Citi’s Gulru Atak and SmartStream’s Andreas Burner discuss how COVID-19 is propelling use cases for artificial intelligence in financial services, and how the best approach lies in collaboration between human and machine.
COVID-19 has triggered a ‘new normal’ in many aspects of our lives, and banking is no exception.
Previously unseen patterns in data, which have torpedoed normal forecasting, together with widespread disruption to working practices, have made a compelling case for artificial intelligence (AI) and Cloud adoption.
“In the banking environment, we’re dealing with a massive amount of data as the number of transactions increases around the world, but banking software is still 10 or 15 years old,” says Andreas Burner, CIO at financial software and managed services supplier SmartStream. “We’re working on projects with several Tier 1 banks and whenever we look at potential use cases for AI, it’s immediately a big data use case. So, you’re really looking at two technologies: machine learning, which can deal with a lot of data and Cloud computing to set it up.”
Citi is also busy looking at the best use cases for AI across its banking divisions.
“One good example is around risk management for our corporate clients,” says Gulru Atak, who heads up Innovation for Treasury and Trade Solutions, part of Citi’s Institutional Clients Group. “We have introduced Citi Payment Outlier Detection, which looks at corporate clients’ past payment behaviour and predicts potential outliers that might be operational mistakes or cyberattacks. It gives the client another decision point to assess whether or not it’s a valid, legitimate payment.
“It’s a great tool and one that’s becoming more important these days because, following COVID-19, we’ve seen bad actors trying to benefit from people working from home through sophisticated cyberattacks. Our new tool enables our clients to better predict those outlier payments that are outside of their typical patterns.”
Late last year, the first technology to emerge from the innovation lab that Burner heads up for SmartStream in Vienna, was launched. SmartStream AIR, which stands for AI in Reconciliations, an intelligent reconciliation engine, has seen rapid uptake and Burner believes it’s an example of how AI can provide a bright outlook for financial services. In fact, he goes further.
“We want to make financial services software as simple as looking for a weather forecast, or using a navigation system,” he says. “You simply enter your city on your device and you get your forecast. They have a lot of data behind them, but you don’t feel it. I believe that financial services can be that simple. We are thinking of new and simpler ways to distribute applications, and exploring smart assistance that can help you to run workflows and better complete daily tasks from home.
“It’s actually a good time for technology. The new normal will push us to a new level, and that’s good.”
Atak agrees: “I think AI will fuel a delightful, personalised client experience,” she says. “Show me that you know me is something that our clients tell us all the time and AI enables us to do that.”
Racing to real time
The move to real-time processing over the past few years has already had a consequential impact on liquidity management and reconciliation at banks. But, since the pandemic, real time has become the top priority.
“What we’ve seen from clients over the last three to four years is the desire for everything to move to real time, and demand is now moving to just-in-time. With COVID-19, this has become even more of a priority for our clients,” says Atak. “It’s an area where we see great benefits from machine learning, predictive analytics and AI. Being able to reconcile your accounts receivable is traditionally a very human-based activity, but you can leverage machine learning to automate the process.”
Burner gives another example: “We’ve been working on a machine learning project with a bank in Asia. It’s a big team and they have to do a reconciliation each day, which is a very exhausting task for a human. When you analyse the data, you see that in the morning everything is matched correctly, but in the afternoon, when the team gets stressed as they rush to hit deadlines, the matching quality significantly decreases.
“When you run machine learning, you can learn from the behaviour in the morning and then warn users if they match something that a machine would not match, increasing the quality in the afternoon. Or you can fully automate it, to help overloaded users process the messages.
“We didn’t plan for COVID-19, obviously, but the interesting thing is that people who’ve typically reconciled with pen and paper are now looking for alternatives – the easier the better. So, we need to adapt all our products for this new normal.”
So, the good news for those worried that machines will soon render us obsolete is that both the software provider and the banker believe the real value of AI lies in using it to augment human decision-making, not to replace it. This hybrid approach is largely derived by AI’s ability to process data at much faster speeds than mortals, whether that’s detecting outliers or providing personalised information on customers, but the fact is it still requires the human touch.
“There are several studies in the medical domain where they tried to use machine learning for radiology imaging,” says Burner. “They found that the machine has a certain percentage and humans have a certain percentage, but the combination of the two outperforms everything. You need that combination in banking as well. Machines aren’t yet clever enough to understand all the relevant outliers.”
The purpose of AI is to help eliminate pain points for humans, agrees Atak: “We focus on automating certain human-based, manual tasks and providing tools for better decision-making.”
Through one such tool, Citi Global Collect, the bank is eliminating a particularly painful point for its global client base – foreign exchange exposure. Citi clients can now upload and send digital invoices through Global Collect, and the payer chooses when they want to pay and what currency they want to use. So. if payers want to pay 60 days after receiving the invoice, they pay the agreed amount when they received the invoice, regardless of rate changes later.
Fintechs are playing a big role in Citi’s technology strategy. The bank developed Citi Global Collect through a partnership with HighRadius, a newly ‘unicorned’ accounts receivable startup that Citi Ventures, the bank’s corporate venture arm, has invested an undisclosed amount in.
“We’re not just relying on our own build,” says Atak. “We’re leveraging partnerships with the likes of HighRadius. Another company called Cashforce supports our cashflow forecasting. We’re also currently piloting a smart contract negotiation tool for our legal teams. While they’re negotiating contracts with corporate clients, they have the opportunity to be able to learn from previously-negotiated contracts, leveraging natural language processing and machine learning together, which is really fascinating.”
AI has been widely deployed since COVID-19 emerged to plot the likely spread of the pandemic and in developing drugs to combat the disease. But when the dark cloud of pandemic clears, if Burner and Atak are right, its impact in financial services will be no less critical.