Discovery Insure says its artificial intelligence (AI)-powered Advanced Driving Instructor (ADI) has reduced risky driving events by up to 10% among clients using it.
Introduced on the Discovery Insure app in January, ADI analyses individual driving patterns in real-time and provides targeted guidance to help motorists address specific driving habits. These include speeding, harsh braking, harsh acceleration, harsh cornering and mobile phone use behind the wheel.
It works by allowing clients to receive personalised driving goals, as the insurer increasingly uses behavioural science to improve road safety.
Michael Rodewijk, chief analytics officer at Discovery Insure, tells ITWeb that the early results indicate the technology is helping clients make measurable improvements, with some motorists maintaining better habits beyond the initial coaching period.
The findings are consistent with research conducted by Discovery, in partnership with Johns Hopkins University and Affective Mobility, which informed the development of the feature.
“Clients with ADI goals have reduced risky driving events by 3% to 10% across the five targeted behaviours: phone use while driving, harsh acceleration, harsh braking, harsh cornering and speeding.
“These improvements are sustained, with 68% of clients who achieve a goal maintaining the improved behaviour the following month. Consistent engagement is also associated with stronger outcomes. Clients who use the app daily have a 49% higher goal achievement rate,” Rodewijk explains.
Discovery Insure provides short-term insurance products, including vehicle, home and business cover, alongside its Vitality Drive programme, which rewards safer driving.
AI coaching
ADI operates within the existing Discovery Insure app and uses driving data to determine which behaviours need attention.
Based on motorists’ driving patterns, the system assigns personalised goals and sends coaching messages intended to encourage gradual improvements rather than attempting to change multiple habits simultaneously.
According to Discovery Insure, the AI models and proprietary algorithms analyse driving data over time to identify recurring behavioural patterns.
The system then prioritises the behaviour, presenting an approach for improvement, translates it into a measurable short-term goal and delivers personalised notifications, continuously reassessing performance to determine subsequent coaching interventions.
According to Discovery's 2025 Integrated Annual Report, Discovery Insure covered more than 278 000 vehicles in SA as at June 2025, while its Vitality Drive programme had over 550 000 local and international drivers.
“Between 45% and 50% of clients receive a personalised goal in any given month. Since its introduction, more than 30 million coaching notifications have been sent, while daily activity on the app has increased by 25%,” notes Rodewijk.
He says the insurer is also using behavioural research to refine how the technology communicates with motorists, including the wording, frequency and presentation of its notifications.
“We continue to refine ADI through client feedback and further experimentation. Streak goals, where a client has a specific target they must meet over a specific number of days, have been a particularly promising development, showing 3.7 times the improvement in driving behaviour.”
AI-generated alerts that identify and discourage motorists from using their mobile phones while driving have generated the highest frequency of coaching activity and the strongest response from clients, according to Discovery Insure.
The insurer says progress on one goal can also encourage motorists to pursue further improvements. Clients who achieve their previous target are 2.6 times more likely to achieve their next one.
ADI forms part of Discovery's broader Vitality Drive programme, which uses financial incentives and behavioural interventions to encourage motorists to reduce their risk on the road.
“Discovery complies with all applicable POPIA requirements and maintains rigorous data protection and security standards. All analytics use only data that clients have expressly agreed to share,” Rodewijk concludes.
Share
