What You Should Know:
– For decades, the “revenue cycle“—the complex machinery of medical billing, coding, and reimbursement—has been the unglamorous back office of healthcare. But in 2025, it has become the frontline of AI adoption.
– According to a new survey released by the Healthcare Financial Management Association (HFMA) and AKASA, the adoption of Generative AI has moved past the experimental phase into operational necessity. With nearly 9% of hospital revenue now evaporating due to documentation errors, health systems are turning to Large Language Models (LLMs) to decipher the complex reality of patient care into the rigid language of billing codes.
The Surge: From Novelty to Necessity
The speed of adoption is striking. The survey indicates that 80% of health systems are currently taking action on GenAI tools for revenue cycle management—whether exploring, piloting, or fully implementing. This represents a 38% increase in less than two years.
This isn’t merely about automating administrative tasks; it is about survival. As patient care becomes more complex, the administrative burden of documenting that care has outpaced human capacity. “The findings suggest that GenAI could become a crucial lever to both capture the quality of care delivered and improve revenue integrity,” notes Malinka Walaliyadde, CEO and co-founder of AKASA.
The “Digital Divide” in Healthcare
While the headline numbers are bullish, a closer look at the data reveals a worrying bifurcation in the market. Operationalizing AI requires capital, and smaller organizations are struggling to keep up.
- Large Systems: Among larger health systems, 64% are actively piloting or implementing GenAI solutions.
- Small Systems: For organizations with revenues between $500 million and $1 billion, only 20% have reached the pilot or implementation stage.
Despite recognizing the value, these smaller entities remain stuck in the early adoption stages, hamstrung by budget constraints and the sheer difficulty of scaling new technology. This divide creates a risk where larger systems become more efficient and financially robust, while smaller community systems continue to bleed revenue through inefficiencies.
The High Cost of Bad Data
Why is the industry rushing toward GenAI? Because the status quo is prohibitively expensive. The survey quantified the cost of “incomplete or inaccurate documentation” at 8.49% of total revenue. For a multi-billion dollar health system, that is hundreds of millions of dollars left on the table.
The pain points are specific and acute:
- 89% of organizations say inaccurate codes significantly impact revenue.
- 60% believe GenAI’s biggest opportunity is identifying missed reimbursement.
- 57% see it as the key to uncovering gaps in clinical documentation.
Jackie Josing, Vice President at LCMC Health, emphasized that success requires more than just buying software. “Success with AI depends on partnerships that bring people and technology together with purpose,” Josing said.
Barriers: It’s Not Just About the Tech
Despite the optimism, implementation is not frictionless. The survey found that integration with existing systems remains a top hurdle. Healthcare IT environments are notoriously fragmented, often consisting of a patchwork of legacy electronic health records (EHRs) and billing systems.
For large health systems specifically, cost and budget constraints were cited as the single biggest challenge (52.5%). This underscores a critical tension: systems need AI to save money, but they need money to deploy AI.
The Future of CDI: Preventive and Collaborative
Looking forward, the role of Clinical Documentation Integrity (CDI)—the teams responsible for ensuring medical records accurately reflect the care provided—is poised for transformation.
Over the next five years, 72% of respondents expect CDI to evolve from a reactive, back-end function into a “preventive, collaborative” force. The vision is for AI to handle the retrospective review, freeing up human experts to focus on denial prevention, appeals, and prospective reviews before claims are even submitted.
