Independent revenue recovery review · UAE clinics
We review claims your current process has already handled, identify what still looks recoverable, and pursue the cases worth pursuing. You pay nothing unless money arrives.
Team, provider and systems unchanged.
Denial reason codes, counts and values.
We are paid only from verified incremental collections.
You send an anonymised denial extract. We do the analysis independently. Your existing billing operation continues exactly as before.
Anonymised reason-code counts are enough to start. No patient data, no system access.
We identify what appears worth pursuing, with the evidence behind each line.
If you say yes, we pursue the selected cases and share in what is recovered.
Deliberately narrow: one payer, one specialty, one denial class, over a three to six month window. Narrow enough to measure honestly in 60 to 90 days.
The review phase runs on anonymised reason-code counts. Claim-level analysis, and the per-claim scoring described further down this page, apply to the recovery phase and require a signed data processing agreement first. You see the review before that question arises.
Scope widens to more payers, more specialties or prevention work only if the first engagement shows the economics hold.
One ledger, with your own figures. Illustrative example
| Pattern | What it looks like | Why it is missed |
|---|---|---|
| Paid, but not fullyShort-paid | Billed AED 4,800, reimbursed AED 4,100, marked complete. | The AED 700 never becomes a task, because the system has no reason to raise one. |
| Recoverable claims age outUnworked denial | Appeals worked in the order they arrived rather than by what they are worth. | High-value recoverable claims can expire while low-yield items are cleared. |
| Small differences, large totalsTariff variance | One payer reimbursing 4.8% under contracted terms on a single procedure family. | Invisible claim by claim. It only becomes a contract issue across four hundred. |
| Recoverable until the deadlineFiling window | Claims sitting a few weeks from a payer submission deadline. | Urgency is not visible in an aging bucket, and the window differs by payer. |
| The same issue, every monthRepeat rejection | One missing field or code pairing, recurring across a specialty. | Each instance is handled individually, so the pattern behind them is never surfaced. |
Your RCM team has to run the entire revenue cycle. We have one narrower job: find financially meaningful residual value across claims that have already been processed.
The examples below show the kinds of difference a second review looks for. Some compare payers, which a first pilot deliberately does not: pilot one is scoped to a single payer.
That is the entire commercial model for our first pilot engagements.
A performance fee is only honest if both sides know what “extra” means. Before any work begins we fix the baseline in writing: historical collection rate, denial and recovery rates, payer and procedure mix, the measurement window, and which claims are excluded.
We then separate genuine recovery from ordinary growth, price effects, volume effects and mix effects, so the invoice is never an argument about attribution. Where a slice of eligible claims is deliberately left unworked as a control, that is agreed up front and used to evidence the difference our work made.
The proposition
We will tell you in three weeks, using your own numbers. Either there is recoverable revenue and we go after it together, or there is not, and you have an independent answer at no cost.
Everything below is the detail behind the review. None of it is needed to decide whether to send us an export.
We screen the full claim population, look for patterns that are hard to see case by case, rank the economically meaningful opportunities, and send only the strongest cases to human review.
This describes the full method. A first pilot is narrower: one payer, one specialty, one denial class, starting from anonymised counts. Per-claim screening and ranking apply to the recovery phase, under a signed data processing agreement.
Traditional billing systems tell you what happened to a claim. We build models around a different question: given everything we know about this claim, payer and history, what should happen next, and what is it worth pursuing?
Claim-level scoring requires claim-level data. It applies once a data processing agreement is in place, not to the initial anonymised review.
| Payer | Procedure | Denial | Age | Amount | Prior result |
|---|---|---|---|---|---|
| Payer A | CPT 97110 | AUTH | 21d | 8,400 | Recovered |
| Payer A | CPT 97110 | AUTH | 38d | 6,750 | Recovered |
| Payer B | CPT 99214 | DOC | 16d | 2,900 | Written off |
| Payer A | CPT 97110 | AUTH | 19d | 9,100 | Recovered |
Modern tabular prediction methods are particularly useful here because clinic datasets are structured and often relatively small.
We benchmark methods including TabPFN, gradient-boosted trees and conventional statistical baselines against held-out outcomes. The operational model is selected on measured out-of-sample performance, not technology preference. Accuracy, calibration, false positives and realised recovery are tracked separately by payer and workflow. If a model stops adding value, it stops being used.
