Data and A.I. in Insurance — An Interview with Dr. Antje Fitzner
I met Dr. Antje Fitzner, Data Scientist at Eucon Digital, at the Big Data & Analytics conference in Leipzig last month. Eucon was founded 20 years ago and is one of Germany’s leaders in the application of A.I. based technology in digital transformation projects and process automation within the automotive, insurance and real estate sectors. The company’s footprint spans 40 countries, with 400 employees, and just this year they’ve won the Digital Leader Award for their Smart Claims project in the insurance sector.
We sat down and spoke about her views on A.I. in claims management, implementation challenges, image recognition and more.
The interview has been lightly edited for clarity.
The combination of human and artificial intelligence is really where the future lies.
Albert: Taking a top level view, what impact can Artificial Intelligence (A.I.) have on the insurance sector?
Antje: A.I. will fundamentally change the entire customer experience in the insurance industry. The combination of human and artificial intelligence is really where the future lies. For instance, in terms of workflow and processes, A.I. will enable much faster processing of large amounts of data, which will work almost completely automatic. People will intervene in these processes, of course, but more in a controlling function, and in order to specify the underlying rules. Along those lines, service providers and subject matter experts can then be involved much more intelligently and assigned automatically to more complex cases.
For the first time, insurers will be enabled to manage claims much more precisely and also avoid unnecessary steps along the whole process, all of which is fueled by A.I.. That would also include the early recognition of claims which do not need further analysis, but also of those that may be fraudulent.
Albert: If we take a deeper look at claims management, to what extent is A.I. able to improve this historically relatively manual process today?
Antje: Within claims and fraud management, the usage of A.I. based technology is especially relevant, because apart from standardised processes, you’re also able to draw upon very large amounts of available data. Irrespective of the immense processing power of these systems, people will still be needed. For instance, complex cases will still need to be reserved for highly skilled claims experts. However, the share of administrative tasks will decrease gradually. Ultimately, A.I. allows insurers to execute existing processes faster and with a higher degree of reliability, which leads to a reduction in costs and improved margins. Most importantly, it frees up resources and improve processes in a way so that the customer relationship can become the focus point again.
A.I. doesn’t have anything to do with SciFi or Superman.
Albert: The projections of how A.I. will affect business are often difficult to imagine. What do you think will be the impact on, for instance, reducing the mountains of unstructured claims documents insurers deal with daily?
Antje: At Eucon we believe that in the future only about ten percent of documents will need to be checked by a person. With the remaining 90 percent, it will be possible to automate end-to-end processes to an extent that no human involvement will be necessary when it comes to checking claims documents by hand. The combination of human and artificial intelligence may be reminiscent of superhero comics, where some superhuman being is solving problems with superhuman powers. But A.I. does not have anything to do with SciFi or Superman — it is really all about the intelligent automation of processes with the help of data, technology and knowledge.
Irrespective of the immense processing power of these systems, though, people will still be needed.
Albert: Where are the limitations and pain points when it comes to the effective implementation of A.I. in claims management?
Antje: The large amount of unstructured data in it of itself is a big challenge. Apart from the fact that very few documents which we receive are still hand written, the majority are PDFs. Not everything can always be read out with Optical Character Recognition (OCR), and human effort is still necessary to control the results in order to avoid errors, and to structure and label the data. Eucon is further automating processes using intelligent prediction models in order to reduce human interference and accelerate claims handling. Naturally, this process will improve as times goes on. Another factor is the quality of the data we receive. Since the quality of every model is dependent on the quality of the underlying data, our work is highly dependent on solid inputs.
Albert: Where do we currently stand with image recognition technology, especially as it pertains to analysing insurance claims?
Antje: Image recognition and analysis is a big challenge! If you take car insurance claims and the corresponding type of damages as an example, you are faced with several challenges. Apart from the fact that you may have access to images, many are unclassified (note: lacking proper labels of every detail), and so it is very difficult to interpret and dissect them properly. You may have reflections or dirt on the actual object and in addition image quality can also be an issue, especially if they’ve been scanned. Furthermore, for depth recognition you’d need to take at least two images from different angles. We’re currently looking for technology partners who are “best in class” in this specific field, but so far we’ve not been able to find anyone who has a solution in place which can yield good results.
I hope you have enjoyed this interview. For questions or comments, or if you want to receive updates, get in touch via Email info@connectingthedots.cx or Twitter @cngthedots.
