News

Responsible AI in Release of Information

A Framework for Health Information Leaders

By Angela Rose, MHA, RHIA, CHPS, FAHIMA

AI has quickly moved from an emerging technology to an operational reality across health care. It is reshaping everyday work for HIM teams, and its evolution has only just begun.

Health information (HI) professionals have already experienced rapid AI advancements in departmental operations, particularly across clinical documentation integrity, medical coding, patient access, and revenue cycle workflows. Early successes with robotic process automation and machine learning provide a foundation for what comes next: increasingly intelligent automation that supports the expertise of HI professionals, including release of information (ROI) teams.

In this quickly advancing era of AI technology, HI leaders need to understand where AI creates measurable value in the ROI process and how responsible governance delivers long-term success. Top areas for improvement in disclosure management include faster turnaround times, workflow automation for requests, EHR integration, and digital delivery of requested information. Furthermore, effective implementation and optimization of AI will help HI leaders navigate workforce shortages, meet compliance demands, and improve patient experiences today and for years to come.

Why AI Matters More Than Ever
According to recent AHIMA workforce research, nearly two-thirds of organizations reported understaffing in HI positions over the past two years. At the same time, record volumes continue to increase, regulatory requirements evolve, and patients expect faster, more seamless access to their health information.

Protected health information (PHI) disclosure management has never been more complicated. Simply asking staff to do more with fewer resources is no longer sustainable.

AI offers one potential solution to optimize staff efficiency and effectiveness, but only when applied responsibly. A large-scale study from Stanford University examined where workers want AI assistance vs its actual capabilities today. The study offers an important perspective for HI leaders: Employees generally prefer AI to automate repetitive, structured, and administrative work.

The same pattern is evident within ROI. Today’s AI technologies are well-suited to support record release activities, including:

• intake and classification of incoming requests;
• extraction of information from request forms;
• automated medical record retrieval;
• document organization and workflow routing; and
• routine quality checks.

By contrast, ROI activities requiring expert judgment—such as the interpretation of complex legal requests, regulatory compliance analysis, or compassionate interactions with patients—remain firmly in the hands of experienced professionals. This distinction is vital because successful AI adoption enables people to spend less time on repetitive administrative work and more time applying their specialized human expertise.

Within ROI specifically, AI-powered record retrieval has demonstrated meaningful operational improvements. Organizations implementing automation have experienced significantly faster turnaround times while allowing ROI specialists to focus on complex requests, exception handling, compliance review, and customer service.

Where Automation Ends and Human Expertise Begins
Consider a busy health system receiving an electronic request for medical records from an attorney representing a minor involved in a personal injury case. The AI intake engine ingests the request, performs optical character recognition on the authorization form, extracts key identifiers, verifies identity, and searches the EHR system. It locates the requested encounter range and organizes the matching clinical documents.

However, the automated check flags two nuances: The patient turned 18 during the course of treatment, and the records contain state-protected behavioral health and substance use disorder notes. At this exact juncture, the technology hands off the file to an ROI specialist.

Before any PHI is released, the ROI specialist reviews the file and catches something the AI never could: This attorney also represents the child's other parent. The other parent lost custody after a domestic violence finding and currently has a restraining order in place. None of that lives in the medical record. It surfaces through a phone call, a caseworker's note, or something staff simply know from handling the family's case before.

The specialist halts the release and checks with legal before anything goes out because disclosing these records, even to a legitimate-looking request, could put the child at real risk. The form itself was flawless. The signature was valid, dates correct, and identity confirmed.

In this scenario, AI knew the document was authentic, but it didn’t know that authenticity wasn’t safe. That judgment call, connecting a name to a hidden risk, is where the human step isn't optional. AI handled the mechanics in seconds. However, it was a human that protected the child.

Similar to the example above, the health care industry is filled with proven examples of how AI amplifies staff productivity. Yet every optimization opportunity also introduces quality risks. Unintended side effects and challenges of AI may include:

• poor-quality source data producing inaccurate outputs;
• increased error potential if automation is deployed without proper validation;
• new technical infrastructure requirements;
• additional quality assurance responsibilities;
• complex regulatory reporting considerations;
• evolving workforce training needs;
• algorithmic bias; and
• staff overreliance on automated recommendations.

Perhaps the greatest misconception surrounding AI is that it removes the need for human oversight. In reality, successful AI programs require more structured governance, stronger quality monitoring, and clearer accountability than traditional workflows. A collaborative partnership between machine and human is considered industry best practice, with established guidelines for the responsible use of AI technology across the entire enterprise.

Responsible AI Is an Organizational Priority
Organizations require formal frameworks for responsible AI implementation. This is especially true as capabilities expand across departments. Proper AI usage is not the sole responsibility of health IT departments. It is an enterprise governance issue that directly affects compliance, privacy, quality, and patient trust.

