Prior authorization runs on documentation. A payer wants the diagnosis, the medical-necessity rationale, and the conservative care already tried, all pulled into a form. A thorough visit note carries most of that clinical story, so an ambient scribe that drafts a complete note cuts the legwork. It can’t file the PA, though, and this page is honest about that line.
So let’s draw it clearly. An AI scribe is not a prior-authorization tool. It doesn’t fill payer forms, it doesn’t touch a payer portal, and it doesn’t submit anything. What it does is hand you a more complete note, which is where a PA gets its evidence. Real lever, narrow one. The 2025 AMA survey puts the burden at about 13 hours of physician and staff time a week, so the part a scribe can cut is worth understanding precisely.
Key takeaways
- The 2025 AMA survey found practices complete 40 prior authorizations per physician per week and spend ~13 hours on them; 95% of physicians say PA delays care.
- A PA needs the diagnosis, the medical-necessity rationale, conservative care tried with dates, and supporting findings. CMS says it’s the same information already required for payment, just earlier.
- A complete visit note covers most of that clinical content. It does not cover the form, the codes, or the submission.
- An AI scribe drafts the note and suggests ICD-10 codes. A person still assembles and files the PA.
prior authorizations per physician per week (2025 AMA survey of 1,000 physicians)
of physician and staff time on PA each week (2025 AMA survey)
of physicians say prior authorization delays access to necessary care
Why does prior authorization create so much documentation work?
Because a PA is a payer asking you to prove the case before the service, in writing. And the proof lives in the chart.
The numbers are not small. The 2025 AMA Prior Authorization Physician Survey, which polled 1,000 practicing physicians (400 primary care, 600 specialists) in December 2025, found that practices complete an average of 40 prior authorizations per physician every week and burn about 13 hours of physician and staff time getting them through. That’s most of a full working day, every week, spent largely on paperwork rather than patients. The same survey found 95% of physicians say PA delays access to necessary care, 79% report patients abandoning treatment over authorization struggles, and 26% report a PA that led to a serious adverse event for a patient. This is not a minor admin tax. It’s a patient-safety problem with a documentation core.
Here’s the part worth sitting with. A chunk of that 13 hours isn’t the submitting. It’s the reconstructing. Someone has to find the diagnosis, dig out which conservative treatment the patient already failed and when, and write up why this drug or this imaging study is necessary now. When the note from the original visit is thin, that’s a research project. When the note is complete, it’s a copy job. The documentation burden of a PA is, to a real degree, the cost of a note that didn’t capture enough the first time.
What documentation does a prior authorization actually need?
Less mysterious than it feels. CMS puts it about as plainly as it gets: prior authorization “does not create new clinical documentation requirements,” it “requires the same information that is already required to support Medicare payment, just earlier in the process.” The patient’s medical record, CMS says, has to “contain documentation that fully supports the medical necessity for services.” Read that twice if you’ve ever felt like a PA was asking for something exotic. Mostly it’s asking for a good note.
What a typical payer PA wants, and where a well-drafted visit note covers it:
| What a typical PA asks for | Where a complete visit note covers it | Does the note cover it on its own? |
|---|---|---|
| The diagnosis, coded specifically | Assessment, with an ICD-10 suggestion to confirm | Mostly. You confirm the code. |
| Medical-necessity rationale | Subjective and assessment: why this, why now | Yes, if the note captured your reasoning |
| Conservative treatments already tried | History and plan: prior meds, therapy, outcomes, dates | Yes, when those details were said and captured |
| Relevant exam and test findings | Objective: vitals, exam, results referenced | Yes, for what was discussed in the visit |
| Symptom duration, severity, functional impact | Subjective: onset, frequency, how it limits the patient | Yes, when it surfaced in the conversation |
| The specific payer form and fields | Not in the note | No. A person fills the form. |
| CPT / procedure or drug codes for the request | Not produced by the scribe | No. That’s coding, downstream of the note. |
| Submission through the payer’s portal or fax | Not in the note | No. The note is evidence, not a transmission. |
The top five rows are the clinical substance of a PA, and they’re exactly what a thorough note already contains. The bottom three are the parts a note will never be: the form, the procedure codes, the act of submitting. That split is the whole honest story of this page. Worth flagging too: exact requirements vary by payer, plan, and service. Treat the table as the pattern, not a checklist any single insurer guarantees.
