For AI prior-authorization and UM vendors

Proof a doctor actually decided.

A growing number of states require a physician, not an algorithm, to decide medical-necessity denials. Attending gives every AI-assisted denial a signed receipt: a license that's active in the patient's state, evidence the physician engaged with this patient's records, and their own rationale, bound to the exact case file.

Designed so patient data stays in your environment · Embeds in your reviewer screen

Signed determinationverified
Case
UM-24-88213 · sha256:4be1…c07a
Determination
Deny · inpatient rehab, days 8–14
Physician
MD · license active in AZ
Engagement
5 of 5 key records · 11m 40s on case
Rationale
Patient-specific · own words
Pre-check
Independent AI (different vendor) · no flags

Why now

The laws say a physician decides. Today's logs can't show one did.

CA

physician decides

California SB 1120, in effect since January 1, 2025: a denial, delay or change of care based on medical necessity must be decided by a licensed physician or qualified health care professional, not an algorithm. Source

AZ

individual review

Arizona HB 2175, effective July 1, 2026: a medical director must individually review each medical-necessity denial and may not rely solely on recommendations from any other source. Source

0

independent proof

in a typical reviewer log. A "reviewed by" field shows someone clicked approve. It doesn't show a licensed physician engaged with this patient, and the vendor is grading itself.

Other states have passed or are considering similar rules. Summaries, not legal advice; check the statute text for your state.

Try it

Sign a denial. Attending decides whether it counts.

A synthetic case. Open the records you'd need, write a rationale, and sign. Or skip straight to signing, the way a rubber stamp would, and see what happens.

Case UM-24-88213 · Arizona membersynthetic data

AI recommendation: Deny continued inpatient rehab (days 8–14). Does not meet criteria.

Key records (click to open):

No record open.

Simulated for illustration; demo time thresholds are shortened. Signatures use real Ed25519 keys generated in your browser. Nothing from the demo leaves this page.

What a sign-off proves

Four facts, signed and bound to the case file.

01

Licensed for this patient

The signer's license is active in the member's state, checked against NPI and state board records at signing.

02

Engaged with the records

Which key records were opened, and time on case. Seconds-per-denial review doesn't produce a receipt.

03

Their own rationale

Patient-specific reasoning in the physician's words. An independent AI from a different vendor flags boilerplate before signing.

04

Bound to the exact case

Signed over a digest of the case file and determination. Change either one afterward and the receipt stops matching.

Built for your stack

Designed so patient data stays in your environment.

Attending is designed to receive only fingerprints (hashes), timestamps, license checks and signatures, never PHI. The AI pre-check is designed to run inside your environment with your own model keys, so only its signed verdict leaves.

  • Embeddable sign-off step. One widget in your reviewer screen, with a passkey signature and engagement capture.
  • Receipt API. Every determination returns a signed, verifiable receipt you store with the case.
  • Verification page. Regulators, auditors and appeals can check a receipt without access to your systems.
  • State rule packs. Starting with California and Arizona: physician decides, and individual review before denial.

For AI UM vendors

Your plans need proof. Hand it to them.

Plans in physician-review states need to show that a doctor decided. Your own audit log can't carry that weight, because you're grading yourself. An independent, signed receipt per denial can.

Early access is free while we build it.

Get early access

Make every AI-assisted denial one you can prove a doctor decided.

Join early access. We'll share the API, the sign-off widget and the verification page as they're ready.