With annual revenues over 8 billion USD, Autoliv, the car safety supplier, is one of those "invisible giant" companies.
I worked on a variety of projects for them, from a todo manager app to a driving score safety app; But the most notable project I had the privilege of working on was an AI-based safety coaching program and bot.
From safety data score to real life safety
Using the data from Driving Avatar app (from DA to a safety coaching)
Autoliv had an app called Driving Avatar that gathered data on your driving - if you drove too fast, if you broke too hard,… this gave you a very good picture of how safe of a driver you are. We wanted to take it to the next level, turn knowing how safe you are to making you a safer driver. From data to transformation.
What is the one thing you should know Before driving today?
If we use a sports analogy, Driving Avatar gave you all the stats post match. This new program was all about getting you the information you would need BEFORE the game to win it. What is that specifically you should pay attentions to before getting into your car for it to be a safe drive.
Understanding our users attitude
From the work I had done with the french safety board, we could divide most drivers into three groups.
A. My car my kingdom. Those who who feel that safety rules don't apply to them. C. those who are unsafe because they lack the driving know how. They don't look at their phones, or speed, but they are just not good at driving.
The vast majority of people fall into category B. They want to be safe drivers, they believe they are safe drivers, BUT they find some wiggle room with the rules. Ex. I speed but only on the streets I know,… The program was aimed at mostly at them, as we needed to make coincide their desire to be a safe driver with their behaviour.
A Safety coach not a lecturer
Get to know what motivate you, how to talk to you.
A key factor for behavioural change is understanding the users motivation. Understanding that motivation allows us to use the right triggers to re-motivate them when their behaviour is slipping.
Using the conversational interface makes it so much easier than a form to pick up on the individuals motivation.
Build to be a conversation.
Coaching is a conversation. That is why our entire interface is build as such. Designing it around Dialogflow means that conversation will be available inside the car itself.
Only talks to you when it has somethimg to say / A conversatoin wherever you want it
Unlike a newsletter, and just like a real coach, the coach only says something when it has something worthwhile to to you to say and that it is the right time to say it. If not, if the user starts to ignore the coach, trust is lost and coaching becomes impossible.
Part of the design was defining when are those right moments, what is more valuable to say today,…
Figuring out what you need to know and when
A customised program for each user via flowcharts and variables.
Though we called it a coaching program, it is more akin a set a flowchart that takes into account variables. It couldn't be a set series of steps, it needed to adapt to outside factors, to the users progress, to how often we've contacted the user in the past days,… These variables mean that every users experience would be different and specific to their own needs. We could then asses the behaviours changes and the their flowchart paths to see how we could improve it.
Getting buy-in through experiences not lecturing
The same way you don't plug a 110W toaster into a 220W socket, we want them to drive at 50kmh because they understand why it is the safe thing to do, not just because it is the law.
So we designed small interactive experiments allowing come to the same conclusion as the road safety board about what is safe and what isn't.
The driver's variables (how they drive, what they believe, outside factors like ice,…) decide how they move along the flowcharts which determines WHAT content, experiments, conversations, they get; And the personalisation affects HOW we communicate with them (tone of voice, triggers,…).
