What would you do if someone owes you money… for a long time?

You might think: if he is a trustworthy friend, a kind reminder should suffice. However, if it is a so-called “deadbeat” who never repays the debt on time… A little push might be needed.

This is exactly what our voice AI tries to do when serving different customers. 

For each type of customer, it generates a highly personalized response, all for achieving the predefined goal. But how is voice AI able to achieve this? 

There is a complex mechanism behind this, but we can break it down into three basic steps. 

 Step 1: Collect & Classify 

Collect customer data and group customers with similar characteristics.

Before calling a customer, Voice AI firstly gathers related information about him/her from previous conversations, along with big data from other platforms (with consent).

Based on dimensions like gender, age, habits, personality, etc, voice AI categorizes millions of customers into small groups with similar traits. 

Take our voice AI for loan collection as an example. Suppose now we need to remind two customers (one named Mary, and one named James) for repayment based on the following data:

According to the data, voice AI categorizes Mary and James into the following customer groups.

Mary: Creditworthy but forgetful type

James: Habitually default type

Step 2:  Customize voice AI for each customer group 

For different customer groups, voice AI adopts corresponding voice types and scripts. 

For example, for “creditworthy-but-forgetful” Mary, voice AI uses  “gentle female” voice +”kind persuasion”.  

For “habitually default” James,  a “serious male” voice + “ determined persuasion” is better.

By matching Mary and James with the right type of voice and scripts, voice AI talks with them respectively.

With Mary

With James

 

Step 3: Allow real-time adjustment based on the context

What is worth mentioning is, to offer the best-in-class interactive experience, the response of voice AI not only depends on how it categorizes the current customer, but real-time context in each conversation.

For example, in the conversation between voice AI and James:

“Sir, sorry to disturb you”  is an answer based on the actual context. That is because:

· James made it clear that the calling time is too early

· Big data suggests that James usually does not use his mobile phone, computer and other electronic devices by this time.

Therefore, it is almost certain that James is still at rest at that time, which calls for a timely apology to comfort the customer and collect the debt.

 Last note

Above is a simple example of how voice AI takes a personalized approach to deal with different types of customers. 

However, in actual operation, by no means will customers be classified only as the “creditworthy but forgetful type” and the “habitually default type”. There can be thousands of customer groups and the algorithm keeps renew and add more groups on a daily basis. Meanwhile, the analysis of the actual context is much more complicated, with the analysis for new contexts added every day.

According to statistics, as the division of a large group deepens, the internal difference in each small group will be increasingly smaller, which makes the personalized service for each customer possible. 

In an era when “personalization” is increasingly powered by technologies, personalized customer experience is not only a trend, but also what we are working on right now. 

If you are interested to know more, or have a live demo talking to our voice AI, feel free to reach us at sales@airudder.com.