What is the strategy behind Scotiabank’s newly unveiled plans to expand its relationship with Google?
The $946.9 billion Toronto-based bank will now leverage the tech giant’s data analytics and artificial intelligence (AI) capabilities to grow its scale and computing power, Grace Lee, Scotiabank’s global chief analytics officer, told Bank Automation News.
While Google works with plenty of banks, the Scotiabank partnership is a “trailblazing move,” said Zac Maufe, the tech giant’s head of retail banking.
“What we’re going do with Scotia is something that I think a lot of people are looking to do, but this is one of the first times that we’re actually really doing it in practice,” he said.
Scotiabank has AI capabilities — including natural language processing and voice capabilities — but it needs a large-scale, integrated solution and the required corresponding compute power to use those technologies to create a personalized experience for customers.
Enter Google.
“Certainly, we’re able to do text and speech analytics … we just can’t do it fast enough, and we can’t do it all at the same time,” Lee said. “Now we’ll be able to use natural-language processing the way that we currently do … at scale, to listen to our calls and make sure that we’re understanding when our customers complain about something — what is it they’re complaining about? How can we do topic modeling better and understand intent better, as we’re routing customers to the right agents, and understanding exactly why they’re calling?”
This will allow the bank to automate more document processes and offer a more personalized experience to customers, Lee said. The bank is also piloting a chatbot that leverages Google’s solutions and has already seen improvements in accuracy due to the analytics, she added.
Data barriers to analyzing at scale
Scotiabank’s previously siloed internal systems made it challenging to move its cloud capabilities. While the bank tried to build data lakes to integrate data, it ran into the same barriers that other companies have — pooling the data while maintaining a useful structure to support analytics. While Lee didn’t specify which data sets the bank will move to Google’s cloud, she said it would come from the datasets needed to create a better customer experience and increase the bank’s efficiency.
“We are going to move data very thoughtfully,” Lee said. “We will bring the appropriate data alongside to make sure that we’re executing and we’re enabling those use cases. It won’t be a wholesale lift and shifts, because frankly, we’ve tried that before, as an industry; we set up these (data) warehouses and set up these (data) lakes and appropriately named (data) swamps because that wasn’t done thoughtfully. We don’t want to repeat those mistakes.”
Overwhelming amounts of data is a challenge that Google — with its Cloud Platform and BigQuery service for large data sets and corresponding advanced analytics — is well-known for solving. The data and analytics capabilities are the key reasons Scotiabank chose to partner with Google “as recognized leaders in the analytics space,” Lee said.
Ultimately, supporting personalization is about data, according to Maufe.
“Consumers really are craving having a more personalized banking experience,” he said. “So you really have to understand someone with a lot more depth to really be able to make thoughtful, personalized recommendations, decisions, all of those things. The key to that is, obviously, data.”
Building better models
The bank’s team of data scientists will be able to deploy AI models and use them to automate model development. The model operations process will ensure the data teams are taking advantage of tools that support AI model explainability and monitor for bias drift in an automated way, Lee said.
“We’ll be able to build hundreds of thousands of models with the same number of people and do so in a way that makes us more confident about the outcomes and make sure that we are continuously improving those models over time, in a way that’s governable,” Lee said. “If you have thousands of data scientists building models outside of a structure, governing that is a highly manual effort and it takes a lot of people and a lot of a lot of time to do.
“By consolidating it into a common platform, I can point to that and say, ‘No, we have perfect visibility into what’s going on.’ I have metrics that will help me define whether or not we’re doing that in the right way. And we can hold people accountable,” she added.