If you want to judge how much of that top half a note actually captures on your visits, book a demo and watch a structured note land about two minutes after a sample encounter, then read the assessment and history sections against this table yourself.
How does an ambient scribe cut the prior-authorization legwork?
By making sure the clinical story is captured at the visit, complete and specific, so nobody has to rebuild it later from a four-line note.
The AI Medical Scribe by Patient Square, one module of Practice Copilot, captures the visit as it happens and hands back a structured SOAP note, ICD-10 suggestions, and a prescription draft, ready to review and sign about two minutes after the visit. For prior authorization, the value is concentrated in that note. A note written while the patient is describing which medication they tried last year and why it didn’t work tends to capture the conservative-care history a PA hinges on. A note built from memory at 9pm tends to lose it. The ICD-10 suggestions narrow the diagnosis code your team confirms, and the medical-necessity reasoning lands in the assessment because it was said out loud in the room.
What that buys you is upstream, and it’s specific: when your biller or care coordinator opens the chart to build the PA, the clinical evidence is already there, legible and dated. They’re transcribing a complete record into a form, not chasing you down the hall to ask what the patient failed before. That’s the legwork an ambient scribe cuts. If you want to see how complete the notes come out on your own visits, our pillar on cutting charting time walks through the time math, and the pricing page lists exactly what ships today, with no PA-submission asterisks.
Can’t the AI just do the prior authorization for me?
No. And we’re going to be blunt here because this is where the category gets sold dishonestly.
AI Medical Scribe by Patient Square does not submit prior authorizations. It does not fill out payer forms. It does not integrate with payer portals or clearinghouses. It does not generate CPT procedure codes or E&M visit levels, and it does not e-prescribe. It drafts a clinical note and suggests ICD-10 diagnosis codes for your team to confirm. Full stop. A person still reads the note, decides the service, fills the payer’s form, attaches the record, and submits the request. The human owns the PA from form to fax.
If a vendor’s pitch implies the AI “handles prior auth” end to end, ask them three things: does it connect to my payers, does it submit the request, and who’s liable when an auto-submitted PA is wrong. Those questions usually end the overclaim. We hold the line on the boundary the same way we do with coding, which we lay out in the honest version of ICD-10 suggestions: a scribe suggests, a person decides. The honest tool helps you build the documentation faster. It does not pretend to be the payer-facing workflow, because it isn’t one, and a draft an LLM wrote has no business being auto-filed to an insurer.
Where does a better note stop helping your PAs?
At the edge of the documentation. Past that, a note does nothing, and pretending otherwise is how this topic goes wrong.
A complete note helps the denials and delays that turn on clinical evidence: insufficient medical-necessity documentation, missing history of conservative care, a diagnosis that wasn’t specific enough. Those are real, and they’re common. But a thorough note can’t move a denial caused by a formulary exclusion, a coverage limit, a plan that simply doesn’t authorize that service, or a request that went in on the wrong form. In the 2025 AMA survey, 32% of physicians said PA requests are often or always denied, and plenty of that is policy, not paperwork. No scribe touches policy.
So the honest framing stays narrow. Documentation is a large, fixable part of the prior-authorization burden. An ambient scribe attacks that part by capturing a complete note at the visit. It does not attack the form-filling, the codes, the submission, or the payer rules, and we won’t dress it up as if it does. Buy it as a way to make your notes more complete and your PA paperwork less of a research project. Don’t buy it as a prior-authorization engine, because that’s not what it is. If documentation completeness is the bottleneck you actually feel, book a demo and watch a structured note appear about two minutes after a sample visit, then run the 7-day free trial and check whether your real notes come out complete enough to build a PA from. For the post-submission side of revenue, where the same documentation logic applies to bounced claims, see whether better notes cut claim denials.