Health insurance · UAE
Pre-authorisation document intake for TPAs and insurers under DHA Directive PD-05-2025. Compliance with a clock you are currently missing, not an efficiency story.
DHA Directive PD-05-2025. Six hours on elective outpatient pre-authorisations. Twenty-four hours on elective inpatient. Emergency immediately, with written confirmation inside 24 hours. Payment posted within 45 days for first adjudication, 30 for resubmission.
The clock starts when the pack arrives. If the pack is still being opened, classified and keyed when the six hours elapse, the breach is already on the file.
You have six hours on elective outpatient pre-authorisations. How many did you breach last month, and what did each one cost you?
Pre-authorisation document intake. Analyse, identify, classify and map the pack so a human can decide inside the window. Arabic-native, including handwritten, with per-field confidence and citation back to the source region. Originals preserved. Every transform logged.
The blocker is not a general preference for privacy. Pre-authorisation packs contain clinical detail, identities and policy data that, for this buyer, cannot be sent to a public parsing API or a cloud agent workspace. PD-05-2025 still runs while the file sits in a vendor region. A metered cloud parser that leaves the country is not a delayed decision. It is a non-starter before the six hours are counted.
Private AI Foundation installs once in the TPA's own environment: air-gapped, customer tenancy in-region, or bare metal. DocxIntel attaches as the intake application. Packs never leave the perimeter. The Model Gateway, policy, DLP and audit sit in the path of every document. The TPA holds the keys. Open weights sit on disk they control.
Operators keep their queues and their core. Intake writes a structured pack and a citation trail. A human still authorises. AgentLab is not part of this first deal.
Receive the claim or pre-auth pack as PDF, image or mixed formats. Classify documents, extract billed items, diagnoses, member and provider fields, and map to the TPA or insurer schema. Originals preserved. Per-field confidence and citation. A human decides. The system does not pay the claim.
Match a resubmission to the prior file, highlight what changed, extract the delta, and queue against the shorter resubmission clock. Avoid re-keying the entire pack from scratch.
Route low-confidence fields, missing pages and policy mismatches to the right queue with a reason code. Assist the analyst. Do not silent-post to the core. AgentLab is optional after Foundation is live. Model Gateway logs every prompt and tool call.
Flag duplicate packs, inconsistent demographics, unusual billing patterns and citation gaps for human review. Assistive only. Not an automatic fraud conviction. The audit trail stays inside the perimeter.
Compare extracted fields to the HIS, TPA or core system of record. Surface mismatches before payment posting. The source of truth remains the customer system. BizfyLabs LLC does not become the ledger.
These controls run on Private AI Foundation for every document ask. AgentLab, if used later, shares the same path.
Thirty to forty-five days. Fixed fee. In the customer environment. One document type. Agreed real volume. Success criteria and conversion price agreed in writing before starting. If it misses the threshold, the customer keeps the report and owes nothing further.
This is not a free trial and not a sandbox that pretends to be residency. The cable can be pulled. The packs are yours.
A paid proof of value, 30 to 45 days, on one document type in your own environment. Success criteria agreed in writing before we begin.
In your office, on your documents, with the cable pulled out. No competitor's sales engineer can do this.
