that human RCM staff performs key revenue cycle tasks while AI supports and learns from their actions. Semantic differences impact compliance and risk of error.
- Limit changes to legacy systems. Many organizations are concerned about the cost, time, and risk of overhauling traditional PMS and EHR systems. The good news is that you don’t have to disrupt your systems of record to leverage the processing power of AI for your revenue cycle. Most products have APIs for access and writeback, and you can add human review to ensure changes to your system of record are accurate.
