Customer Relations: AI is a great approach but should not be seen as an end in itself

Mathieu Choux, Partner at Ekimetrics, has been interviewed by Vincent Ducrey, CEO of the HUB Institute.

When we ask our customers where they are in terms of the relational promise, they find it increasingly difficult to give a clear answer. The relational promise tends to get lost somewhere in the technological promise. This often results in a loss of coherence in terms of brand messages.

Mathieu Choux takes the example of insurance companies that want to change their image i.e. from “payers in the event of an accident” to “business partner companies”. To do this, companies need to address two transformation challenges i.e. refocusing on customer expectations and simplifying their operational approach.

In terms of customer expectations, the two key moments for policyholders with their insurer are: making payment (once a year) and submitting a claim in the event of an accident. All key moments generate varying levels of frustration. Thanks to AI, a score on the probability of the contract being terminated can be created.

At the operational level, it is necessary to move away from static models (customers and aggregated events) toward sequential models including a notion of time. For certain customers, we can see the dissatisfaction mounting slowly but surely as there has been no contact since the time of purchase and the end of the contract: the customer would have appreciated a phone call.

AI can “give insurance companies superpowers” in terms of trying to meet policyholders’ needs. It is in the insurance company’s interest that the model can identify which customers are at risk of terminating their contract. In terms of the policyholder, we can reduce the level of dissatisfaction by meeting relational goals without the need to earmark additional resources.

Transformations succeed when they are driven by business and not technological opportunities. You need to see AI as a way of developing a brand rather than a goal in and of itself. And what do I need to do that? Collect data, create models and test them; that’s 20% of the job, the remaining 80% is all about adoption and industrialisation.

—