Synchrony minds HR as it develops AI
Synchrony Financial, a bank and a provider of cobranded credit card programs, is deploying artificial intelligence in myriad ways: It’s using machine learning to detect fraudulent transactions, robotics process automation to handle mundane operations tasks, and a virtual assistant named Sydney to answer basic questions by text chat.
“We'll see AI across the company,” Margaret Keane, Synchrony's CEO, said in an interview. “We've taken an active stance and worked with McKinsey to study the areas of our company that could be most impacted.”
At the same time, Keane says, the company is trying to be conscientious about how these deployments will affect employees.
“Some people are saying 40% of jobs will go away,” she said. “We have to be very thoughtful about how we approach this. We're trying to take a stance that we need to help our employees transform during this.”
With training and “reskilling” programs, tuition reimbursement and mapped-out career paths, Synchrony is trying to soften the blow of eliminated jobs and help employees take on new roles working with AI.
Smarter fraud detection
Synchrony started applying AI to fraud detection after surveying customers and hearing that reporting potential fraud on their accounts was one of their more painful experiences.
When a customer called to say their card was used in, say, a grocery store in California when they were at their home in Connecticut, they used to get grilled by the customer service reps.
“They would say, 'OK, tell us some more: Where were you last Tuesday? How do you know it wasn't your transaction? Do you have any siblings or family members in California?' ” said Michael Bopp, who runs enterprise analytics as Synchrony's chief engagement officer. “You’d go through this terribly manual and customer unfriendly process to confirm the truth of what this person is telling you.”
There was good reason for the scrutiny. Some callers do try to fool the call center reps. Sometimes they’re not who they say they are. Sometimes they conducted a transaction and were trying to not pay for it.
Bopp, who oversees a group of 170 data scientists, had his team create a machine learning model that predicts the likelihood that a transaction is truly fraudulent.
The model was trained with information about fraud investigations that have already been conducted. It’s constantly fed transaction data for all the accounts a customer might have with Synchrony. (So if a customer has a Gap card and an Amazon card, both of which are Synchrony clients, the fraud analytics team can see all transactions on both cards.) This is helpful because fraudsters often conduct their schemes across multiple accounts. The model also receives geolocation data.
The model is trained to look for suspicious patterns. If a customer has a billing address in Stamford, Conn. and has had nine transactions in the Stamford area and one transaction in California during the week, the odds of that West Coast transaction being fraud are high.
Bopp said the model can predict with certainty the legitimacy of a transaction more than 90% of the time and give it a red or green light that’s passed along to the call center representatives, so they can handle calls in a more informed way. (Yellow-lighted transactions are investigated by humans, the results are then fed into the fraud model.)
"We also have to be realistic that there are manual jobs that are going to go away. If that's true, how do we make sure we help that transition for folks?"
Now, far fewer investigations are conducted by humans. This saves a lot of time, as these investigations typically take two weeks and involve pulling data from different sources. This means customers can get new cards faster and start spending again more quickly.
Eventually, fewer people will be needed for fraud investigations. But so far, Bopp said, these people are doing higher-value work, like deeper investigations. The company is also working to build career paths that teach people to become data scientists and that enable data scientists to climb the corporate ladder, he said.
Another place Synchrony is using machine learning is in its virtual assistant, Sydney, which answers basic questions on behalf of all of the company’s major card clients, such as, “When is my payment due?”
Carol Juel, Synchrony’s chief information officer, said Sydney was built by in-house developers who focused on understanding the intent of customers’ questions.
“People ask the same questions in all different ways,” she said. “So even though it may appear to be a simple answer, you have to make sure you understand the intent of the question. If you're not understanding what someone is saying, it's game over, you're not going to be able to help the them.”
The bank is planning to deploy Sydney to call center reps to help them answer customer questions. It hopes this will help provide consistency across digital and human channels. Eventually, Sydney will replace the knowledge base currently used in the call centers, called Genius.
Synchrony is also using AI in credit underwriting and in authenticating new customers during the account onboarding process.
The next phase, Keane said, will be AI for intelligent voice interactions with customers over Alexa, Siri and Google Home. (The company already has an Amazon Skill.)
Work done by bots
Synchrony is using robotics process automation software, which some might call low-IQ AI, to automate manual work throughout the company, such as data entry and cutting and pasting information from one application or website to another. It uses Automation Anywhere’s software for this.
Operations and finance are among the departments that will be affected.
“In institutions like ours, there are certain jobs that never merited the investment to build a solution for that because the cost benefit wasn't there,” Juel said. “But RPA allows you to automate those manual processes.”
At the same time, new jobs will be created, Keane said.
For instance, today Synchrony has call center reps who focus on fraud, collections, or customer service. With Sydney handling many basic questions and only forwarding more complex queries to humans, call center reps could handle all of these things, making their jobs more interesting, Keane said.
“The job becomes more fulfilling because you're doing multiple things and you're hopefully giving even better service to the customer because it's one stop,” Keane said. “It will take time to get there, but it's not impossible that you can get to that level of service.”
"We can't place everybody and I'm not saying we're going to be miracle workers, but if we can't, how can we help them, what other things can we do?"
Synchrony executives have begun referring to these multitopic reps as “super reps.”
In all the places Synchrony is deploying robotics and AI, Keane says, it considers the impact on jobs.
“We want to look at it from both sides,” Keane said. “In some ways we can make jobs better. I think we also have to be realistic that there are manual jobs that are going to go away. If that's true, how do we make sure we help that transition for folks?”
The company tries to brace employees for the new world of AI in a few ways. It’s mapping out career paths as Bopp described for fraud people and data scientists.
It pays up to $20,000 a year for employee tuition, so employees can pursue other lines of work or rise in the organization.
“In our call centers, we have a number of employees who put themselves through university or college, which is pretty cool,” Keane said.
A career development program called Step lets call center and operations develop new skills; 70% get promoted at the end of the two years.
“It's a great way to build retention inside our company because people feel a lot of loyalty to the fact that they were able to work, go to school and then get promoted to this program,” Keane said. “Our goal is to have them go from hourly employees to exempt employees.”
So AI is likely to cause the company to shift from having many lower-paid employees to fewer, better-paid people.
“The opportunity is there for that,” Keane said. "It's not going to happen tomorrow.”
Overall, Keene sees responsible AI as being sensitive to the impact on employees.
“If you just thought as a finance person, you could say, 'If I roll out these bots I can cut 500 heads, that's great,' ” Keane said. “I think, that's great, but how are we going to help the 500 heads? I think going into this we have to be very thoughtful about helping the organization transform and what that means for skill sets and planning. It may mean that we can't place everybody and I'm not saying we're going to be miracle workers, but if we can't, how can we help them, what other things can we do?”
Editor at Large Penny Crosman welcomes feedback at firstname.lastname@example.org.