The models find patterns and rank opportunities. Experienced people determine whether the evidence is sufficient to act.
Because a claim can be operationally complete while the economic outcome is still suboptimal.
We do not assume your team made mistakes. We apply a second analytical pass to transactions that already moved through the normal process.
A claim is billed at AED 4,800 and reimbursed at AED 4,100. The payment posts, the claim is marked complete, and the file closes.
The remaining AED 700 never becomes a task, because nothing in the process has a reason to raise one. On a single claim that is hard to spot. Across four hundred claims with the same payer and procedure family it becomes a contract issue.
A billing operation is measured on throughput and has to serve the whole workflow. A review has one job and can spend its attention on closed and low-probability cases that would never justify a person’s time individually.
UAE law restricts the transfer, storage and processing of health data generated here to inside the country. We treat that as an architectural requirement rather than a policy paragraph.
UAE law restricts the transfer, storage and processing of health data generated here to inside the country. We treat that as an architectural requirement rather than a policy paragraph.
Databases, processing and model inference all run in a UAE region on infrastructure we control, and backups stay in-country. The scoring models run on our own hardware, so claims data is never sent to a third-party AI service. Access is read-only wherever possible and named only, with a complete audit trail of every record opened and every decision taken.
Databases, processing and model inference all run in a UAE region on infrastructure we control, and backups stay in-country. The scoring models run on our own hardware, so claims data is never sent to a third-party service. Access is read-only wherever possible and named only, with a complete audit trail of every record opened and every decision taken.
We do not start by taking over your revenue cycle. We pick one payer, one specialty or one denial category, narrow enough to measure clearly within 90 days.
Read-only. No integration project.
See exactly what appears recoverable before deciding whether to continue.
Define what counts as incremental cash before we touch anything.
Your authorised team approves filings.
No performance recovery means no performance fee.
For the review, anonymised counts are enough: denial reason code or family, how many claims sit in each, the billed or outstanding value, the date window, the payer and the specialty. That is a spreadsheet, not a patient record. No names, no Emirates IDs, no clinical notes.
No. You export, we analyse. We do not log into your EMR, your billing system or your bank. Extracts normally come out of your EMR or your ProviderLink / DHPO access, and your team stays in control of what leaves the clinic.
Then we still do the review. The audit runs on anonymised reason-code counts. Claim-level data is only needed for the recovery phase, and only under a signed data processing agreement. You can see what we found before that question ever arises.
Deliberately narrow: one payer, one specialty, one denial class, over a three to six month window. Narrow enough to measure honestly within 60 to 90 days. Scope widens only if the economics hold.
Nothing. They keep the account, the scope and the fee. We do not bid for their work, touch their workflow or report on their staff. If they are doing well, the review finds little and says so.
In writing, before any work starts. A baseline annex fixes the historical recovery rate and method, which claims are in and out of scope, how concurrent internal or incumbent effort is treated, how write-offs and timing differences are handled, and the window within which recovered cash is attributed.
Incremental recovery is then actual recovered cash on in-scope claims, less what the baseline method would have predicted. Attribution disputes are what kill arrangements like this, so the annex is locked before go-live rather than argued afterwards.
You do, always. The clinic remains the creditor and the sole recipient of every insurer payment. We never hold remittances, never touch clinic bank accounts, never net our fee against a payment. We invoice you separately, as an ordinary receivable.
Inside the UAE. Storage, processing, model inference and backups all stay in-country. Claims data is never sent to a public AI service. Access is named and read-only wherever possible, with an audit trail of every record opened.
Only with your approval, case by case. We never alter a clinical record, never invent a service and never upcode. Coding responsibility stays with your authorised staff, and nothing is filed without a person approving it.
Write-offs you have already approved in writing, eligibility failures with no resubmission rights left, anything under a fraud hold, and any change to a clinical record. Those are agreed in the scope annex, not discovered later.
Roughly three weeks from receiving an extract to handing back the review. If you then choose to proceed, the recovery phase runs 60 to 90 days from the day the baseline is locked.
You get the ledger anyway, showing what was checked and why it came back thin. That is a useful answer about your own operation, and there is no recovery fee, because there is no recovery.
Good fit if you have meaningful insured claim volume, six to twelve months of historical claim and remittance data, and an existing billing operation.