Given the festive break, this time I am writing about a fascinating presentation I attended recently, out of personal interest, on how the third most important German health insurer uses Artificial Intelligence* to optimise administrative processes. Every HR function can be seen as a tremendous administrative processing machine, so the subject remains in my favourite remit of HR Analytics. In this zero-growth environment the Western world is experiencing, competition pushes companies to drastically optimise administrative processes. In a completely stagnant health insurance market which grows only at a rate of 1.1% per year, Allianz is acquiring new business at a rate of 5.5% per year. Its operating profit grows at a rate of 3% per year. Allianz Health is achieving this by constantly working on simplifying its product design, streamlining its cost base and by improving the customer experience.
To give you an idea of the challenge I am talking about please note that the Allianz logistic centre receives 53.000 health insurance claims every day or close to 22 million claims per year. In weight this corresponds to more than 95 tons per year. 750 employees do nothing else than claims management in a highly regulated environment. To do a hip replacement, I was told, there is a 120-pages guideline to respect, also to avoid any potential fraudulent activities by doctors and patients.
Of course, the solution is aiming for a full digitalisation of the process. Currently 10% of the claims are submitted via an app and 5% are submitted via an online upload. These submission channels are fast growing, bearing in mind that there is a limit on how much technology older policy holders can cope with. It is also well known that those are the ones more likely to submit health insurance claims. Once the claims are inserted in the health insurance processing system, there is an automated routing system in place to send the claims to the right internal administrator to process them. An online tracking tool and a “best of” bonus check is presented to the customer, which as far as I understand, is a refund option offered to customers based on their healthiness prospects (but is retroactive too?). Customers seem very happy giving very high net promoter scores in their feedbacks. The turnaround in getting the health insurance reimbursement, I have been told, is very fast.
The main business issues of handling all these claims in the digital world are the following two and this is where artificial intelligence comes into play:
- Identifying the complex aspects of the claim to avoid staff to go through the claim item by item.
- Inefficient allocation of the time that administrative staff spends between “low value” and “high value” claims, which is often based on rules of thumb.
The solution is a spline based regression model which covers over 370 individual input variables. The solution furthermore foresees an instant self-learning system which constantly improves the prediction model.
Bespoke regression models are needed to do all this. Allianz has been collecting data extremely early, i.e. for the last 20 years, which required a massive investments in storing them, but now the company can utilise these to forecast disease progression for example. The company is on track to have by 2018 a fully automated claims system that covers 85% of the claims representing a huge efficiency gain. Even complicated medical cases can be settled error free. The speed and the transparency at which the claims are dealt with increases the customer satisfaction.
Personally, I find all this very fascinating especially knowing that open source software like R is being used. Personally, I am not too concerned on the ongoing debate about the relevance and validity of statistical approaches of Artificial Intelligence.
* Wikipedia’s definition of Artificial Intelligence: a flexible rational agent that perceives its environment and takes actions that maximise its chance of success at some goal.