May 28, 2018 12:35
Photo : Freepik
The French company Shift Technology has developed a platform entirely dedicated to the detection of insurance fraud by the artificial intelligence. The company already employs fifty customers in 25 countries. It will soon arrive in Canada.
Samuel Falmagne is responsible for the international development at Shift Technology. There was a passage in the recent edition of Insurtech Quebec.
Shift Technology was founded in 2014 by three young entrepreneurs, two of whom were completing an internship at the insurer AXA. They have seen the difficulty of modelling the types of fraud. They have found a partner and launched the company from Paris.
The technology developed by Shift is based on the detection of fraud in claims after a disaster. According to Mr. Falmagne, it is estimated the fraud to insurers of $ 3 billion (G$) per year in Canada, and more than$ 40 billion in the United States. «We want to help insurers detect fraud, but also to better serve customers in honorable really victims of a disaster. «
Bet on the modeling
The majority of the personnel Shift is made up of specialists in data analysis and mathematicians holding a phd. All of these people develop algorithms thanks to the constant progress in artificial intelligence. They were able to model the work of investigation, the collecting of clues and the intellectual progression that will bring in the experts claim to conclude to the possibility of fraud.
The platform developed by Shift, named FORCE Fraud Detection Solutionis based on a wide variety of typical scenarios of fraud. Each claim is well decked out with a note that corresponds to the probability of a fraud.
In addition to the score, the tool also provides the reasons that determine the likelihood of fraud. «This is the most probable scenario used by the fraudster, as well as the elements identified in the data that allow us to arrive at this conclusion,» said Mr. Falmagne.
We use the data provided by the insurer, but it is crossed with other sources of information that allow to validate or not the statements of the victims of the disaster. This way, the analysis of the data is highly structured, from internal or external sources, such as GPS information of the vehicle or cellular data, weather data, or unstructured data, such as police reports, recordings of call centers, photos, PDF documents and other.
«Insurers collect a large amount of data. The challenge is to give them a meaning, » he says.
The platform allows to clean the data by eliminating duplicates and by tracing down the master of errors at the collection stage, the single shell type is wrong.
Finally, the tool of detection of non-fraud can automate the process of settlement of claims. «Some customers come to us just for this function,» says Dr. Falmagne.
Shift has created a veritable library of scenarios of fraud. We can thus verify if the strategies of a network of fraudsters in auto insurance in a given country may have been recovered elsewhere in the world.
The alerts are sent through an Internet platform. For each block of 100 alerts per Shift, the fraud’s actual average of 75 %. Among the first customers, where the platform has been installed for a longer time, the relevance rate even reached 90 %. Much has also eliminated the » false positives «, a common problem in computerized models, he adds.
Rapid growth and sustained
Already 120 people working at Shift Technology at the beginning of April. This number is expected to climb to 200 by the end of the year. We are also working to develop a second product based on processing claims that have passed the first stage where the learning algorithms were not able to detect any pattern well-established fraud to the insurer.
Shift has managed to raise some $ 40 million US ($M) in capital to fund its development, including$ 28 Million in October 2017 with a group of institutional investors. With clients in 25 countries, the platform Shift has made it possible to treat a block of more than 90 million claims insurance auto, home, or travel, as well as $ 400 million of claims for health-care coverage.
Not for sale
A good number of startups looking to be bought by insurers, but this is not the ambition of Shift Technology, emphasizes Samuel Falmagne. The shareholders aim to achieve a market capitalization of at least a billion dollars. «We aim to make an initial public offering within two or three years. We are on track to get there. This will be easier with future canadian clients, » he says.
Shift does happen, however not in uncharted territory in Canada. The firm with the technological development V-NEO, based in Quebec, has connected a «version» vanille » of the platform Guidewire, which is used by several of damage, with the platform for data analysis Shift. The interface thus developed can already be implanted in the canadian insurer attempted by the tool Shift.