By the start of the project, Greater Than developed a pilot plan with four different partners in order to achieve two main objectives: training the AI to predict risk and optimizing communication tools to influence driving behavior. Greater Than's choice of partners with varied geographical reach, business models and demographics aimed to maximize the insights to be drawn from the pilot across different factors. The pilot plan was divided into four stages: the first and second stage included embedding the pilot into the business strategy and gauging market uptake, and the third and fourth stages included testing the pilot with partners and validating the technology - of which the last stage will be a continuous process.
To meet the pilot’s two objectives – optimizing the AI and end-user communication tools – the former has been analysed by collecting claims data and running analyses on the statistical significance of the correlation between risk level and claims cost, as well as risk level and injury, and the latter has been analysed by launching new features with a reduced customer group and comparing behavioural changes with the control group. At the end of this first period, Greater Than's AI and data analysis module were successfully completed. The data analysis findings show that Greater Than’s AI technology is making accurate measurements, which predict accident likelihood and injury occurrence. The tools that have been introduced to incentivize accident prevention, such as gamification and app use, have been found to significantly reduce accident risk. Equally, user and customer experience of these apps has been central to improving the efficacy of these tools.
A significant part of preventing accidents before they happen will rely on communication with high-risk drivers in real-time. We have conducted an experiment to evaluate the efficiency of different means of communication with drivers in high-risk situations. The experiment relied on two types of warnings: alarms and voices. After each communication, respondents were invited to grade the alarm/voice on two scales, the participant’s perception – i.e. whether it was irritating/stressful vs. enjoyable/pleasant – and the alarm’s effectiveness – i.e. how effectively said warning was in conveying the fact that one is driving in high risk. One overarching conclusion is that voice communication is both more effective and pleasant for users. Equally, personalization is key for user satisfaction. As such, any future implementations ought to include multiple choices, including gender as well as the option to record a personalized voice. This personalized voice could be used for an emotionally closer relationship, such as a child, husband, daughter, friend or other.
With regards to legal preparations and agreements we have identified and established the documents needed in legal matters to fulfil the partner legal requirements and to make the sign on process with new partners as smooth as possible. The review of what are requested in legal terms have been validated with several customers in several region globally.
Focusing on business model development, we have analysed our offerings to adjust the business and pricing model to the market willingness to pay for the service. We have within this development identified and established a structure to make it easy to prepare an offering to the partner, which we know is competitive both in pricing and with added value of the services and features.
