As AI transforms medical coding, CDI, and health information management, leaders must prepare professionals for the work where their expertise creates the greatest value.
Blog Post Written by Andrea Faber, Shearwater Vice President, Quality & Training
Every conversation about artificial intelligence in healthcare seems to arrive at the same reassurance: “There will always be a human in the loop.” It sounds responsible. For employees worried about their future, it can also sound like a promise that their jobs will remain largely unchanged.
I believe we need to stop using it that way.
Human involvement can be an important part of an AI workflow. But the phrase tells us little about what that person will do, what authority they will have or how their involvement improves the outcome. Used as reassurance, “human in the loop” becomes a comfort phrase rather than a workforce strategy.
Medical coding, Clinical Documentation Integrity (CDI), health information management (HIM), quality, and auditing will change. Leaders should embrace that change while being honest about what it means for the people doing the work.
How Has Technology Changed Medical Coding and Health Information Management?
When I think about the changes ahead, I think back to medical transcription. For many in health information, the shift from dictated reports and manual transcription to speech recognition and electronic documentation felt almost overnight. The broader transition took years, but its impact on a department or an individual career could be sudden.
A useful milestone is 2009, when the Health Information Technology for Economic and Clinical Health (HITECH) Act helped accelerate electronic health record adoption. That was not the year transcription disappeared, nor was it the beginning of speech recognition. It marked a major policy investment in changing how healthcare information was captured and used. [1]
Traditional transcription work contracted, and the disruption was real. Speech recognition shifted some work from producing a first draft to editing and verifying machine-generated text. The Bureau of Labor Statistics describes transcriptionists reviewing speech-recognition drafts and correcting errors in medical terminology and meaning. [2] The work changed, but the need for accurate, meaningful clinical documentation remained.
As electronic documentation expanded, priorities also changed: documentation quality, information integrity, workflow support, and helping clinicians use new systems became increasingly important. These were opportunities to apply existing knowledge differently, although moving into them required training and did not guarantee a new position for every displaced worker.
That history shapes how I view AI today. We should acknowledge disruption without equating a changing job description with the demise of our expertise. Medical terminology, clinical understanding, attention to detail, and knowledge of how information is used can remain valuable even when the method of producing the record changes. Our responsibility is to help professionals carry that expertise into the next form of the work.
How AI in Medical Coding Is Changing the Role of Human Expertise: Preserving expertise without preserving every task
In my work across coding, quality, and training, I see a clear distinction between preserving professional expertise and preserving every activity in a job description. An experienced professional’s value extends well beyond the number of records reviewed or codes assigned.
Yet it is easy to introduce automation and retain the entire manual workflow around it. A system recommends a code, a coder checks it, and an auditor reviews the result. That arrangement may be appropriate during implementation or for particular risks. It should also be evaluated as the technology demonstrates its performance.
If a system can reliably complete a defined activity, requiring someone to repeat it indefinitely deserves scrutiny. What errors does that review prevent? What does it miss? Could the professional’s time create greater value elsewhere?
The decision should depend on demonstrated performance, case complexity, and the consequences of error. Human oversight should be purposeful, with clear responsibilities and authority to intervene. Having a person present does not, by itself, establish quality or accountability.
Where Human Expertise Still Matters in AI-Assisted Medical Coding and CDI
The strongest opportunities are in work that brings together interpretation, investigation, communication and responsibility for outcomes. In coding, CDI and HIM, several areas deserve immediate attention.
Complex interpretation and exception management. Consider a record with conflicting diagnoses, inconsistent procedure documentation or a clinical picture that does not support the suggested conclusion. The expert’s contribution is to investigate the conflict, apply relevant guidance and determine the appropriate next step. A useful future role will involve both resolving these cases and helping define which cases should reach a specialist. Those review criteria must be tested; a system’s confidence score alone should not determine whether a case is safe to release.
Clinical documentation and provider engagement. A potential documentation opportunity is only a starting point. A CDI professional must consider whether the complete record supports it, whether clarification is appropriate and how to communicate the issue without leading the provider. For example, an isolated clinical indicator may warrant investigation but cannot be treated as sufficient support for a diagnosis. Skilled provider engagement also turns recurring documentation problems into education and improvement.
Regulatory interpretation, appeals and disputes. Difficult coding questions, clinical validation disputes and payer denials often require professionals to assemble evidence, explain their reasoning and apply guidance to the facts of a particular case. The value lies in a defensible conclusion and the ability to communicate it clearly. AI can support that work, while designated professionals remain responsible for the organization’s decisions and escalation process.
AI governance and quality. Experienced coding and CDI professionals should help determine where automation is appropriate, what evidence is required before expanding its use and what triggers a return to more intensive review. Their work should include investigating recurring errors, evaluating performance across specialties and documentation patterns, and verifying that changes produce the intended results. One incorrect recommendation may be an individual defect; the same error across hundreds of records is a system problem.
