AI and machine learning in a GDPR environment.

GDPR, the four-letter acronym for the EU’s upcoming data protection regime, will please data privacy advocates, and need not be the bitter pill insurers fear. In the insurance industry, rich data is the lifeblood of the business. It is an industry built on the fundamentals of behavioural analysis. As a result, the EU directive requiring insurers to anonymise data that is not used, and giving people the “right to data erasure” could pose challenges to insurers’ effective data use in areas like sales, underwriting and claims.The insurance industry has invested heavily in AI modelling that optimises many elements of the insurance pricing, sales, conversion and retention process. Optalitix have implemented AI models for clients that do everything from marketing channel recommendation, identification and product selection through to conversion rate optimisation and risk management. The challenge we needed to address was whether anonymisation of data (as prescribed by GDPR) would break the AI models we had in place, and more importantly compromise the models’ ability to continue learning.Trend analysis is complicated by the fact that the new GDPR initiative has set out new guidelines surrounding the term of data retention, and only data that is actively being used can be retained. Older data that is no longer expected to be part of active data requirements needs to be deleted or anonymised.At Optalitix, our fusion of actuarial thinking, technological expertise and a deep understanding of the financial sector has allowed us to find innovative solutions for this upcoming change. It means that the start of the regulations on the 25th of May 2018 need not require a return to the guess work surrounding data use in business acquisition, retention and underwriting in the insurance space.Most statistical algorithms don’t require specific information about an individual, and they don’t require deep personal data. With more generic data, Optalitix is able to re-categorise client data and create fresh, anonymised data sets prior to discarding the data for GDPR purposes. The data used in the models is anonymous, but still provide valuable insight and continue to make efficient and accurate recommendations.Working with a leading insurer with a business model dependent on rich data to not only underwrite, but also to promote and maintain their loyalty and rewards program, meant Optalitix was able to test and prove that these new data sets would be as accurate and efficient as they were before GDPR. Optalitix built an anonymisation engine into the client database which automatically deletes data once it had reached a certain age or expired but most importantly we maintain fully anonymised data sets that enable algorithms to be trained efficiently and effectively.Optalitix were able to prove that insurers can operate in a GDPR compliant environment and not lose out on business intelligence due to data anonymisation.