A roadmap for cultivating a data-driven culture
In my previous post, I suggested that it was possible to provide a road map that would help with the introduction of artificial intelligence, advanced analytics and machine learning into insurance companies. This post outlines the process.
The first area to address is applications where the adoption of analytics will have an immediate impact on cost reduction and efficiency. The obvious point is process automation. In many insurance companies, the first projects involving advanced analytics and machine learning models are the digitisation and optimisation of processes. These projects tend to be rapid, with very fast returns on investment. They lend themselves to building business cases and assessing the benefits in advance. These applications often require work to harmonise and structure data sources and to ensure that data transformation procedures are fit for purpose. It is also necessary to monitor and improve data quality.
Examples of these types of applications include the automatic analysis of internal help desk activity, call centre notes and statistics, customer complaints, and hospital reports. Natural language processing techniques are used to automate categorisation and responses because the documents concerned are usually written in an unstructured way, in the form of free text. Analysis identifies the links and correlations with other structured data. For example, analysing customer complaints and linking them to data about sales channels can identify whether there are problems in a certain channel or territory and allow prompt intervention if necessary. The major benefits of applying these techniques include reduction of labour costs with low added value and some fundamental insights into organisational inefficiencies and the processes that cause customer dissatisfaction.
The extension of analytical applications to performance analysis will reduce costs related to claims and operational management. Accurate cost analysis using statistical methods enables organisations to understand the areas of greatest impact and need and, therefore, the targets for possible improvements. This is a step beyond process optimisation because it requires data – or, better, quality information – to support analysis.
Statistical models can be used to analyse and prioritise the alerts for potentially fraudulent claims, increasing their quality and reducing false positives. This allows investigators to focus on the most critical cases and the ones most likely to be recoverable.
