Who Is Accountable When the EHR Becomes a Decision Engine?

Photo By: Usman Yousaf

The electronic health record, or EHR, was originally created to make it easier to document patient care. Doctors and other healthcare professionals could use it to record visits, store test results and share information.

That role is changing quickly.

Artificial intelligence is now being built into many parts of healthcare. AI can help write clinical notes, suggest medical codes, review claims, support prior authorization and manage other parts of the revenue cycle. The EHR is no longer simply recording what happened. It can also influence what happens next.

That raises an important question: Are today’s healthcare policies and oversight systems prepared for an EHR that is increasingly helping make decisions?

The answer is not straightforward.

AI tools are already being used to support medical coding and other administrative tasks. These tools can save time and reduce some of the workload for healthcare organizations. They also introduce new risks.

An AI system might review a doctor’s documentation and suggest a billing code. That may sound like a routine administrative task, but the code can affect how much a healthcare organization is paid. It can also influence claims data and other financial information.

Jim Maguire, co-founder of Government Market Strategies, points to the potential benefits on both sides of the healthcare system. “AI is enabling providers to generate cleaner claims faster, reduce administrative burden, and accelerate reimbursement, while giving payers powerful new tools to audit and validate those claims at scale. If oversight frameworks ensure that claim edit rules and decision logic are applied consistently and transparently, AI could become one of the most important levers for reducing administrative costs and improving efficiency across the entire healthcare industry.”

The Centers for Medicare & Medicaid Services is already examining these issues. CMS has asked for input on the use of AI in diagnosis and coding, including how AI recommendations should be presented to human reviewers and what compliance risks need to be considered.

AI is also being used by health insurers and government programs to review claims and prior authorization requests.

CMS’s WISeR model, which began in 2026, uses AI and machine learning along with human clinical review to examine certain Medicare services. The goal is to identify care that may be inappropriate or wasteful.

These developments show why healthcare AI is becoming more complicated from a regulatory perspective. AI is no longer limited to tools that support doctors and nurses. It is increasingly involved in financial and administrative decisions that can affect providers and patients.

Current regulations were not necessarily designed for this environment.

The Food and Drug Administration regulates certain types of medical software, but not every AI tool used in healthcare falls under FDA medical device rules. Some clinical decision-support software may be outside those requirements depending on what the software does and how it is used.

HIPAA presents another example. HIPAA provides important protections for patient health information and requires organizations to maintain privacy and security. Privacy is only one part of the AI question, however. HIPAA does not by itself answer whether an AI system made an accurate coding recommendation or whether an automated claims decision was fair.

The issue becomes even more important when different AI systems are connected.

Imagine an AI tool creates a clinical note. Information from that note is added to the EHR. Another AI system then uses the information to suggest a diagnosis or billing code. That information could eventually be used in a claim or another financial decision.

Each individual step might seem reasonable. The problem begins when one system makes a mistake and that mistake moves through the rest of the process.

That creates a difficult question about responsibility. Who is accountable when AI influences a healthcare decision? Is it the software company, the healthcare organization, the clinician, the billing team, the insurer, or some combination of them?

Transparency is another concern.

Should patients be told when AI played an important role in a decision about their care or coverage? Should providers be required to keep records showing how an AI system influenced a claim? How much human review is enough when a person is simply approving an AI recommendation?

These questions suggest that healthcare may need to rethink how it oversees AI.

The focus may need to move beyond individual software products and toward the entire decision-making process. Regulators and healthcare organizations may need to understand how information moves from the patient record to an AI system, from the AI system to a recommendation, and finally from that recommendation to a clinical or financial decision.

The EHR was originally designed to document healthcare.

AI is turning it into something more. It is becoming part of the infrastructure that can influence clinical, administrative and financial decisions.

The challenge for policymakers is making sure the rules keep pace with that change.

The question may no longer be whether AI belongs in healthcare. AI is already there.

The bigger question is this: When an AI-enabled EHR influences a healthcare decision, who is responsible when it gets the decision wrong?