Helvetia has set itself the goal of significantly reducing payouts for wrongly submitted insurance claims in the coming years. Artificial intelligence (AI) is meant to help identify even complicated cases of suspected fraud and contribute to combating insurance abuse (BVM).
A preliminary study launched in 2020 defined the future environment for combating insurance fraud (BVM) and worked out suitable solution scenarios combining standard software with the application of in-house capabilities in data and analytics.
Challenge
The market for fraud detection software is broad. Alongside sector-neutral products, providers specialising in insurance and using AI are increasingly gaining a strong competitive position.
Yet the maturity and benefit of advertised AI components are not obvious at first glance. When making an investment decision, the performance of AI-based products needs to be clearly weighed against conventional, rule-based products.
Claims cases and the information that can be gathered about them vary widely. The effectiveness of a fraud solution grows with its ability to make diverse information usable for assessment. The ability to enrich individual cases with additional information from internal and external data sources is therefore a key factor in achieving expected success rates, and thus in evaluating purchased software.
This gave rise to the following areas of focus for the preliminary study:
Fraud detection realises its potential precisely when it is available to the case handler early in the recording process, allowing it to be built into the natural workflow. This close link to the process demands flexible integration options for the chosen solution, along with a high level of transparency around the indicators used to assess fraud.
The wide range of information required for fraud detection places high demands on data provision from the core insurance processes and is a major factor in assessing technical and time feasibility.
Wide variance and low transparency in the advertised ML/AI-based functionality of fraud detection offerings call for intensive product analysis and a rigorous selection process with clear criteria and metrics, to create as broad a basis as possible for comparing the performance of the solutions.
Approach
The preliminary study applied the KDX.Partners approach model for software selection. The final candidates from the selection process were tested against Helvetia's business case scenarios. Alongside the procurement process with the chosen solution provider, a multi-layered plan was developed covering several accompanying transformation projects.
The selection process ran through several rounds, starting with an initial market study, moving through a Request for Information (RFI) and on to a Request for Proposal (RFP). For the RFP, around 130 functional and non-functional criteria were compiled and assessed across several evaluation dimensions. Particular attention was given to working out the benefit and approach of the ML components used, through criteria, interviews and reference calls.
Helvetia's Advanced Analytics team tested the option of a self-developed fraud detection solution built on Helvetia's analytics infrastructure as part of a proof of concept. KDX.Partners evaluated the results of this in-house approach, standardised them for the software selection, and factored them into the decision-making process.
Based on the results of the RFP and the proof of concept, several business case scenarios were calculated to reflect the characteristics of the BUY and MAKE variants, and reviewed by the relevant architecture and leadership bodies.
An implementation roadmap was drawn up for the chosen scenario. Particular attention during planning was paid to analysing dependencies on several extensive transformation initiatives affecting Helvetia's underlying application and process landscape for fraud detection.
Throughout Helvetia's procurement process, the negotiation through to contract signing was moderated in substance, to secure financial and operational success factors for the longer-term collaboration between the software provider and Helvetia's operating organisation.
Result
KDX.Partners guided the selection and procurement process as well as the development of the project roadmap towards a transparent investment decision, through continuous facilitation, clear structure and the involvement of all relevant business and IT areas.
The result: from a wide range of very different options, the team identified the most future-proof solution and reached broad consensus to approve the project roadmap. It covers all content and timeline requirements of the business case and integrates the project with related transformation initiatives.
KDX.Partners managed to structure our feasibility study despite a complex starting point and involve all necessary parties in the process. Thanks to the continuous facilitation, we were able to reach a broadly supported and comprehensible decision in the end.
