AI in the Front Office

AI and Prior Authorization: The Current State

What AI can and cannot do for prior authorization today, what the CMS rule changes from 2026, and a workflow that keeps clinical judgment with people.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing7 min read

On this page
  1. The size of the problem
  2. Two sides of the same request
  3. What provider-side tools can do now
  4. What tools cannot do
  5. What the CMS rule changes
  6. A workflow with automation in the right places
  7. Scripts for the front desk
  8. Evaluating a prior authorization tool
  9. Example: a busy orthopedic office
  10. Measuring progress
  11. Frequently asked questions
  12. References
AI in the Front Office: AI and Prior Authorization: The Current StateArticle coverAI IN THE FRONT OFFICEAI and PriorAuthorization: TheCurrent StateAva Medical IntakeInsightsAI in the Front Office: AI and Prior Authorization: The Current StateArticle coverAI IN THE FRONT OFFICEAI and PriorAuthorization: TheCurrent StateAva Medical Intake

Key takeaways

  • Prior authorization takes an average of 13 hours of physician and staff time a week, and 40% of physicians employ staff just for it[1].
  • Provider-side tools help most with finding out whether authorization is needed, assembling documents and tracking status. Clinical justification still needs a clinician.
  • Since January 1, 2026, impacted payers must decide urgent requests within 72 hours and standard ones within seven calendar days, and give a specific reason for denials[2].
  • Automation should start with status tracking and patient updates, which are low risk and save the most phone calls.

The size of the problem

Prior authorization is one of the heaviest administrative loads in a practice. In the AMA's 2026 survey, physicians reported completing an average of 40 prior authorizations a week, consuming 13 hours of physician and staff time, and 40% said they employ staff who work exclusively on authorizations[1]. About a third, 32%, said requests are often or always denied[1].

Physicians' experience of prior authorization, 2026Share of US physicians in the AMA prior authorization survey
Physicians' experience of prior authorization, 2026Bar chart. Say it delays access to necessary care: 95%; Say it contributes to burnout: 94%; Employ staff working only on prior authorization: 40%; Say requests are often or always denied: 32%.Say it delays access to necessary careSay it delays access to necessary care: 95%95%Say it contributes to burnoutSay it contributes to burnout: 94%94%Employ staff working only on prior authorizationEmploy staff working only on prior authorization: 40%40%Say requests are often or always deniedSay requests are often or always denied: 32%32%Physicians' experience of prior authorization, 2026Bar chart. Say it delays access to necessary care: 95%; Say it contributes to burnout: 94%; Employ staff working only on prior authorization: 40%; Say requests are often or always denied: 32%.Say it delays access to necessary careSay it delays access to necessary care: 95%95%Say it contributes to burnoutSay it contributes to burnout: 94%94%Employ staff working only on priorauthorizationEmploy staff working only on prior authorization: 40%40%Say requests are often or always deniedSay requests are often or always denied: 32%32%

Source: American Medical Association, 2026[1]

Show the data
ItemValue
Say it delays access to necessary care95%
Say it contributes to burnout94%
Employ staff working only on prior authorization40%
Say requests are often or always denied32%

The human effect is just as clear. Ninety-four percent of physicians said prior authorization contributes to burnout[1], and 95% said it delays access to necessary care[1]. Front desk teams feel it as well: in MGMA polling, eligibility and prior authorization were named the most time-consuming phone tasks for staff, ahead of scheduling[3].

It is no surprise that practices look to automation here. What matters is being precise about which parts of the process software can safely take on today, and which parts still depend on a clinician's reasoning and a payer's decision.

Two sides of the same request

Every authorization has a provider side, where the practice decides what is needed and builds the request, and a payer side, where the plan reviews it. Automation is arriving on both, and the two do not always speak the same language.

Where automation sits on each side of a prior authorization request.
StepProvider side todayPayer side today
Is authorization required?Rules engines and payer lists check the service and planPublished lists, often changed during the year
Assemble the requestTools pull notes, codes and results from the EHR into a draftPortals and fax lines receive requests in different formats
Clinical justificationClinician writes or approves the medical necessity rationaleClinical criteria applied by reviewers and automated rules
Submit and trackPortal submission and automated status checksStatus updates through portals; APIs required for impacted payers from 2027
Decision and appealStaff draft appeals; clinicians signDecision with a specific denial reason required for impacted payers

What provider-side tools can do now

  • Flag at scheduling or ordering whether a service is likely to need authorization for the patient's plan, so nobody finds out at check-in.
  • Gather the documents a payer usually wants, such as recent notes, imaging, lab results and prior treatments tried, into a draft packet.
  • Pre-fill payer forms and portal fields from the chart and the order.
  • Check status automatically on a schedule and update the work queue, replacing manual portal logins and calls.
  • Send the patient a plain-language update when the status changes.
  • Draft an appeal letter structure from the denial reason for a clinician to complete and sign.

These are document and status tasks. They save staff hours because they remove searching, copying and checking, not because they replace clinical reasoning. Seven in 10 physicians in the AMA's 2026 AI survey said they see AI as a tool to automate tasks that drive burnout[4], and prior authorization paperwork is the obvious example.

What tools cannot do

  • Decide that a service is medically necessary. A clinician owns that judgment and signs the request.
  • Guarantee approval. Payer criteria, plan rules and reviewer decisions sit outside the practice.
  • Replace peer-to-peer calls. When a payer requests one, a clinician needs to be on it.
  • Know about undocumented clinical context. If it is not in the chart, a tool cannot include it.
  • Work evenly across payers. Coverage of payer portals and forms varies, so many requests still need manual steps.

Rule of thumb: let software assemble and track; let clinicians justify and sign; let staff talk to patients.

