The next, the best: Leveraging AI in improving the customer experience

In the pursuit of providing the right services to the right customers at the right time, the OCBC AI Lab started its journey to provide the next-best conversation to customers. Hear more from Andrea Pisoni (pictured in the middle), Head of Data Science at the OCBC AI Lab, on how the Bank is utilising AI to do this.

Q: What is the rationale behind the next-best conversation project?

Andrea: Providing consistent communication and offerings across channels to customers is not a simple task. An example is our frontline staff, who often have limited information about our customers as well as limited time to learn about them during their brief interaction with them, be it via the phone or at branch. At times like these, it is crucial for the bank to demonstrate an understanding of the customers’ needs.

Through next best conversation, our AI system analyses the full profile of the customer, leveraging information such as their life stage or spending habits, to determine what the next touchpoint conversation with a customer should be. For example, following a purchase of air tickets, the Bank could offer travel insurance as the next product offering. This conversation could range from anything from a financial product, a change of address to as simple as wishing the customer a Happy Birthday!

The end goal is to enable seamless and relevant communications across our contact centres and branches, thus improving the experience of both customers and frontline employees.

Q: Could you take us through the project process?

Andrea: The data science team, data engineering team, and customer analytics team made up the core project group, with our contact centre, customer relationship management team and marketing teams involved at different stages of the project. The planning and design phase weren’t without its setbacks, as we encountered difficulties differentiating the right algorithm to deploy for this project. When collaborative filtering did not work as expected, we decided to build a custom solution.

The amount of data required to train the AI on the entire bank customer base was extremely large, too large for a single server to handle. We overcame this by making this the first data science project to be deployed on the bank’s enterprise big data platform. We created the big data platform and AI infrastructure required to accommodate this project in analysing these 120 billion data points - to handle more data, much faster, including some in real time.

Resolving these issues allowed us to directly integrate the AI model with the bank’s systems and push AI-recommended suggestions directly to the frontline channels, thereby impacting the customer experience in a positive way.

Q: What is most memorable about this project?

Andrea: I enjoyed working in this fast-paced and focused project team, and it’s not often that one gets to work with such a large and rich data set. It was an excellent opportunity for the team to quantify the effectiveness of AI in actual business applications, through capturing of customer responses to our recommendations. Since its launch, we’ve measured take-up rate of up to 12 times than before, with more customers saying “yes” during conversations with the bank.

Taking a step back, this project gave the team the opportunity to revaluate how our work with AI could directly improve the relationship with our customers. Data scientists love working with data, and it is easy to fall into the trap of considering AI and data in isolation. It’s refreshing to be reminded to be customer-centric at all times, even and especially when developing AI systems.