The last chore after the last patient
After the documentation is done, the coding remains: translating the clinical work into billing and diagnosis codes, accurately enough to be paid and defensibly enough to be audited. It is exacting, repetitive work that sits squarely between the clinician and the end of the day. We are building coding assistance for JamEMR to take on the drafting.
Coding assistance is on our roadmap and does not exist in JamEMR today. Here is what we are planning and the principles it will be built under.
What we are planning
- Codes proposed from the note. After a note is signed, JamEMR is planned to analyze the documented encounter and suggest applicable diagnosis and billing codes, each linked to the documentation that supports it.
- Evidence, not just answers. A suggested code without its justification is a liability. Every proposal is planned to cite the specific documentation it rests on, so review means reading a line, not re-reading the note.
- Documentation gap flags. When the note supports more specific coding than it currently justifies, the clinician can be shown what is missing — while the encounter is still fresh.
- Human review, always. Suggested codes are drafts. Nothing is finalized or submitted without review by the clinician or coding staff, and every acceptance is auditable.
Built on documentation that is already better
Coding assistance is only as good as the note beneath it, and this is where JamEMR starts ahead. Ambient documentation produces complete, structured notes; the Clinical Knowledge layer holds the encounter’s problems and findings in structured form. Coding suggestions drawn from that foundation can be grounded in what was actually documented — not inferred from fragments.
Local processing, as always
Coding runs on the most detailed clinical narrative a practice produces. Like every AI capability in JamEMR, coding assistance is planned to run on the practice’s own dedicated local GPU hardware — encounter documentation is not sent to third-party consumer AI clouds to be coded.
Practices with strong opinions about where coding help matters most — evaluation and management levels, specificity, denials — are invited to share them through our contact page as we prioritize this work with pilot practices.