Data integrity and education. HIM expertise can help trace an unreliable output to its source: incomplete documentation, conflicting information, an interface problem or an incorrect workflow assumption. Educators can then translate those findings into targeted learning for providers, reviewers and operational teams. This work connects individual findings to lasting improvements in the information on which healthcare depends.
What Will the Role of a Medical Coder Look Like With AI?
Medical Coding Specialist — AI-Assisted Coding: Resolve complex cases and coding exceptions; validate selected AI outputs based on risk; apply official coding guidance; collaborate with CDI and providers to clarify documentation; support appeals; and identify recurring errors that improve system performance. Requires strong coding foundations, sound judgment, clear communication and the ability to evaluate AI recommendations. Success is measured by accuracy, defensible decisions and prevention of recurring errors, alongside productivity.
How to Maintain Coding Accuracy and Quality With AI: Quality must evaluate the whole process
As automation expands, quality programs should retain appropriate case review while also examining how the process behaves over time. Reviewing only cases flagged by the technology can leave errors in supposedly routine cases undiscovered. Independent sampling remains important for understanding what the system fails to flag.
Quality leaders should ask which errors recur, how consequential they are and whether they cluster by service line, documentation source or case type. They should also examine missed opportunities, unnecessary escalations and the effect of changes to the system or workflow.
An overall accuracy percentage cannot answer every question. A strong aggregate result may conceal a recurring error with substantial compliance or reimbursement consequences. Quality reporting should distinguish frequency from impact and connect findings to corrective action, ownership and follow-through.
The future quality professional’s value will increasingly include identifying the cause of a problem and preventing its recurrence across the process.
What Skills Will Medical Coders Need in an AI-Driven Future? Training must move with the work.
Preparing professionals for these roles requires more than teaching them how to use an AI tool. Coding foundations, anatomy, pathophysiology, documentation standards and official guidance remain essential. Professionals need that knowledge to challenge a plausible recommendation and explain why it is unsupported.
Training should also develop investigation, evidence-based reasoning, data interpretation, provider communication and escalation decisions. Assessments should ask learners to reconcile conflicting evidence, recognize an inappropriate query opportunity or identify a recurring pattern across multiple cases.
Organizations should create practical pathways into this work through supervised assignments, calibration, mentoring and measured progression. Professionals need time and access to develop new skills. Telling employees to “adapt” without providing those opportunities is an incomplete response to workforce change.
How Healthcare Leaders Can Prepare Their Medical Coding Workforce for AI: Leadership owes the workforce an honest plan
My drive to move this conversation forward comes from what has always motivated me: helping people grow. Working with international teams has reinforced that belief. Without the willingness to push forward, invest in education and challenge assumptions about where expertise belongs, many of the professionals on my teams would not have the place in global healthcare they have today. Their opportunities grew because we prepared for new work and demonstrated what they could contribute. I bring that same commitment to AI: helping people develop the skills and confidence to take their expertise into what comes next.
Promising people that their jobs will never change is not protecting them. Preparing them for how their jobs will change is.
That preparation starts with concrete decisions: which activities are candidates for automation, which require continued review, what performance evidence supports those choices and which skills the organization needs next. Leaders must then connect those decisions to training, staffing and career opportunities.
Some tasks will disappear. Some roles may shrink, and others may emerge or expand. Moving professionals into more advanced work will require deliberate investment; it will not happen automatically because routine work has been removed.
I believe we should embrace AI’s potential and build a workforce capable of directing, evaluating and improving how it is used. Human expertise will create its greatest value by resolving difficult cases, strengthening documentation, investigating failures, communicating with clinicians and holding the overall process accountable.
“Human in the loop” describes involvement. A workforce strategy must explain the work, develop the expertise and define the responsibility. That is the conversation healthcare leaders should be having now.
This conversation is ever evolving. What I describe today may look different a year from now—or even a few months from now. As technology advances, we will need to revisit our assumptions, priorities and the ways we apply our expertise. Our willingness to keep learning will help us shape what comes next.
My challenge to each of us is to embrace change and prepare ourselves to ride the next WAVE! Ask questions, build new skills and help others move forward. Our children and grandchildren will inherit a world of work that continues to evolve. We can lead by example, showing them how to meet uncertainty with curiosity, judgment and the confidence to grow. Our expertise has a future, and we have a responsibility to help shape it.
References
[1] Office of the National Coordinator for Health Information Technology. Non-Federal Acute Care Hospital Electronic Health Record Adoption, 2008–2024. https://healthit.gov/data/data-briefs/non-federal-acute-care-hospital-electronic-health-record-adoption-2008-2024/
[2] U.S. Bureau of Labor Statistics. Occupational Outlook Handbook: Medical Transcriptionists. https://www.bls.gov/ooh/healthcare/medical-transcriptionists.htm
