Artificial intelligence in banking and personal finance is marching on, and it’s taking lessons from institutional adoption of advanced technologies, said panellists at the AI expo in London.
There are several ways that AI pops up in retail: predicting customer behaviours from big data, customer interaction with chatbots, or customized experiences based on how users are interacting with the system.
Ash Booth, HSBC’s head of Artificial Intelligence for the bank’s Digital Assets Technology unit, explained that AI can be defined as “this idea of learning, planning, reasoning. Things we associate with human intelligence embedded in our technology.”
For Wells Fargo, it’s about AI’s ability to sift through a huge amount of data the bank has collected over the years.
“We are really moving from reactive to predictive,” said Hélène Alunni-Botteri, SVP for Wells Fargos’ Innovation Group.
But all the panellists, which included Roshan Rohatgi, a senior member of the RBS Innovations and Solutions team, were quick to point out that it’s not about replacing humans.
“We don’t want people crunching numbers, we want computers and algorithms crunching numbers and people building relationships,” said Booth.
From institutional to retail
In the markets and trading world, algorithms have been making classifications and predictions for some time and have matured, said Booth.
But in terms of back office or customer interaction, it’s quite new, and so there’s still a lot of “low hanging fruit” being picked.
That could be technologies that recommend other products for cross-selling, or chatbots for both customers and internal use.
“For many years, we’ve been shouting: we can do this with AI. This big boom in AI has enabled us to pick the low hanging fruit and we’re still doing that,” he said.
Not that the hype is entirely a rose garden, particularly as companies make dubious claims about AI capabilities to get funding, but the side effect, said Booth, is that AI has made big strides in the last five years.
“If the people I am pitching to for funding have some idea what I am doing, what I am talking about, that’s great,” he added.
Low hanging fruit
Wells Fargo pointed to its mobile app for predictive banking, which customers can choose to use as an option to humans. It can perform a number of tasks, like keep tabs on spending, for example.
In the back office, RBS’ Rohatgi noted that machine learning approaches are beating legacy systems when it comes to flagging corruption, like money laundering.
“Those (rules-based) archaic systems can’t compete with the new world,” he said. “The unsupervised learning approach, we are not attaching rules to anything anymore.”
In such systems, a vast quantity of payments data is analyzed, clustered and then sampled to find anomalies, he explained. One result is reducing false positives, but another is flagging up issues the bank is unaware of.
This, Rohatgi explained is the easy part. The tough part is having the people and systems in place to derive insight from what the new methodology has delivered.
“We are not set up to derive things the new way, but we are adapting very quickly,” he said.
AI ethics and governance
The approach of European data regulation GDPR, which is extraterritorial, along with widespread discontent over how personal data is used by corporations is certainly a factor in any future developments.
Moreover, it’s happening at a time when trust in banks has been seriously eroded since the financial crisis. All the banks represented on the panel have had to contend with numerous scandals.
Not surprisingly, the panellists were keen to get across the message that how AI is being developed is all part and parcel of rebuilding trust with customers as well as staff in an era when many are questioning how jobs will be affected.
Rohatgi said: “We don’t just throw tools in there to automate process and do things smarter, and eradicate jobs. I think you may have a headcount increase over time as you hire new types of skills.”
Those new types of skills for banks will be about “doing things in a more automated fashion using that big data approach,” he added.
HSBC’s Booth said that if customers don’t want their data used, they can just say no: “It will slow things down if the cause is to advance AI as fast as possible. In banking, it’s not.”
Big, big data
Booth’s background includes stints in the hedge fund world, which were early adopters of AI for automated trading executing millions of transactions a day on AI-based systems.
But customer data doesn’t have quite the same dynamics, he noted.
The question arises, that machine learning systems being adopted require huge datasets for training. Is there enough training data to build models at the retail level?
Certainly at Wells Fargo, noted Alunni-Botteri.
The bank has collected 375 petabytes of customer data around behaviour and preference from its interactions with customers. Digital customers, she pointed out, interact 10 times more than customers using regular channels.
To put that in perspective, 1PB of data could store the DNA of the entire US population, twice over again.
“This is why the time is so right for AI, because you can see that with these numbers, it’s impossible to do the data mining,” said Alunni-Botteri. “We need the machines.”