The direct-to-consumer era of e-commerce is changing the way customers behave and is having a ripple effect on merchants’ customer outreach strategies, requiring enhanced speed and accuracy in decision making.

According to Dimi Dosis, president of Mastercard Advisors, the payment giant’s business-to-business consulting arm, technology-driven changes in consumer behavior and expectations are changing how businesses respond to their needs and, in turn, how Mastercard helps businesses map out their strategies around new products and services.

Dosis said his team uses artificial intelligence, analytics and an abundance of anonymized and aggregated transaction data, as well as third-party data, to help businesses do “rapid experimentation,” or quickly test products that go to market. Mastercard’s Test & Learn self-service analytics and recommendation platform, for instance, allows businesses to vet ideas for products and services they’re considering bringing to market. Based on their responses to questions prompted by the automated platform, businesses can instantly analyze data sets that drill down into each metric, category or location in play.

“What Test & Learn does throughout the whole innovation process is it effectively helps you rule out the areas you will not want to go into and direct your focus on the areas that have the highest probability of increased return on investment,” Dosis explained. About 40% of the ideas vetted by the Test & Learn tool don’t pass muster on the platform, which he said has translated into a lot of costs savings for clients. Mastercard acquired cloud-based analytics provider Applied Predictive Technologies, which built the Test & Learn platform, for about $600 million in 2015.

In an interview, Dosis told Bank Innovation how and why his team uses AI and data analytics. An edited version of that conversation follows:

We hear a lot about AI, but what are you actually doing with it?
I don’t think about AI as a product but an ingredient to many, many things we’re doing, and those use cases are constantly expanding. AI, effectively, is being deployed across Mastercard today. To me, AI is an evolution of what we used to call analytics. It’s a manifestation of what we can do with the increased amount of data, the facilities we now have and the velocity with which we can deploy this data.

For example, in a traditional analysis, you would run a kind of data segmentation, take the results, make your slides, come up with recommendations based on those slides and then do a quality check and give it to the customer. This process would likely take 90 to 100 hours. With AI, it takes only 10% of the time it used to because the data analysis, slide production and recommendations are all being handled automatically. We are now deploying this across the business because not only does it save time, which people can devote to other things, but it also drives higher accuracy of data. The only piece we hold onto is quality control at the end, which is still done by a person.

How are you using AI with Test & Learn?
Test & Learn is very much based on the notion of having a hypothesis to test. Hypotheses traditionally have been along the lines of, ‘I, as a bank, think I should sell X, Y, Z product in this branch,’ but these hypotheses are not derived from a data set, in many cases. What we’ve been doing is building an automated hypothesis generation engine, which looks at data in an unbiased way and says, ‘These are the hypotheses you should be using because they have the highest probability of being successful.’ We are applying AI for this, and those hypotheses are being tested.

What do clients look for from Mastercard Advisors?
It started with a very basic proposition like, ‘I run a card portfolio. How do I make more out of it?’ The reason why Advisors came into existence — almost two decades ago — was to help customers answer that question. This is the bread and butter of the business but, from there, we’ve moved to helping customers with things like geographic expansion into other markets. Then it moved very much into innovation and technology over the last few years.

How proactive versus reactive is your team?
The majority of our time is spent connecting with customers on the latest trends and innovations, and they are hungry. Their bosses are asking them questions about how to respond to things like AI or workflow-embedded data systems. They also are pressed to react to regulations in the most profitable way possible. When GDPR [the EU’s General Data Protection Regulation] and PSD2 [the EU’s Revised Payment Service Directive] came up, we were in the center of it in terms of helping them think through the consequences of those regulations for their business.