What inpatient autonomous coding should (and shouldn't) do

Artificial intelligence (AI) is changing the landscape for healthcare revenue cycle leaders, especially as pressure grows to adopt autonomous coding. But inpatient coding is unique. Its complexity and depth make inpatient environments fundamentally different from outpatient settings. As you consider automation, it’s critical to set realistic expectations, avoid shortcuts and choose the right starting point.
 

The limits of retrofitted outpatient logic

Inpatient coding isn’t just outpatient coding at scale. Unlike single, same-day discrete outpatient visits, inpatient encounters can span days or weeks, involve multiple diagnoses, procedures, complications, comorbidities, etc., and require synthesis of a complete patient story from layers of clinical notes. Inpatient coding involves concurrent coding as well as post-discharge final coding efforts. As we look toward automation, we also must look at technology differently and cannot fit models built for outpatient coding requirements into inpatient coding ones. It is a case of trying to fit a square peg into a round hole. 
 

Understanding inpatient complexity

Success in inpatient autonomous coding means respecting the depth and complexity of inpatient hospital care. Truly addressing inpatient coding challenges requires solutions and evaluation criteria that account for the full clinical journey, not just isolated data points. Automating coding in the inpatient setting requires more than just a large language model. It needs AI that can accomplish what a highly skilled medical coder can do, bringing the patient story together and using critical thinking skills to solve the puzzle and come up with a complete and compliant final code set.
 

Assessing readiness for inpatient coding automation

Every organization is different. Taking a candid look at your status quo and capabilities will get you on the right track. Here are six key actions to focus on:

• Assess the quality and consistency of clinical documentation across inpatient encounters.
• Review how coding automation would fit into current workflows, including concurrent and final coding.
• Inventory all documentation sources coders rely on to build the full patient story, including physician documentation, ancillary documentation and other relevant records.
• Confirm that automation can reliably access the documentation required for accurate and compliant coding.
• Identify strengths to build on, and roadblocks that could slow adoption.
• Set clear expectations at all levels, from executive sponsors to front-line coders, to support smoother adoption and minimize disruption.

Choosing a strategic starting point

In my experience with healthcare organizations and going through go-lives myself, I believe that doing a phased implementation is the way to go. The most successful teams identify well-defined, lower-variability cases (perhaps a service line or admission type) where documentation is strong and coding rules are clear. Starting small builds confidence, gives you measurable outcomes and lets you iterate based on real feedback. Gradual expansion to more complex scenarios happens naturally as your team’s trust grows and workflows adapt.
 

Thoughtful change, lasting results

It’s easy to be swept up in AI promises, but true progress in inpatient autonomous coding depends on realism, transparency and a people-first mindset. Focus on continuous learning, involve your experts and remember that inpatient automation is a journey requiring deliberate steps. With clarity and intention, revenue cycle leaders and HIM directors can guide their organizations to smarter, more sustainable automation built for the true complexity of inpatient care.

Shawn Wells, global product owner, inpatient autonomous coding at Solventum.