Asked where State Street plans to deploy artificial intelligence software, Moiz Kohari, the custody bank's chief technology architect, flips the question to, “Where first?”

“In my two years at State Street, I haven’t encountered a single business unit where that capability does not apply or wouldn’t help,” said Kohari, who is also a senior vice president. "Our problem is more where do we start to apply it first, second and third because ... you have limited resources."

David Weiss, a principal at the advisory firm Market Structure Metrics, views State Street as close to the leading edge of AI deployment compared with other banks.

He pointed out that to be competitive on a margin basis per employee, high-volume transaction processors like State Street have to either reduce their headcount or be better at what they are doing to improve profitability.

“How are they going to do that? Through these technologies,” he said. “Everybody is looking at leveraging the employees they have now to increase productivity or be more profitable.”

Survey of how many large banks have studied or implemented AI

Michael Spellacy, a senior managing director at Accenture, concurs.

“Many of the global asset management firms are trying to determine where to apply artificial intelligence to bend the cost curve or to raise their revenue profile, and more importantly, serve customers, engage customers and drive a better customer experience,” he said. “It is a pervasive topic in the industry.”

Kohari discussed the AI efforts underway at the bank, which can be summed up in five categories.

Alerts and reconciliation

State Street’s global services group, which provides asset custody and accounting services, generates hundreds of thousands of alerts each night for transactions that happened during the day, like, say, the sale of 10,000 oil company shares at a certain price. The group also handles all reconciliations between clients.

Such work is hard to automate because it is difficult to define every possible scenario that might arise.

This is where machine learning comes in.

“With machine learning, instead of me having to program the machine how to do every single thing, the machine itself looks at previous actions and starts to create new algorithms on its own that help me deliver the service that I’m trying to deliver,” Kohari said. “Then I don’t have to spend hundreds of thousand of hours a night doing menial jobs.”

This could lead to large efficiency improvements.

“State Street is a huge custodian,” Weiss said. “The more efficiently they can process all the paperwork, the better.”

A crystal ball for markets

The bank is working on predictive analytics that could tell clients the likelihood of certain events occurring and the effects those events would have on their portfolios.

Today, clients might ask how a war in Iran would affect them. State Street would then create models that could predict the probability of such a war, and an analyst would determine how much the war might influence the price of oil and could point out, for example, that the client's portfolio holds 29 oil or energy stocks.

“With artificial intelligence, the software looks at planetary data and specific client data, starts to draw correlations and starts to alert the client ahead of those types of events,” Kohari explained. “That provides you with much better operational capabilities, and clients end up getting intelligence they wouldn’t get otherwise.”

Other banks and Wall Street firms are also using AI to have machines generate predictions, Spellacy said.

What if a machine makes an off-the-wall prediction?

“The honest answer is, let’s not prognosticate on the future,” Spellacy said. “Not in the near term. What we have today is use of this technology to help human-based interaction, patterning and decision-making, and provide greater insight that makes better decisions with fewer risks.”

Client communication

Even in this modern age, State Street still receives faxes from clients about transactions.

“That’s a manual process — somebody has to look at these faxes and transactions, and it’s error-prone because humans are touching them,” Kohari said.

The bank has invested in natural-language-processing software that can read the faxes and extract the entities — the service provider and the service consumer, for example — and start to apply relationship-mapping to figure out what should be done with each fax.

Early-warning systems

State Street is also using AI to discover anomalous behavior in transactions.

AI software can analyze historical data and new transactions and identify anything odd.

“This is a very interesting area of implementation because it helps you identify problems before they occur,” Kohari said. “If I normally log in from Boston and all of a sudden I log in from China, that’s an anomaly. You want to be able to identify that. Credit card companies do that really well.”

Building data lakes

State Street is undergoing a large information management project to make large volumes of data available to all of its AI engines while ensuring clients’ data privacy.

“You have to be able to control who’s gaining access to what, at what level, when, who they have permissioned, and are there any red flags on different types of profiles that may impact that access?” Kohari said. “If a client accesses this data out of China at night, does that match the profile of activity?”

Conor Allen, global head of data architecture at the bank, is leading an effort to identify all transaction data and define roles, responsibilities and access control mechanisms around that data so that it is viewed and stored only in appropriate ways.

“Say you have two clients: Morgan Stanley and Goldman Sachs,” Kohari said. “When somebody from Morgan Stanley logs in and they want to look at and drive intelligence from that data, obviously Goldman Sachs wouldn’t want its data included. That’s very difficult to do if you don’t have the access control at a very granular level.”

So the bank is creating a data lake in which all the data is available, but it says controls are in place to minimize problems.

Data is "the lifeblood for all these applications of this technology," and many organizations are trying to build data lakes, Weiss said.

“Everybody’s trying to do that,” Weiss said. “The question is, how much success is anybody having with that? These lakes can be very wide and deep, like the Great Lakes. Boats have been known to sink in the Great Lakes.”

Editor at Large Penny Crosman welcomes feedback at penny.crosman@sourcemedia.com.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access