MEDKONG
Our manifestoAKOSBuilt by AKOSBuilt on Palantir Foundry

One rulebook. Checked from both sides.

MEDKONG deploys Agents on both sides of the table, working from the same rulebook.

We built the reviewer’s Agent first: seven gates, with the policy cited behind every finding. Provider Agents clear the same gates before they submit. Every decision feeds back into the rulebook. Decided once, and it holds.

Read the principles
What MEDKONG is

Agents in your stack. Powered by the same rulebook.

For providers and for the reviewers who decide their cases. Same rules on both sides, deployed into each side’s existing stack.

Provider side · 8 AgentsReviewer side · MAC review Agent · 7 gates
Eligibility & BenefitsPrior AuthorizationCharge CaptureCoding ReviewClaim QA & SubmissionDenials & AppealsPayment PostingAR Follow-upCompletenessEligibilityProgram scopeGoverning policyDiagnosis coverageModifiers & attestationDocumentationOne rulebookCoverage policy · payer rules
Provider caseChecked against the rulebookDecision back to the provider
Provider side · 8 Agents
Eligibility & BenefitsPrior AuthorizationCharge CaptureCoding ReviewClaim QA & SubmissionDenials & AppealsPayment PostingAR Follow-up

One rulebook

Coverage policy · payer rules

Reviewer side · MAC review Agent · 7 gates
CompletenessEligibilityProgram scopeGoverning policyDiagnosis coverageModifiers & attestationDocumentation
What it does

MEDKONG is a set of deployable Agents for critical revenue cycle workflows. AKOS adds them to the stack you already run, and you own them. The same record, Agents and governance serve two desks: the provider preparing and defending a claim, and the Medicare Administrative Contractor reviewing a prior authorization request.

  • Eight provider Agents, from eligibility to AR follow-up
  • A MAC review Agent for prior authorization: seven gates and a decision
  • Deployed into your stack, never rented. Your data stays in your systems
  • The Agent prepares the work; a named person signs every decision
How it’s built

Each Agent is infrastructure as code, built by AKOS on Palantir Foundry. It ships as a versioned package into your environment, and both sides execute against one versioned rulebook.

Delivery
Each Agent as a versioned module: ontology object types, governed actions, functions, automations, operator apps
Runtime
Your Foundry environment. No MEDKONG tenant holds your records
Record
Patients, encounters, codes, authorizations, claims and accounts as objects every Agent reads and writes
Rulebook
LCDs, NCDs and coverage articles resolved by category, state and jurisdiction, plus your own payer policies
Harnesses
Provider Agents and the MAC review Agent run the same checks against the same rulebook
Controls
Governed actions, not direct writes. A rerun never overwrites a verified human decision
What we believe

Four principles.

What we believe, what the industry says instead, and how MEDKONG puts it into the software.

Principle 01

One rulebook, both sides of the table.

We believe

A denial is not a billing error. It’s two parties reading the same policy differently. It ends only when both sides work from the same rules.

The industry says

Shift left. Catch errors earlier. Appeal faster.

How MEDKONG applies it

MEDKONG runs on both sides of prior authorization. The reviewer’s Agent checks every request through seven gates; the provider Agent builds the packet against those same gates before submission. Same LCD, NCD and rule-group logic, so the outcome is predictable before the request is sent.

Principle 02

Defensible decisions, not automated tasks.

We believe

A decision nobody can explain isn’t finished; it’s an appeal waiting to happen. What matters is the decision that holds up, not how many tasks got automated.

The industry says

Automate more. Measure the automation rate.

How MEDKONG applies it

Every automated action carries its inputs, its reasoning and a named reviewer. Findings quote the evidence and cite the policy text. The Agent proposes, a person records the verdict, and a verified decision can’t be overwritten by a later automated refresh.

Principle 03

Policy is a system, not tribal knowledge.

We believe

Payer policy changes constantly. When it lives in people’s heads, training decks or a physician champion’s memory, it fades the day that person leaves.

The industry says

Train the clinicians. Build denial-prevention teams. Hold monthly reviews.

How MEDKONG applies it

MEDKONG resolves each case to its governing coverage article, LCD and NCD by category, state and MAC jurisdiction. When a policy updates, the change reaches every affected case, and every verdict feeds back into the rules instead of someone’s memory.

Principle 04

Own your decision making. Don’t rent it.

We believe

Your denial patterns and review history are your most valuable operating data. Outsourcing the decision means paying someone else to learn from it.

The industry says

Outsource follow-up. Buy SaaS. Hold vendors to a scorecard.

How MEDKONG applies it

MEDKONG is deployed, not hosted. Each Agent ships as versioned infrastructure into your own environment on Palantir Foundry, inside the stack you already run. You run it, govern it and extend it, so what the Agents learn stays with you.

The relay problem

The hops cost days. The loops cost months.

Every prior authorization passes through a relay, AI or not. The fixed hops are annoying but finite. What turns days into months is going around again.

The hops
Provider submitsMAC forwardsReviewer decidesUTN issued

Fixed and finite. A few days per pass.

The loops
Incomplete submissionMiscategorized requestNon-affirmation↺ Resubmit, start over

A non-affirmation can’t be appealed. Every miss is another full trip through the relay.

What we do

We built the reviewer’s side first.

MEDKONG’s reviewer Agent checks every request through the seven gates a reviewer is required to check. We know the gates because we built them from the reviewer’s side.

Providers scale up to match that rulebook. Provider Agents check your case against the same seven gates before it leaves your building.

The seven gates · both sides
  1. 01Completeness
  2. 02Eligibility
  3. 03Program scope
  4. 04Governing policy
  5. 05Diagnosis coverage
  6. 06Modifiers & attestation
  7. 07Documentation
The provider promise

Submit once.

A case that arrives right goes through the relay one time.

✓ 01

Complete

Every required document and field present before submission, checked against the completeness gate.

✓ 02

Correctly categorized

The request lands in the right program and service category, so it’s reviewed against the right policy.

✓ 03

Policy cited

The governing LCD or NCD named, and the evidence mapped to what it requires.

How it’s measured

First-pass affirmation rate

Affirmed on the first submission, with nothing required from the MAC.

Where it leads
Clean first passesHigh affirmation rateGold card eligibilityOut of the relay

WISeR offers gold card exemptions to providers who meet affirmation-rate thresholds. A gold-carded provider skips prior authorization for those services.

What we don’t believe

The consensus, and our view.

The consensusOur view
Faster AI review fixes turnaround.The relay isn’t the delay. The loops are.
Keep pace with payer AI.Bots racing bots is a treadmill. Make fewer decisions worth disputing.
Free staff for high-dollar appeals.An appeal is the most expensive way to be right.
Physician champions fix documentation.That’s tribal knowledge with a better title.
Automation rate is the KPI.First-pass affirmation rate is. An overturned decision was never automated, only deferred.

Decide once, correctly, in a way both sides can check.

Deployable Agents for the critical workflows in your stack. The Agent prepares the case, a person signs it, and the rules are the same on both sides of the table.