Wall Street's AI dreams are running into the cold reality of thousands of databases and messy data
- JPMorgan co-president Daniel Pinto spoke to Business Insider recently about the complexity of making usable the massive amounts of data that the bank collects from customers.
- Artificial intelligence can only be unleashed once data is made usable in a clean and consistent way, Pinto said.
Wall Street's dream for artificial intelligence is running into the hard facts of what's needed to bring it to life.
At JPMorgan, the largest US bank, there are thousands of databases that still need to be cleaned and made usable before AI or machine learning techniques can be fully unleashed, according to co-President Daniel Pinto, who spoke with Business Insider on the sidelines of the World Economic Forum in Davis, Switzerland.
Chart that workload across dozens of large banks, not to mention investment firms, and the scale of the work ahead for the industry is a staggering reminder that the robot revolution is still years away.
Sign up here for our weekly newsletter "Wall Street Insider," a behind-the-scenes look at the stories dominating banking, business, and big deals.
For years, JPMorgan built databases for particular purposes in one system only to build another for a different purpose in a second system, according to Pinto. For sophisticated AI techniques to be most effective, that data needs to be assigned the same name and migrated into the same system, or at least housed in interconnected systems. That project is just one of many being covered by JPMorgan's $11 billion in annual technology spend.
"We are in a massive process of making that data usable, in a very clean, consistent way," Pinto said. "We have plenty of data across multiple systems that was developed over time, so often the same thing is called X here and Y there. It takes time, money and effort to really clean up all of that."
JPMorgan isn't alone. For decades, banks were at the forefront of data collection, hoovering up information about stock and bond trades, credit-card transactions and mortgage loans. But for most of that time, firms were content to take in the data and store it, with few spending much time thinking about how it might be retrieved or compared to other datasets nestled in other parts of the firm.
According to a July 2016 McKinsey article, about 50% of the time spent by employees in finance and insurance is used for collecting and processing data. That and the large amounts of data involved in the industry make it one of the area's most ripe for disruption, according to the consultant.
At Credit Suisse, the bank has focused on ensuring that any data that gets fed into AI tools is of the highest possible quality, according to the Swiss firm's Chief Technology Officer Laura Barrowman. For an AI-based tool to be efficient, the data it analyzes needs to be complete and accurate. While that might seem like a basic request, Barrowman said it's a critical one and not easily achievable in a company the size of Credit Suisse.
"Making sure that your basics are right is a fundamental for everything," said Barrowman, also speaking on the sidelines of the Davos event.
If a company is able to create consistencies across the data sets that sit within its organization, the opportunities for what AI can do at Wall Street firms are huge, she said. And they will undoubtedly provide banks revenue opportunities they might not have realized previously.
As banks wade into that new realm, they'll have to be careful to protect customer data. While many clients like the analytics banks are providing around their own data, customers are incredibly uneasy about having their own data shared with others, according to Pinto. The topic is fraught with drama in the wake of consumer-data breaches at tech giants like Facebook.
"You need to be very careful to protect client privacy," Pinto said. "A lot of clients don't want their data used elsewhere, even in aggregate."