A brand new next generation AI-powered service helping insurance companies and their underwriters and decision experts to make binary and multi-class decisions by performing binary or multi-class predictions and classifications for example or sanity checking their existing predictions/classifications. By doing those, Butterfly AI will assist insurance underwriters to automate the process of loss estimation and risk evaluation while reducing underwriting workflows. It can help by analysing the risk associated with customer profile and to reduce the repetitive tasks.
Some more examples:
1. Predict if a driver will involve in an accident and file an insurance claim: Butterfly AI can help auto insurance companies assess the risk level associated with each driver (e.g. a new applicant). Knowing the key data features of thousands of past customers’ profile and journeys including number of past insurance claims, the number of times they have been in 500 m radius of an accident when it happened while driving during past year, the top three most frequent travel routes, number of accident hot spots within the top three routes, age, gender, number of years the driver has held a driving incense, past speeding fines, number of license penalty points, highest speed limits across driver’s most regular commute routes, travel regularly in early mornings (yes or no), travel regularly late night (yes or no), the weather status across the regular routes during the usual times of commute and during those past recorded journeys (rainy, cloudy, windy, lightening, storm, extreme cold and etc.), the lighting status across the top travel routes, the make of the car, the number of similar car models involved in accidents in recent months across the driver’s regular travel routes, history of driving under influence of alcohol, the known distracting factors that led to past accidents across driver’s regular travel routes (for example a direct blinding sunlight on the front screen during specific time of the day in parts of the regular routes or sudden black ice during the cold season on the road across those regular travel routes, etc.), MOT health score, millage on the car, number of breakdowns in past months, smoker, eyesight score, heart disease history, stroke history, high blood sugar, high blood pressure, diabetes, medication taken that may make driver drowsy or lose concentration, number of drivers with similar age, license, travel pattern and health status to the target driver who have been involved in accidents across target driver’s top travel routes, and having verified, labelled and recorded whether those past target drivers during those recorded journeys were ultimately involved in any accident or not (the target and label of the prediction), predict whether a new insurance applicant or existing customer (driver) will involve in an accident or not (and present the probability and certainty) provided that the data features (describe above) are available for that insurance applicant.
2. Help insurance underwriters to estimate the loss level of commercial buildings (and to automate the process): Butterfly AI can help underwriters to classify the commercial buildings based on their potential loss and risk level. Knowing the key data features of hundreds of past insured commercial buildings including building age, number of floors, square footage for each floor level, the overall score of building in terms of being fire resistant (for example fire-resistant by spraying fire-retardant chemicals and using building materials such as concrete or brick and fire-proof roofing materials), number of past insurance claims, category of past claims (e.g. fire, flood, wind damage, etc.), quality grade of water pipework (1,2,3), quality grade of gas-work (1,2,3), quality grade of electric wiring (1,2,3), quality grade of electric outlets (1,2,3), quality grade of heating system (1,2,3), quality grade of cooking equipment and kitchen facilities (1,2,3), the lightening protection system grade (1,2,3), quality grade of construction material used for exteriors (1,2,3), type of the construction material used on exteriors, e.g. fire resistance (1,2,3), quality grade of material used in flooring (1,2,3), quality grade of material used in ceiling (1,2,3), quality grade of material used on walls (1,2,3), earthquake resistance grade (1,2,3), quality grade of material used in wall insulators (1,2,3), the quality level of maintenance practices (1,2,3), the quality level of repair practices (1,2,3), history of bushfire across the surrounding vegetation, distance from epic centre of past earthquakes, on the route of past tornadoes, distance from epic centre of past hurricanes, distance from the frequently flooding rivers, possibility for damaging winds, local wildfire risk, history of the external floods, history of internal flooding, the distance from firefighter brigade station, numerical ranking of combustibility of contents, which measures their effect on the structure under fire conditions, numerical ranking of susceptibility of contents, which measures potential damage to materials or merchandise from the effects of fire, smoke, and water, the quality grade of the sprinkler and extinguisher protection credits, the nature and number of claims within the surrounding buildings if they are exactly same as the target building, presence of vegetation or dead vegetation within ten feet of the commercial building, presence of the any combustible materials within rain gutters, double-pane windows, building’s security grade, hazardous material stored in the building, hazardous material stored in the neighbouring buildings, data on adjacent buildings, including exposed walls, hazards, construction, and distance (for example, a property near a high-hazard operation or next to a storage tank with flammable liquids), the number and category of claims of the buildings in that district which are the closest to the target building in terms of building profile (the most similar buildings to the target building in terms of data features described above including material, grades, and number of floors, etc.), and having verified, recorded and labelled the past loss/risk level of those commercial buildings ( as 1 (lowest), 2, 3, 4, 5 (highest), the targets and labels of the classification/estimation), classify a new candidate commercial building based on the loss level (and present the probability and certainty of classification), provided that the data features (described above) are available for that new candidate building.
3. Predict customer mortgage affordability: Knowing the existing insured customers’ data such as credit history, background, insurance claim history, car, spending, income, mortgage, house value, etc, and whether those customers have managed to pay the mortgage payments or not, decide or predict whether a new or existing customer will afford to maintain the mortgage payments?
4. Decide if an insurance claim is genuine or fraudulent: Knowing the past insurance claims’ details and knowing their applicants’ data including age, demographics, location, drug, alcohol or smoking history, criminal history, health history, credit history, house value, loan and credit history, their job, insurance claim history, car type, spending, income, mortgage, house value, etc, and whether their insurance claim later proved to be fraudulent or not, decide or predict whether a new insurance claim is also fraudulent or not?
and any other binary or multi-class insurance classification or prediction use cases.
Log in/Register Request a Demo Instruction Manual
