​Pegasystems helping CommBank, Sprint leverage AI for customer retention

Pegasystems CTO Don Schuerman believes organisations need to implement pragmatic artificial intelligence technologies just like the Commonwealth Bank and Sprint to avoid getting swept up in the AI hype.

Artificial intelligence (AI) is simultaneously overhyped, misunderstood, and already hugely impactful, according to Pegasystems CTO Don Schuerman, who said many organisations were already successfully using the technology before it became an industry buzzword.

Schuerman told ZDNet that he has been encouraging his clients to take a look at what pragmatic AI technologies they can employ to deliver real benefits to how they interact with customers, rather than succumbing to the hype.

The best example is what the Commonwealth Bank of Australia (CBA) is doing, he said.

"They've got Pegasystems' product called the Customer Decision Hub, which uses predictive analytics and adaptive or self-learning analytics to figure out in real time what the best offer or service is, or recommendations are, to put in front of a customer either to improve their satisfaction rate and retention rate, for example," he explained.

In implementing the technology, CBA has increased its branch sell rates by 13 percent, which Schuerman said is a significant impact to the business.

He said clients like United States telecommunications provider Sprint are building nine-figure business cases around the use of Pegasystems technology, and in some cases have been doing it for the last five or six years.

"This isn't something that's new or in the future -- this is something that is real and proven at scale at some of the world's largest organisations," he said.

In the US, Schuerman explained that the telecommunications industry is capped out, with barely any net new customers for Sprint to sign up. As a result, the only customers the organisation can get are by stealing from other providers.

After suffering from a "really horrible" churn rate, Schuerman said Sprint worked with Pegasystems to fix customer retention.

"In nine weeks, using real-time analytics and machine learning, [Pegasystems put in place] a system that looked across a set of data about a customer and made a real-time decision when the customer was on the phone as to what their risk of churn was ... and how much Sprint was willing to spend to keep them around as a customer," he said.

"What they saw after implementing that system was a 30 percent increase in offer acceptance. But more important, a 10 percent increase in customer retention six months after that conversation."

To Schuerman, enterprise IT companies talk too much about parts of AI that deliver the least potential value. This is perverse when there's AI that's proven to deliver tremendous value and is underused today, but if there is one thing Schuerman wants to make clear, it's that AI technology is already delivering real value to businesses.

"AI is certainly the buzzword of the industry right now, and especially in the area of enterprise software I think there's a lot of noise around AI as this magical thing, and lots of organisations are talking about potentials,' he said.

"There have been lots of examples of AI running amok -- AI is only as good as the human governance and judgement you wrap around it.

"AI is one input into a decision that a company needs to make, but human judgement needs to figure into that as well. To me, the real promise comes not AI standalone -- regardless of the amount of data you throw at it -- but AI plus human governance and human judgement."

Using the Sprint example, Schuerman said that the concept of big data is also one talked about at length. One of the things he said he has experienced in working with enterprises, however, is that there is still a lot of little data in the enterprise that an organisation hasn't fully managed to exploit yet.

"[Pegasystems] focused on a relatively small set of data -- I think we only used 50 or 60 data fields for the first appointment -- because sometimes organisations feel that they have to have some sort of massive data store to get started with this, but the fact is they can start with a small data set and optimise it over time," Schuerman said.

"We generally start by helping an organisation to exploit their little data, and then pull their big data in once they begin to see some initial value and initial return."