Curacel

The removal test leaves manual claims adjudication, which is the state the company was founded to replace. Models vet every claim automatically to identify fraud, waste and abuse, analysing large datasets for suspicious patterns, and process claims in real time with human involvement reserved for quality control. Risk assessment and payout decisioning run on the same layer. Handling the volumes described with the staffing available to African insurers is achievable no other way.

The oversight position is stated plainly and it is a deliberate design rather than an afterthought: claims are digitised and settled automatically with human intervention required only for quality control, so people are positioned where judgement matters rather than on every routine claim. Automation is described as operating with minimal human intervention.

Held at B because no threshold, sampling rate or escalation rule is published, so what proportion of flagged claims a person actually reviews, and whether a denial can issue without human sign off, is unknown.

Three quantified outcomes are published and, unusually, they are corroborated across independent sources rather than resting on company material alone: around 25 percent reduction in fraud, waste and abuse payouts, more than 70 percent reduction in claims cycle time, and up to ten times the claims throughput. Volume processed is stated at over 750,000 claims, which gives the fraud figures a base.

What is missing is the other side of the measurement: no detection accuracy, false positive rate or validation result appears, and for a system whose value is expressed as payouts avoided, the rate of wrongly avoided payouts is the figure a supervisor would ask for.

Three of the largest insurers operating on the continent are named as customers, alongside more than 800 hospitals across three countries at an earlier stage and expansion to more than ten markets since. Volume is stated at over 750,000 claims processed.

Three outcome figures appear consistently across independent sources rather than only in company material: fraud, waste and abuse payouts reduced by around 25 percent, claims cycles shortened by more than 70 percent, and up to ten times more claims processed. The backer set is unusual for the region, spanning a leading accelerator, two global technology companies, a card network's programme and named fintech founders as angels.

No boundary statement was located. Fraud detection improves by observing confirmed abuse across a wide base, and the customer base includes insurers competing in the same markets alongside the hospitals whose claims are being scrutinised. Nothing states whether patterns identified at one insurer inform scoring at another, whether provider level fraud histories follow a hospital across insurers, or what a customer contributes by participating.

No data protection agreement, retention schedule, subprocessor list or consent framework was located, and the holdings are among the most sensitive in this index. Health insurance claims contain diagnoses, treatments and provider records for patients across more than ten countries with differing and in several cases recently introduced data protection regimes, and the platform sits between insurers and more than 800 hospitals. Technology profiling also indicates hosting outside the region, which raises a cross border question the company does not address.

No attestation, certification, trust centre or enumerated framework was located. Major multinational insurers have completed supplier assessment before connecting claims systems to the platform, so review has occurred at a serious standard, and nothing is published for the hospitals and smaller insurers in newer markets to rely on when the platform holds patient claim records.

No regulator, statute or supervisory framework is named by the company. Independent coverage references the national insurers association and the industry vehicle insurance database, which indicates the operating context, and that is journalism rather than a disclosure. Operating across more than ten jurisdictions with differing insurance regulation and, for health claims, differing medical confidentiality rules, makes the absence of any named framework a substantive gap.

The access argument is inverted from the usual one in this index and is the sharper for it: fraudulent, wasteful and abusive claims cost African insurers billions annually, which makes them cautious and risk averse toward customers, so reducing that loss is framed as what allows insurers to cover more people rather than fewer. The embedded product extends distribution into digital services reaching populations conventional channels do not. The counterweight is unaddressed and serious.

A 25 percent reduction in payouts means claims are being refused, some proportion of them legitimate claims wrongly flagged, and in health insurance a wrongly denied claim means someone does not receive treatment. No false positive rate, appeal rate or analysis by provider or patient type is published.

No guarantee, indemnity or correction process was located, and the exposure falls on parties with no relationship to the vendor. A patient whose health claim is flagged as fraudulent, wasteful or abusive may be refused treatment cover, and a hospital flagged for a claims pattern may find its submissions scrutinised or its relationship with an insurer affected. Nothing describes whether either learns the basis, whether an appeal exists, or how a wrongly flagged claim is corrected.

No model provider, base model, hosting arrangement or subprocessor is identified. Training data is the company's own claims corpus accumulated across insurers and providers, which the company describes as tailored to the operational realities of African insurers rather than imported from elsewhere, and that is a meaningful distinction without being a provenance statement. Nothing states rights, permissions or how the corpus was assembled.

The integration achievement is the network rather than any single connection, linking insurers to hospitals, travel agencies, automobile companies and security firms through cloud based tools and interfaces, which is what allows a claim to be submitted and adjudicated in real time rather than passed between organisations on paper. The embedded product goes further, letting technology companies place insurance inside their own services. No named claims, policy administration or hospital information system appears and no developer documentation was located.

No hosting, region or residency commitment is published by the company. Technology profiling indicates servers located outside the continent, which if correct means health claims data on African patients is processed abroad, and several markets served have introduced or are introducing localisation requirements for personal and health information. That the position has to be inferred from profiling rather than read from disclosure is itself the gap.

No pricing, packaging or basis of charge was located. The business spans claims automation sold to insurers, network connectivity to providers, and an embedded distribution product sold to technology companies, which would ordinarily carry three different commercial models, and none is described. For a fraud reduction product a share of savings arrangement is common and nothing indicates whether one applies.

Coverage extends well beyond insurers to the parties around them, connecting insurance companies with primary care hospitals, travel agencies, automobile companies and security firms, and reaching technology companies through the embedded product. Lines of business span health and motor. Geographic reach is stated at more than ten countries across Africa and other emerging markets, having started from three. Held at B because market by market depth is not evidenced and the most expansive claims about organisations served come from directory listings rather than the company.