An AST product · In pilot at a U.S. healthcare facility

Medexa.

Trusted AI for payer-ready approvals.

Medexa listens to the visit, codes it with evidence, and carries the claim all the way to a payer-ready approval — eligibility, prior authorization, submission and denial recovery, with a human approving every decision.

Ambient captureFHIR R4 coreHuman-in-the-loop
The problem

The paperwork is electronic. The work is still manual.

From US payer portals to the Gulf's mandated e-claims rails, eligibility checks and prior authorizations still mean staff re-keying data, chasing requests by phone, and racing hard regulatory clocks. The result is the same everywhere: slow approvals, avoidable denials, and clinical teams doing clerical work.

UAE

1-hour pre-auth response window in Dubai — a hard regulatory clock on every request.

KSA

NPHIES-mandated e-claims in Saudi Arabia — new rails, same manual re-keying behind them.

USA~24%of claim denials begin at registration and eligibility — the single largest cause in US denial-index data
USA86%of denials are potentially avoidable — the information existed; the workflow missed it
USA$19.7Bspent by US providers every year just adjudicating and fighting denials

US figures from industry denial-index and provider-spend studies, rounded; regulatory facts per DHA and NPHIES mandates.

How it works

From conversation to payer-ready approval.

One loop, end to end — the encounter is captured once, and Medexa carries it the rest of the way.

01/06

The visit is captured ambiently

Medexa listens to the clinician–patient conversation. No dictation, no forms — say “Hey Medexa, start session” and the visit documents itself.

Keep scrolling ↓
02/06

Codes surface with their evidence

ICD-10 and CPT codes appear live in the room, each linked to the exact words that justify it — coding review starts from proof, not guesswork.

Keep scrolling ↓
03/06

Everything becomes one clean record

Patient, coverage, encounter and codes normalize into canonical FHIR R4 — a single source of truth every downstream step reads from.

Keep scrolling ↓
04/06

AI agents draft the decision

Eligibility and prior-auth agents consult a deterministic Rules Engine and draft the decision with a risk-readiness score — never a black-box “confidence.”

Keep scrolling ↓
05/06

A human approves — always

Your reviewer sees the agent's reasoning and the named rule it cited, then approves, overrides or escalates. No unsupervised submission, ever.

Keep scrolling ↓
06/06

Payer-ready, on your rails

Approved requests are signed and submitted over the rails you already run — DHPO / eClaimLink, NPHIES — and every outcome feeds the learning loop.

Keep scrolling ↓
01

The visit is captured ambiently

Medexa listens to the clinician–patient conversation. No dictation, no forms — say “Hey Medexa, start session” and the visit documents itself.

02

Codes surface with their evidence

ICD-10 and CPT codes appear live in the room, each linked to the exact words that justify it — coding review starts from proof, not guesswork.

03

Everything becomes one clean record

Patient, coverage, encounter and codes normalize into canonical FHIR R4 — a single source of truth every downstream step reads from.

04

AI agents draft the decision

Eligibility and prior-auth agents consult a deterministic Rules Engine and draft the decision with a risk-readiness score — never a black-box “confidence.”

05

A human approves — always

Your reviewer sees the agent's reasoning and the named rule it cited, then approves, overrides or escalates. No unsupervised submission, ever.

06

Payer-ready, on your rails

Approved requests are signed and submitted over the rails you already run — DHPO / eClaimLink, NPHIES — and every outcome feeds the learning loop.

The agents

Five agents. One claim, carried to settlement.

Each agent owns one stage of the revenue cycle — and every decision it drafts cites the rule it applied.

Live

Eligibility

At registration

Verifies coverage the moment the patient registers — member ID, plan and policy checked against the payer before care begins.

Live

Pre-Authorization

Before care

Triggered by the ordered service itself. Drafts the prior-auth decision, cites the payer rule it applied, and starts the regulatory SLA clock.

Roadmap

Claim Submission

After approval

Assembles, scrubs and submits payer-ready claims over the rails you already use — no re-keying, no portal hopping.

Roadmap

Reconciliation

After payment

Matches remittances to claims automatically and flags underpayments and variances your team would otherwise never catch.

Roadmap

Denial & Appeals

After a denial

Drafts evidence-backed appeals — and learns which arguments actually overturn denials for each payer.

Submission, Reconciliation and Denial & Appeals are on the roadmap — shown in the product today so you can see where the platform is headed.

The complete suite

Not a tool. A revenue-cycle operating console.

Behind the agents sits a full platform — the surfaces your coding, billing and compliance teams live in every day. Click through the console:

Work queue · 19 awaiting review
SD
Lumbar Epidural InjectionPre-Auth Agent · drafted approve · cites rule R-214
SLA 9 min
AZ
CT Abdomen & Pelvis (urgent)“Documented trauma indication matches Daman imaging rule IMG-07.”
Drafted
ApproveOverride…Escalate

Your team stays in command.

Every drafted decision lands in a work queue with the agent's reasoning, the rule it cited, and the regulatory clock — reviewers approve, override with a reason, or escalate.

  • SLA countdown on every pre-auth in flight
  • Override reasons feed the learning loop
  • Full case context — coverage, codes, documents
82%first-pass approval in the working platform today
91%projected at full agent autonomy on a tuned payer mix
87%agent–reviewer agreement, measured in shadow mode
14 minaverage time from drafted decision to approval

Figures from the working platform on representative case data — pilot results will be published as the shadow-mode evaluation completes.

Now in pilot

Already running inside a U.S. healthcare facility.

Medexa is integrated for pilot testing at a U.S. healthcare facility today — wired into real scheduling and documentation workflows, with the agents in shadow mode learning the facility's payer mix before they earn assist. That's the trust model working exactly as designed: prove agreement first, act second.

  1. IntegrateWired into the facility's existing systems — no rip-and-replace.
  2. ShadowAgents run alongside staff, measured on every case. In progress.
  3. GraduateAssist mode unlocks per agent, per payer, as agreement clears the gate.
Why it's different

Built for trust, not just throughput.

A co-pilot, not another RCM

Medexa sits on top of the systems you already run. No rip-and-replace, no data migration — it integrates through the rails you already use.

Every decision cites its rule

A deterministic Rules Engine holds the payer- and procedure-specific rules. Each decision names the rule it applied — explainable and auditable by design.

It gets smarter with use

In shadow mode the agents run alongside your team. Every disagreement your reviewers resolve becomes a new rule — knowledge that compounds.

Trust is earned, per payer

Agents graduate from shadow to assist per agent and per payer, gated on measured agreement with your reviewers — not on promises.

Denials prevented, not appealed

A denial-risk predictor scores every claim before submission, with the factors and a fix-list — so problems are caught while they're still cheap.

Audit-grade by default

Every action — human or agent — lands in an immutable audit trail. SLA cockpits track the regulatory clock on every pre-auth in flight.

Built by AST

From the team that ships healthcare software for a living.

Medexa is engineered by AST — 17 years of EMR systems, clinical integrations and HIPAA-grade platforms running inside real care networks. The same production discipline, applied to our own product.

17 years in healthcare software300+ facilities running our workFHIR R4 · HL7 · payer rails
See it live

Watch Medexa carry a claim end to end.

A 30-minute walkthrough: ambient capture, live coding, the agent's drafted decision with its cited rule, and the review console your team would run. No slideware — the working product.

Request a walkthrough