Six principles guide effective AI initiatives within and beyond HI teams:

1. Protect Patient Privacy: Every solution maintains HIPAA compliance, enforces appropriate access controls, and safeguards PHI. Leading organizations establish clear enterprise policies prior to utilizing patient data for model development, validation, or ongoing algorithm tuning.
2. Preserve Accuracy and Clinical Integrity: AI-generated recommendations—such as coding suggestions, documentation support, or record summaries—are validated by qualified HI professionals. Automation assists decision-making but does not replace accountability for medical record integrity.
3. Ensure Transparency and Explainability: Stakeholders must understand when automated systems, rather than human staff, steer workflows or influence recommendations. System auditability, clear confidence scoring, and documentation that supports regulatory review are essential components of responsible AI.
4. Evaluate Equity and Mitigate Bias: Routine assessments evaluate AI tools for potential disparities across patient populations, including race, ethnicity, age, language, gender, and socioeconomic factors. Bias testing begins prior to implementation and continues throughout the technology’s lifecycle.
5. Keep Humans in the Loop: Consequential decisions include human oversight to review, override, and escalate AI-generated recommendations when necessary. Human expertise remains essential for protecting patients and ensuring regulatory compliance.
6. Establish Governance and Accountability: Effective AI governance teams are multidisciplinary, bringing together HI, compliance, privacy, legal, information security, clinical leadership, and health IT. These teams collaborate to evaluate new AI use cases, monitor system performance, approve implementations, and oversee ongoing quality assurance.

The next several years will bring far more sophisticated automation into HI operations. Generative AI, intelligent agentic agents, advanced document understanding, predictive workflow management, and autonomous process orchestration are already influencing operational strategy.

The Agentic ROI Department
Future ROI workflows will rely on "agentic AI." These are intelligent agents capable of multistep reasoning and semi-independent execution across connected but disparate health care systems. Agentic AI agents operate with defined guardrails rather than full autonomy.

Rather than waiting for a request to enter a static queue, an agentic AI system will independently communicate with payer portals and EHR platforms through standardized interoperability APIs, coordinate cross-system record fetches, and dynamically predict fulfillment bottlenecks before they occur. It will generate precompiled, compliance-checked release packages ready for humans.

Regulations like HIPAA and 42 CFR Part 2 will require documented human accountability at key decision points and review across multiple checkpoints. ROI professionals will transition from record aggregators into strategic workflow orchestrators, exception handlers, and compliance auditors. They will spend more time resolving the ambiguous, high-risk cases that agents escalate to them.

Achieving meaningful operational value remains essential as AI advances. Technology should never be implemented simply for technology’s sake. This principle holds true in HI and across every other department in provider organizations.

Three guiding points ensure value across future AI investment decisions:

• Choose Wisely: Not every operational challenge requires an AI tool. Traditional robotic process automation, workflow software, or simple process redesign often provides a more effective and less costly solution. Evaluate whether AI is truly necessary before introducing technical complexity.
• Think Strategically: Every investment must connect to measurable organizational outcomes. Whether the objective is reducing turnaround times, improving quality, lowering operational costs, strengthening compliance, or enhancing the patient experience, clearly defined success metrics are a necessary precursor to any system implementation.
• Move Carefully: Governance cannot be an afterthought. Establish clear ownership for AI-enabled workflows, define quality monitoring processes, document accountability, and continuously evaluate performance against regulatory and operational expectations. Look for industry validators such as Best-in-KLAS recognition and compliance with Alliance for Health Information Operations and Standards best practices for ROI.

Leaders should prioritize use cases that have already demonstrated measurable value across operations. Walk before you run.

Technologies supporting document classification, automated request intake, record retrieval, workflow prioritization, quality monitoring, and administrative task automation continue to mature and produce tangible improvements. These practical applications deliver stronger, more predictable returns than experimental AI initiatives.

Human Collaboration Matters
As intelligent automation changes the mechanics of ROI, the focus of HI leadership must center on the workforce. AI does not replace the human element in ROI; it elevates it. Prepare your team through proactive change management, targeted upskilling, and clear communication.

• Reframe the Narrative: Empower staff to see AI as a supportive partner rather than a threat. Highlight how automation removes tedious manual tasks. This may include logging requests, searching disparate systems, or indexing scanned pages. Also allow specialists to focus on high-value, satisfying work that requires specialized credentials.
• Upskill for New Roles: The skill set for ROI specialists will evolve as AI automates administrative routines. Invest now in training programs that enhance data literacy, critical thinking, regulatory analysis, and audit management. Staff will need to understand how to evaluate AI outputs, interpret confidence scores, and identify system anomalies. Prompting and context are skills that should be mastered to successfully use AI.
• Foster a Culture of Continuous Feedback: ROI staff are the first line of defense in identifying when an AI model strays or misinterprets a document. Establish accessible internal channels where team members can flag false positives, suggest workflow refinements, and actively participate in tuning the technology. Collaborative intelligence models, where both humans and machines continually learn from each other, are preferred.

By grounding AI adoption in human-centered leadership, HI organizations can build resilient, highly skilled ROI teams equipped to navigate the future of health care information management with confidence.

— Angela Rose, MHA, RHIA, CHPS, FAHIMA, is the vice president for client success at MRO.