What the CMS rule changes

The CMS Interoperability and Prior Authorization Final Rule applies to Medicare Advantage organizations, Medicaid and CHIP managed care and fee-for-service programs, and Qualified Health Plan issuers on the federally facilitated exchanges[2]. Its main dates for practices:

  1. From January 1, 2026, impacted payers must send decisions within 72 hours for expedited requests and seven calendar days for standard requests. Exchange plan issuers are excluded from these timeframes[2].
  2. From 2026, impacted payers must give a specific reason when they deny a request[2].
  3. From January 1, 2027, impacted payers must implement a Prior Authorization API, which is meant to let provider systems check requirements and submit requests electronically[2].

For practices, the practical step is to track decision times by payer against those deadlines and to capture the denial reason in a structured field. Both make appeals faster and give you evidence when a payer is consistently slow. Commercial plans outside the rule may follow different timelines, so keep payer-specific notes.

A workflow with automation in the right places

  1. At ordering or scheduling

    The system checks whether authorization is likely required and creates a task with the service, plan and due date.

  2. Packet assembly

    Software drafts the packet from the chart. A coordinator checks it is complete and correct.

  3. Clinical sign-off

    The ordering clinician reviews the rationale and signs. No request goes out without it.

  4. Submission and tracking

    The coordinator submits; automated status checks run daily and move the task when the status changes.

  5. Patient updates

    The patient gets a text at submission, at decision and if more information is requested, with a number to call.

  6. Denials

    The denial reason is recorded in a structured field, an appeal draft is prepared, and the clinician decides whether to appeal or request a peer-to-peer.

Scripts for the front desk

Patients call repeatedly when they do not know where a request stands. A status line and simple scripts cut those calls.

At scheduling: โ€œYour plan needs approval for this test before we can schedule it. We'll send the request today. Most plans decide within about a week for routine requests, and we'll text you when we hear back.โ€

On a denial: โ€œYour plan didn't approve the request, and they've given us a reason. Your doctor is reviewing it and will decide whether to appeal. We'll call you by Friday with the next step.โ€

Evaluating a prior authorization tool

Vendors in this space range from simple status checkers to tools that claim to handle the whole request. Test claims against your own work, using fictional or properly protected sample cases, and ask for answers in writing.

  • Which of my top payers and services does the tool support end to end, and which only partly? Ask for the list, not a percentage.
  • How does it decide that authorization is required, and how often are those payer rules updated?
  • What exactly does it pull from the chart, and can a coordinator see and edit every item before submission?
  • Does anything ever submit without a person approving it? If so, can that be turned off?
  • How are status changes, requests for more information and denials surfaced, and how quickly?
  • Does it record decision times and denial reasons in a format I can report on by payer?
  • Will it connect to payer Prior Authorization APIs as impacted payers implement them, and on what timeline?
  • Who signs the business associate agreement, and where is patient data stored?

Also look for the failure points that vendors rarely volunteer. Packets built from incomplete notes are denied just as fast as handwritten ones. Payer portals change without warning, breaking automated checks. And automation can create a false sense that someone is watching a request when nobody is. Every open authorization should have a named owner and a due date, whatever the software does in the background. A weekly review of requests older than the expected decision time catches the ones that slipped through.

Example: a busy orthopedic office

Example: a fictional orthopedic practice handles many imaging and procedure authorizations each week. Staff spend much of the day logging into portals to check status. The practice starts with the lowest-risk automation: daily status checks and patient text updates. Coordinators stop most manual status calls and inbound where-is-my-approval calls fall noticeably.

Example, continued: in the second phase, the practice adds packet drafting for knee MRI requests only. Coordinators find the drafts often miss documentation of physical therapy already tried, so the clinicians add a standard line to their notes. Only then does the practice extend drafting to other services. Clinical sign-off stays mandatory throughout.

Measuring progress

  • Coordinator hours per week on authorizations.
  • Days from order to decision, by payer, against the CMS timeframes where they apply.
  • First-pass approval rate and top denial reasons.
  • Inbound status calls from patients.
  • Visits rescheduled because authorization was not ready.

Prior authorization is one of the areas practices most often name as a starting point for AI, though behind scheduling, calls and eligibility[5]. Eligibility checks usually come first in the same workflow; see AI for insurance verification.

Frequently asked questions

Can AI submit prior authorizations automatically?

Tools can draft packets, pre-fill forms, submit through supported portals and track status. A clinician should still review and sign the medical necessity rationale, and many payers still require manual steps.

What are the CMS prior authorization deadlines?

For impacted payers such as Medicare Advantage and Medicaid managed care, decisions are due within 72 hours for expedited requests and seven calendar days for standard requests from January 1, 2026, with specific denial reasons required.

Where should a practice start with prior authorization automation?

Start with status tracking and patient updates. They carry little risk and remove many phone calls. Add packet drafting next, one service at a time, with clinician sign-off kept in place.

References

  1. American Medical Association (2026). AMA survey: Prior authorization reform pledge falls short with physicians (opens in a new tab) โ†‘

  2. Centers for Medicare & Medicaid Services (CMS) (2024). CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) fact sheet (opens in a new tab) โ†‘

  3. Medical Group Management Association (MGMA) (2026). Phones are still a bottleneck costing medical practices time they can't afford (opens in a new tab) โ†‘

  4. American Medical Association (2026). More than 80% of physicians use AI professionally: AMA survey (opens in a new tab) โ†‘

  5. Medical Group Management Association (MGMA) (2026). Automatic for the people: AI moves for medical practices to boost the front office and access (opens in a new tab) โ†‘

Book a demo

See Ava handle your calls.

Book a short demo. Weโ€™ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo