Skip to content

Insurance: Resemblance

The Resemblance product is built on the Luther Platform for Deep Process Automation, with AIA and AXA to deploy TOKO at scale and automate the complex processes of digital asset issuance.

Insurance: Resemblance


In most industries there are similar or twin documents or entities. Work twice on the same customer profile or open and follow similar claims from two different insurers. In this second example, correspondence not only leads to increased operating costs but can be identified as a clear sign of Fraud.

Fraud and abuse occur at many points in the insurance process, and the costs of such issuance have been estimated by the Coalition Against The Fraud (CAIF) at around $80 billion annually, half due to similar claims. The result is not only a higher cost of care but above all higher insurance premiums for all consumers.

What makes similar claims difficult to detect is the inability of insurers to legally share business data with competitors or others and to disclose personally identifiable information without invading privacy.

Resemblance Business Stats

Process Complexity 

The absence of a collaborative and automated product for identifying similar claims with data privacy protection has been a real barrier for AXA and AIA.

This is why they have partnered with Luther Systems to use its unique Deep Process Automation platform to create this product for secure and privacy-wise collaboration.

This network of participants requires the ability to detect matches on documents across each participant’s private repository. In the case of multiple insurance providers, the need is to detect fraud by determining if the same or a similar claim was filed with another insurer.

In many cases, the participants desire to keep the contents of their repository private from the other participants.

Additionally, the document checked against these private repositories should also be kept private. The documents themselves sometimes contain Personal Identifiable Information (PII) and are subject to data storage and processing requirements to ensure user privacy (compliance, regulation, sensitive data).

The documents across the repositories may have slight differences yet still be considered similar to another, which requires systems to include a document similarity metric that detects similar documents and not only exact matches.

To avoid deception, these applications must prevent participants from discovering documents that they do not own. In particular, the system provides safeguards to ensure that querying participants can only query for authorized documents.

Process Problems

Process problems

Business problems

Business problem


The problem to solve was further far beyond the scope of any technical solutions available on the market. This is why AXA and AIA have partnered with Luther Systems to build the first world's international common claims network.

This solution, best for insurers and best for customers, required automation that could handle the inherent complexity of a multi-insurance process involving hundreds of different Insurance companies and processes. It is equally essential that the new network could be built scalable to easily adapt to an always broader audience.

Resemble is a technology developed for Claims fraud. It allows the match detection between documents using a novel “Multi-signature hashing” method, with a resulting tool for collaboration privacy and security proof.

The Resemble (matches detection) product can:

  1. Identify similar claims
  2. Ensure data privacy  between all the participants of the network
  3. Ensure data storage on the Blockchain is GDPR Compliant
  4. It can identify similar claims in less than 1 sec
  5. Create a network that is easily scalable to add new insurance companies

All the data are elaborated in the data storage of each insurance company, then encrypted and only the hashes are shared in the semi-trusted network, where all the companies are.

At this point, the Smart contract will validate the claim by searching for a similar claim, and send an alert to the insurance companies involved in the detection of a similar claims

The image below, by way of illustration, is a representation of the network and the interactions. Insurance company A inserts a claim. The claim inserted by company A is similar to one of the claims handled by company B. In the case of a match, all of the companies with the similar claim receive an alert.



The solution proposed and the developed product address the problem with a scalable, flexible, and privacy-protected solution. Furthermore, the multi-signature hashing function, the strong hash function selected, and the distribution of data in several structures leave the information safe and not available to any malicious attacker.

Commercial results

  • Detection Speed - in less than 1 second
  • $2.7 Million - Cost reductions for Fraud
  • 6% payout reduction incurred due to fraud
  • ROI estimated 10X

Product results

The solution is valuable for:

  • the extensive automation that allows the process to complete a check of matches in less than a second
  • the document similarity metric that detects similar documents, which allows detections of all different similarities not only the exact match
  • privacy protection as a focal point. The solution enables a safe and privacy-protected environment where collaboration has never been easier
  • the scalability, which means multiple insurance companies can join the network and keep their processes and data private.

Technical results

  • The solution leverages Luthers scalable and efficient platform to detect similar claims
  • detect document similarity in less than 1 second
  • keep documents active for a long period of time
  • 1 billion of active documents in the joint repository
  • 0.1% false positive detection rate
  • Security & cost to decrypt a document on the repository

We use cookies to give you the best experience of using this website. By continuing to use this site, you accept our use of cookies. Please read our Cookie Policy for more information.