Coding as a Service: The Allocation Problem Nobody's Solving Well
Automation can carry the volume. Revenue integrity depends on how you allocate the risk.
Part 2 of a 3-part series on the future of coding, auditing, and revenue integrity.
In Part 1, we made the case that middle rev cycle, the connected discipline of coding, documentation, and revenue integrity, is the most under-invested, highest-leverage part of the revenue cycle: the layer sitting between clinical documentation and final claim that nobody in the org chart cleanly owns. Coding sits at the center of that layer. So it's worth asking a more specific question: given the workforce shortage and the rise of AI-assisted coding, what does a well-run coding function look like now?
Most of the market answers that question in binary terms. Either you're still fully human, or you've moved to autonomous coding. We think that's the wrong frame. The better question isn't which side to pick. It's how to allocate.
Volume and risk aren't the same variable
A coding department handles a wide range of encounters, and those encounters aren't equally difficult or equally consequential. A routine outpatient visit and a complex inpatient stay with multiple comorbidities both need to be coded correctly, but they don't carry the same risk if something goes wrong. One is high volume and low variance. The other is lower volume and high variance, with real dollars attached to getting the judgment calls right.
Autonomous coding tools are genuinely good at the first category. They're fast, consistent, and increasingly reliable on high-volume, lower-complexity work like radiology and routine outpatient visits. They're much less reliable on the second category: the complex inpatient cases where a missed diagnosis or an incorrectly sequenced code can swing reimbursement by tens of thousands of dollars.
Treating those two categories the same way, whether that means throwing automation at everything or keeping everything fully manual, wastes capacity in one direction or takes on unnecessary risk in the other. Neither is a strategy. Both are defaults.
What allocation means in practice
A deliberate allocation model starts by segmenting the workload, not by encounter type alone, but by complexity and dollar risk. Low-complexity, high-volume work is routed to automation with light human spot-checking. High-complexity, high-dollar work goes to your most experienced coders, with automation handling first-pass structure so the human effort is concentrated on judgment rather than data entry.
This is a meaningfully different design from most organizations' today. Many coding departments still allocate people by department or by physician, rather than by where the actual risk lies. That made sense when everything was manual, and volume was the main constraint. It makes less sense now, when volume is no longer the constraint and expertise is.
Done well, this kind of reallocation does two things at once. It stretches a shrinking pool of experienced coders further, because they're no longer spending time on the routine cases that automation can handle. And it improves accuracy on the cases that matter most financially, because those are exactly the cases getting focused human attention instead of being split evenly across the team.
Why this is harder than it sounds
Building this model internally requires three things most health systems don't have readily available: a clear, ongoing view of where complexity and dollar risk concentrate in your case mix, coders trained and comfortable working alongside automation rather than around it, and enough flexibility in staffing to shift capacity as case mix changes month to month.
That combination is difficult to build and maintain internally, particularly for organizations already stretched thin on coding capacity. It's also exactly the kind of problem a coding-as-a-service model is built to solve: not by replacing your team, but by giving you the flexible capacity and allocation discipline to route work to the right resource, automated or human, based on where the risk sits.
Where this series goes next
Getting allocation right on the coding side raises a natural follow-up question: who's checking the work, especially the automated share of it? In Part 3, we'll look at auditing and make the case that as automation scales, a growing share of the audit opportunity isn't about catching human error anymore. It's about auditing the machine.
If your coding function is still organized around departments or physicians rather than complexity and risk, that's often a sign that meaningful accuracy and capacity are being left on the table. We'd welcome the chance to help you think through what a better allocation model could look like.
Get allocation right and get help building it from a team that's already closing the coding, documentation, and revenue integrity gap.
