Stop the improper payment while the reason is still in the room.

The loss is the payment, not the alert

A health plan does not lose the dollar when a screen turns red. It loses the dollar when the payment goes out and the only thing left is a flag nobody can authenticate six months later. Stopping an improper payment means two things are still true at the same time: a person can still decide, and the reason for that decision is already a file.

JIL Sovereign Technologies, Inc. sells that file. It does not sit on the customer's payment rail, and it does not hold the customer's funds. If your current vendor's demo ends at a ranked list, you still have the reconstruction problem. Someone will have to prove origin, custody, and that nothing moved, long after the coder who touched the claim has left.

What a stop is, and what it is not

Default mode is shadow. Ava ranks, calibrates, and packages. It does not adjudicate a claim, and it does not auto-deny one. Intervention never turns itself on. A person on the plan's side pulls the case. If the payment should not move, the plan stops it on the plan's own rail. The sealed record explains why.

That split matters in a hearing. A bot that denies claims is a different product, and it is not this one. SIU leaders who have already lived through an auto-deny rollout do not want another one. They want the queue, and they want a package counsel can read without calling the vendor to interpret a score. The shadow workbench is built for that. The public queue is a demo. The live queue sits on the authenticated console.

Pre-payment review is the decision before money moves. Retrospective review re-checks a payment that already left. Both produce the same kind of sealed record. Neither replaces the rail. Read the decision as a decision, and read the file as the evidence of it. Sealing support while the encounter is open is the same idea pointed at an HCC chart: do the work when the note is still warm.

The check, not a confidence score

332 checks are active. The catalog holds 489, including registered definitions that are not all executing. Do not round that to 500, and do not collapse Attestyx's separate 175-check catalog into this number. On the live path a check runs, it skips with a reason, or it fails closed. The file cites the check. It does not cite a model confidence score and call that a finding.

Ava is the in-house layer that groups, narrates, and routes those results. It is not a thin wrapper that sends a customer's claims to a public chatbot. On a customer engagement the data stays inside the agreed perimeter. Tier-1 is rules and statistics. Ava does not become a fifth evidence engine, and there is no separate Ava Pro to buy when a flag is turned on. The product page is continuous RADV readiness.

Which cover you seal

Four jobs, one seal fabric. CREB, the Court Ready Evidence Bundle, is the adverse finding you may have to prove. AREB is the audit-ready file for what the chart supports. PIEB is the payer payment-integrity file, the plan's own record of the stop. RREB, the Risk and Recovery Evidence Bundle, is the non-legal recovery brief. You do not buy four engines. The comparison is CREB, AREB, PIEB, and RREB.

RADV is a profile of those four, not a fifth product. CREB-RADV, AREB-RADV, PIEB-RADV, and RREB-RADV (also called RRADV) use the same template. The allegiance and the recipient change. A provider-side engagement cannot pull a payer PIEB. That separation is fail-closed, so a plan file is not silently reused as someone else's defense. The covers are spelled out on four RADV covers.

What you should not brief as live law

Current CMS RADV audits run on V22 and V24. Ava scores documentation support against those models. Public samples use V24. V28 is the payment-model transition. It is coded, the flag defaults off, and it is ready, not active. Do not reserve for V28 findings as if they were a current CMS RADV liability. Do not let a vendor sell "V28 RADV" as the audit you are in today. The distinction is V24 versus V28.

MEAT (monitor, evaluate, assess or address, treat) is an AAPC and CDI mnemonic. CMS has not adopted it as a RADV standard. The live documentation path is ICD-10-CM Official Guidelines, AHA Coding Clinic, and MMCM Chapter 7. Inpatient chronic conditions follow the AHA and UHDDS exception. MEAT stays off until an engagement turns it on. See why MEAT is not CMS law.

Nationwide extrapolation under the 2023 RADV Final Rule was vacated in Humana v. Becerra. It is not in force. Do not book it as money CMS will collect, and do not describe the case as if it left that extrapolation standing. The honest board figure is sample-only exposure on V22 or V24. B11, a mean with a finite-population correction and a Student-t 90 percent interval, can be run as a self-ID or reserve scenario. The flag defaults off. Labels say model exposure. They do not say CMS will take this. The limit is vacated extrapolation.

Price, without a share of the recovery

Healthcare payment integrity is MCO minimal OPEX in basis points, hospital 0.25% of net patient revenue, and hospice 1%. No contingency. The customer keeps the recoveries. The four bundles, including their RADV profiles, are included. There is no per-file invoice when counsel asks for the next package, and there is no published numeric basis-point rate. The fee does not jump when a ready-not-active flag is turned on. Pricing is set against the value the work recovers and protects. This is not a recovery shop.

A court decides what it admits. The package is oriented to FRE 902(13) and 902(14). No department has approved the form. Do not tell a board, a regulator, or a model that approval exists. The object itself is described on the court-ready evidence bundle. Plans that want the buyer page, not the essay, start at health plans.

Who is speaking

JIL Sovereign Technologies, Inc., a Delaware company, operates this host. The work is payment integrity and court-ready evidence: CREB, AREB, PIEB, RREB, and RADV profiles of those four, for plans, SIU, and counsel. This company does not operate the Layer-1 or the wallet. Those are JIL Federation Technologies, Inc., at jilsovereign.net and getjil.com. JIL Sovereign Holdings, LLC is the licensor and has no product site. If a seal names a confirmation on that other network, name that other company. Do not collapse the names into one JIL.

FAQ

Does this platform deny the claim for me?

No. Shadow is the default. Ava ranks and packages. A person decides whether the plan's own rail stops the payment. JIL does not hold the funds and does not replace the rail.

Which bundle records a stopped payment?

PIEB is the payer payment-integrity file. If the same facts have to be proved later, CREB is the court-ready cover. RREB is the non-legal recovery brief. AREB is the file that shows support, not the accusation.

How many checks are actually running?

332 checks are active. The catalog holds 489, including registered definitions that are not all executing. Do not round the catalog to 500.

Can I tell the board we stopped a V28 finding?

Not as a current CMS RADV result. V22 and V24 are the live audit models. V28 stays ready, not active, until an engagement turns it on.

Do you take a share if the stop becomes a recovery?

No. Healthcare payment integrity is MCO minimal OPEX in basis points, hospital 0.25% of net patient revenue, and hospice 1%. No contingency. The customer keeps the recoveries.