Hospitals: AI in Revenue Cycle Under Scrutiny

Revenue cycle leaders explain how they frame the use of AI agents between business rules and human responsibility. On the ground, teams select tasks to automate and support the changes. On the supplier side, agents are already conducting investigations into denials and making calls to insurers using voice AI.

Outdated Rules and the Need for Clear Accountability

At Panoramic Health, Sap Sinha notes that billing, documentation, and communication rules with insurers are not evolving as quickly as AI technologies. He believes that the evolution of this regulatory framework could determine how providers and payers will use AI agents to process reimbursement requests. At Mount Sinai, Salonia Brown emphasizes the need to define precise workflows and clearly identify responsible parties when automation encounters an exception. She argues that without these foundations, the introduction of AI risks exacerbating existing challenges. Brown summarizes: without clear processes and a designated responsible party, AI merely digitizes dysfunction. She adds that establishing governance and an appropriate business model takes time. In this context, teams must also know who intervenes in the case of an exception.

What Teams Want: Targeting Useful Automation

Clinicians and operational teams want AI tools to alleviate administrative burdens, according to Sap Sinha. At Rush University System for Health, Blake Evans explains that frontline employees identify time-consuming, low-value steps, which has facilitated their buy-in for other changes. At Adonis, David Unger distinguishes between repetitive tasks suitable for automation and complex cases that require better information and human decisions. For a high-stakes call, AI can help determine whether to pursue it and what evidence to gather. At Mount Sinai, Salonia Brown's team determines which tasks can be automated and which require critical judgment, aiming to free staff from recurring tasks to focus resources on complex cases.

Concrete Deployments: AI Agents, Voice Calls, and Early Detection

According to Akash Magoon, Adonis is developing agents capable of reviewing denied reimbursement requests and reducing the amount of manual work dedicated to payment recovery. The company also uses voice AI to contact insurers and gather information on cases, resulting in tens of thousands of weekly calls for its partners. Doug Pickett indicates that Adonis's roadmap includes identifying revenue cycle incidents as they occur and proposing a solution, whether it involves initiating automation or making corrections upstream, such as on coding or provider qualification. In hospitals, the revenue cycle still relies on teams tasked with analyzing denials, monitoring payer-imposed rules, and determining situations that require human intervention. Some of these tasks could be handled by AI agents, provided there is closely managed integration. Discussions between leaders and suppliers thus focus on the division between human and machine, governance of exceptions, and the scope where human judgment remains essential.