The Drawer of Forms: Drafting Sick Notes, Insurance Forms and Referral Letters From the Chart

There is a drawer in every family practice, and an inbox that behaves like one. It holds the insurance form that needs three dates from a chart nobody has open, the disability paperwork that has been waiting nine days, the school note, the pre-operative letter a surgeon’s office keeps faxing about. None of it is clinical work and all of it requires a clinician. It gets done at 8pm, badly, in the order the pile happens to be in.

This is the transformation we would run for that practice. It is presented as a concept, built on the same method we use in client engagements, so the before, the after, and the path between them are all visible.

What does the paperwork actually cost a practice?

The Commonwealth Fund published an issue brief on administrative burden in primary care in October 2025, based on hour-long qualitative interviews conducted in March 2025 with twelve people, eleven of them primary care physicians in family and general internal medicine. That is a small qualitative sample and it is quoted here as evidence of what clinicians describe rather than as a measurement of how many hours it takes.

The measurements come from Canada. OntarioMD’s evaluation of AI scribes across 162 primary care providers found that in routine practice, clinicians reported roughly a three-hour weekly reduction in after-hours administrative time. Canada Health Infoway’s national AI Scribe Program, whose six-month results were published in June 2026 across a cohort of more than 12,000 enrolled primary care clinicians, found nearly 70% reporting reduced administrative burden and almost 80% reporting lower cognitive load.

Both of those measured documentation of the visit. The drawer is what is left over once the visit note is handled, and it is the part no scribe touches, because it is not generated by a conversation with a patient. It is generated by a third party who wants facts the chart already holds.

Before and after: the form

Drag the handle. The before is the blank third-party form beside a chart nobody has time to read; the after is the drafted form designed in its place.

The form does not change, because a third party owns it. What changes is that the facts arrive already pulled, each one showing the chart entry and date it came from, leaving the clinician the part only a clinician can do.

What we built

The concept above is the output of the same engagement steps we run in client work, and two of them are decisions a competent peer could argue with.

• We treat the form as a schema, not a document. Each third-party form becomes a map from its fields to the chart facts that answer them, which is why adding the next form is hours of mapping rather than a rebuild. The alternative, having a model read a scanned PDF and free-write into it, demos better and fails silently.
• We show provenance on every extracted fact, with the chart entry and its date, and we flag anything older than a configurable window rather than using it silently. Stale is the dangerous failure here, not wrong: a resolved diagnosis carried into a disability form is a specific and expensive kind of error.
• We stop the draft at the clinical opinion, deliberately. The system produces dates, diagnoses, medications, investigations and their results, and writes nothing about prognosis, capacity or restriction. This is the decision most likely to be argued with, because those are the fields that take the longest.
• We build the queue to sort by deadline and by who is waiting, replacing an ordering that was previously determined by which piece of paper was on top.
• We wire evals from day one against the practice’s own completed forms, scored on whether each extracted field matched what the clinician actually entered, and every correction becomes a test case.

How it works

The EMR stays the source of truth. An incoming form is matched to its mapping, the mapped fields are drafted from the chart with provenance attached, and anything stale or missing is flagged rather than filled. The clinician reviews the facts, writes the opinion, and signs. Only the signed document leaves the practice. Nothing is transmitted, faxed or filed on the system’s own initiative.

Because chart context stays in Canadian infrastructure and processing runs with zero data retention, the design lines up with what PHIPA expects: a completed privacy impact assessment before deployment, an agreement naming the processor as an agent under section 17, and an audit log on every access to personal health information.

Why this pays back

The Infoway results give the honest shape of the return: clinicians saved over an hour a week during work hours, and only 20% of them turned that time into appointment availability, while 50% spent it on other administrative work. That is the number a practice should plan against. Time returned does not automatically become capacity, and a practice that assumes it will has bought an hour of relief rather than an hour of revenue. Deciding in advance which of the two you want is part of the engagement, not an afterthought.

Break-even for the first slice is about forty drafted forms and letters a week at a typical loaded hourly cost, and the practice should substitute its own figure.

The limit worth stating: this does nothing for the forms nobody has mapped yet. A practice facing a long tail of one-off third-party paperwork, rather than a few high-volume forms, will see a much smaller share of the drawer disappear, and should ask for the volume breakdown before committing.