Accelerating Keypoint's Digital Real Estate Management Platform with AI
Impact
By embedding artificial intelligence in its platforms, Keypoint is able to reduce fraud cases and will guarantee all home repairs are completed properly and at a fair price. Next to that, water damage claims will be prevented through the implementation of smart sensors and AI.
Intro to the customer
Keypoint is the digital assistant for all parties with decision-making power on repairs, maintenance and building management.
Its digital collaboration platform links all the parties involved and makes claim management highly automated and efficient. A claim becomes digital, transparent and relevant at the right time for everyone. To this end, Keypoint also builds numerous links with property management software for its respective target markets. In addition, Keypoint manages its own repair network across Belgium that can repair all types of assignments, both within and outside the insurance sector.
Challenge
In a rapidly growing property market, both insurers and residents require professional, 24/7 ad-hoc insurance services that provide reliable and fast solutions. With Connect, Keypoint provides a unique platform that brings all involved parties together. Through Repair, Keypoint specialises in the organisation of repairs and maintenance of buildings across Belgium. However, Keypoint sees an increase in the need for faster responses (e.g. claim prevention) and more optimisation (e.g. estimate checks through picture recognition and pricing databases). Therefore, Keypoint considered the next step for the evolution of the Keypoint’s Connect platform to be the integration of Artificial Intelligence and linking it with smart sensors (IoT).
Through the implementation of AI, Keypoint aims to make its platform more effective for:
- Repairs pricing: With the implementation of AI, Keypoint can check estimates based on historical data (e.g. similar claims in the past) and via the creation of a pricing database (automatic database creation via approved estimates) compared with the claim pictures and received estimate.
- Fraud detection: In Flanders about four in 1,000 cases in 2017 turned out to be fraudulent. More than one in three fraud cases involved exaggerated or invented claims of the extent of damage or loss.
- Claim prevention: Preventing claims before they occur: based on Assuralia’s statistics, around 40%+ of property claims are related to ‘water’. By applying anomaly detection on water sensor data Keypoint aims to detect leaks and receive notifications through its platform.
Solution
ML6 accelerated the adoption of AI in Keypoint's platforms by focussing on two technical challenges:
- Automatic processing of repair offers (NLP)
- Fraud detection on claims images and weather data (Computer vision and Structured Data)
Automatic processing of repair offers
Based on object detection, OCR techniques, text classification and named entity recognition, we were able to build a pipeline that could parse a multitude of offers from different repair vendors in a structured manner. This way the prices (and other information) from offers can be automatically extracted and compared, line item by line item, with market prices for that line item. This way a price estimator was built to indicate whether an offer is overpriced or not.
Fraud detection on claims images and weather data
Most claims contain one or more pictures of the damage. These are mostly taken by the customer and provide evidence for the damage claim. Based on the type of damage (known in the claim), we could verify through computer vision techniques whether or not the content of the picture was in line with it. Next to that, we provided Google Streetview Images for the address of the claim. This helped the human in the loop to verify whether the requested claim was in line with the sort of house, roof, windows, etc. and whether any damage already existed before the date in the claim.
For storm damage, we can detect fraudulent claims based on weather information and the time and location where the damage occurred. Also, we look at the number of similar claims in the same time period, to verify whether the claim is justified. Outliers are evaluated case by case.
By
Results
By embedding artificial intelligence in its platforms, Keypoint is able to reduce fraud cases and will guarantee all home repairs are completed properly and at a fair price. Next to that, water damage claims will be prevented through the implementation of smart sensors and AI.