MedKong — modular AI for the healthcare revenue cycle
MEDKONG
MEDKONG is a kit of deployable AI and workflow modules for healthcare revenue cycle operations, built by AKOS on Palantir Foundry.
Install the module that fixes your worst queue. Expand when it earns it. No rip and replace.
Representative outcomes. Actual results vary by implementation and provider workflow.
Pre-service to post-service. Deploy one, or the whole cycle.
Each module owns a stage of the revenue cycle end to end, and they all share one model and one integration layer. Start with the queue bleeding the most today; the next one drops in without a re-integration.
Patient Access & Eligibility
Verifies coverage, plan rules, and patient responsibility before the visit.
96% verified pre-visit
Prior Authorization
Decides if auth is needed, assembles the clinical packet, submits, and tracks the deadline. Staff review exceptions, not every case.
SLA tracked to the minute
Charge Capture Support
Reconciles orders, notes, and administered items against posted charges while the encounter is still fresh.
~$212K recovered / month
Coding & Documentation Review
Suggests codes linked to the source note and flags specificity and laterality gaps for human review.
Every suggestion cited to source
Claim QA & Submission
Scrubs claims against your real denial history; held claims name the reason and the reviewer.
Held before the payer denies
Payment Posting & Reconciliation
Matches remittances, explains contract variances, and routes exceptions for review.
88% auto-posted
Denials & Appeals
Classifies denials by root cause and drafts evidence-backed appeal letters.
Tracked by payer & reason code
Reporting, Audit & Oversight
Records the inputs, rationale, and reviewer behind every automated action.
100% of actions traced
Built on Palantir Foundry by AKOS.
Revenue cycle work is scattered across systems that were never built to cooperate. MEDKONG puts one governed model underneath them, then works the queues where margin leaks — with people in the loop and an audit trail on every action.
One shared model
Patients, encounters, claims, payers, and accounts — modeled once on Palantir Foundry and shared by every module. Add a module and it plugs into the same graph, no re-integration.
Human-in-the-loop
Agents assemble and check the busywork. Your staff work the exceptions, not the whole queue, and nothing files without a path to review.
Auditable by default
Every automated action carries its inputs, its rationale, and its reviewer. When a payer asks how a claim was worked, the trail is already there.
No rip-and-replace
Your EHR and PM systems stay. MEDKONG reads and writes back through HL7 / FHIR and 837 / 835 connectors, sitting over your stack — not in place of it.
MEDKONG is one of the systems we build. If your operation runs on queues, handoffs, and systems that don't talk to each other, that's our kind of problem — in healthcare or anywhere else.