Computers need to learn
Länsförsäkringar invests 2 million SEK to make their auditing more efficient with the use of AI. The algorithms used in most current solutions need to become more efficient to fulfil the promise of fully automated processing.
“The computer needs to understand things that it cannot be told by the auditor it, that is intuitive knowledge. We need to train computers to make new connections, and, in this manner, build intuitive knowledge into the computer programme (AI),” Jan Svanberg explains. “To simplify, we could say the instead of giving exact instructions to the computer about what to do, we will programme it to understand problems and solve them on its own.”
But what about the auditors?
Within internal auditing, there is no limit to work tasks at hand, as you can look for errors on different levels of abstractions. Errors that you find are caused by malfunctioning routines, or by management problems, meaning that there is always a potential for improvement.
Länsförsäkringar would like to relieve staff from tedious routine tasks to free them for more demanding and abstract tasks. Auditors will then be able to produce larger quantities of auditing of a higher quality, which in turn will improve business operations.
“Policy holders will benefit from this process in the end. Management of claims will improve in precision and there will be fewer incorrect decisions.”
AI doesn’t necessarily mean fewer jobs
It is true that jobs performed by highly educated staff will be taken over by computers when they are able to solve problems on their own. Photographic models, accounting assistants and machine operators are often considered to be jobs that probably will disappear in the near future. Other threatened professions are cashiers, financial advisors, taxi drivers and IT support.
However, AI doesn’t necessarily mean fewer jobs; many people think the opposite. From 2020, many count on a positive net effect on employment.
“Auditors will still be needed for advice and for contributing with their knowledge as well as with their intuition, moral integrity and emotions, all of which computers cannot easily provide,” Jan Svennberg concludes.
Hosted by the University of Gävle, the two-year long research project at Länsförsäkringar is a collaboration between a number of universities.
Collaborators include Peter Öhman, professor in accounting at Mid Sweden University, Presha Neidermeyer from West Virginia University, Mats Danielsson, professor in computer science at Stockholm University, and a PhD student from KTH.