Every code, next to the sentence that earns it
Why Coda shows the exact wording behind every proposed code, and why “defensible” is the only standard that survives an audit.
A surgical claim is only as strong as the sentence behind it. An auditor does not ask what was done in the operating room. They ask what the operative note says was done, and whether that wording supports the code on the claim. Everything Coda does starts from that fact.
The gap that costs the most is the honest one
Take a laparoscopic cholecystectomy. The procedure codes to CPT 47562. If the surgeon also performed an intraoperative cholangiogram, the correct code is 47563, and the difference is real work that was actually performed.
Now look at the note as it is usually written:
An intraoperative cholangiogram was performed.
One sentence, no technique, no findings. Everyone in the room knows the cholangiogram happened. On paper, the higher code cannot be defended: the descriptor for 47563 asks for more than a mention. So the coder has two bad options. Code 47562 and leave real work unbilled, or code 47563 and hope nobody reads closely.
The only good option is the one that rarely happens: get the surgeon to document what they saw, in their own words, while the case is still fresh. That is the gap Coda is built to close.
What defensible means
We use the word carefully. A claim is defensible when four things hold:
- Every approved code traces to documentation. The evidence trail an auditor would ask for already exists, case by case.
- Flags run in both directions. The software tells you when the note does not support a code, including a code you were about to bill.
- Documentation drives the codes, never the reverse. No code the note does not support, and no leading question to a surgeon to manufacture support.
- Everything is on the record. Every access and every AI action lands in an append-only audit log.
Notice what is missing from that list: revenue. A defensible claim often pays more than a sloppy one, because complete documentation captures work that was previously left out. But that is a consequence, not the goal. Software that optimizes for reimbursement eventually drifts into upcoding, and upcoding is the one thing we will never do.
Why we show the evidence, not a score
Many coding tools present a code with a confidence percentage. We think that gets the interface backwards. A percentage cannot be argued with, cannot be checked, and does not survive an audit.
A highlighted sentence can. When Coda proposes a code, it sits next to the exact wording in the note that supports it. A coder who disagrees can see precisely why the code was proposed before changing it. A coder who agrees approves something they have read, not something they have trusted. And when the note does not support a code, the flag points at the missing wording rather than at a number that dropped.
The same rule applies to questions. When a biller asks whether the documentation supports 47563 rather than 47562, the answer cites the note in front of them and the code descriptor behind it. Every answer shows its sources, and every answer teaches.
Where the line is
Coda asks. The surgeon answers. The biller approves. None of those roles belong to the software.
Coda never writes a clinical fact the surgeon did not state. It never submits a claim. It never overrides a person’s judgment. It advises, in the open, with its reasoning visible and its actions logged. That is a narrower job than “automated coding,” and we think it is the only version of the job that a hospital can stand behind.
If you sign operative notes or defend the claims built on them, we would like to show you a real case prepared end to end. Write to coda@unira.io.
← All posts