What's an effective AI prompt for drafting a prior-authorization request letter to an insurer

The Prompt

Draft a prior-authorization request letter to [insurer name] for [procedure/medication/service].

Patient diagnosis/ICD code: [diagnosis]
Requested service/CPT code: [service]
Medical necessity — why this is needed: [clinical justification, prior treatments tried and failed]
Ordering provider: [name, credentials, NPI]

Write in a formal, factual tone. Lead with the medical necessity justification, since that's what reviewers weigh most heavily. Keep it to one page.

Why This Prompt Works

Prior-auth denials disproportionately come from letters that bury the medical-necessity justification under administrative detail — reviewers are working through a queue and weight that section most heavily, so leading with it measurably improves how the letter reads on first pass.

How to Customize It

Paste the payer’s own stated criteria for that procedure if you have it (many publish medical policy documents) — mirroring their own language back to them strengthens the letter. If a prior authorization for the same patient was recently denied for a related service, reference the resolution briefly, since reviewers often work through a patient’s history as a batch.

Common Mistakes to Avoid

Don’t let the model invent “prior treatments tried” — that has to be the patient’s real documented history, since it’s a specific factual claim to the insurer. Also avoid vague diagnosis or procedure codes — an imprecise or mismatched code is one of the most common reasons a request gets kicked back for resubmission rather than denied outright, which just adds delay.

Bottom Line

The prompt that gets prior-auth letters approved fastest puts real, provider-supplied medical necessity first and treats the model as a drafting tool for structure and tone, not a source of the clinical facts.