Morgan Stanley Center of Excellence Readies Bank for AI’s Data Demands

By Sara Castellanos
April 22, 2019 5:00 a.m. ET

Companies are looking to supercharge corporate decision making through artificial intelligence, but first they need help preparing the troves of customer and business data they have acquired over the years.

Morgan Stanley, for example, has set up a Data Center of Excellence to help address the ongoing deluge and complexity of enterprise data and to help develop artificial intelligence applications that rely on high-quality data.

About 30 experts specializing in data architecture, infrastructure and governance are acting as data advisers to different business and technology divisions within the bank, partly to ensure AI and other applications are being built with the right data.

“The reason we have the Center of Excellence is we want to continue to build on AI, and we understand this is one of the foundational areas that is needed,” said Katherine Wetmur, head of quality assurance and production management, who is transitioning into the role of international chief information officer. Ms. Wetmur works with the team to develop services across business units and oversees separate data-related projects.

Morgan Stanley recently posted a profit of $2.4 billion, or $1.39 a share, on revenue of $10.3 billion. Both are lower than the same period a year earlier, when the firm earned $2.7 billion, or $1.45 a share, on record quarterly revenue of $11.1 billion.

Created last year with staffers spread across New York and London, the Data Center of Excellence works with the bank’s various business and infrastructure divisions to establish best practices and controls around data quality and data security.

The data group is working closely with Morgan Stanley’s AI team to make sure AI algorithms in areas such as commercial real-estate loan analysis, fraud detection and virtual wealth management advisory services are using the appropriate data. “We want to make sure we have high-quality data, because your insights are only as good as the data that underpins it,” said Gez Hester, head of the Data Center of Excellence.

High-quality data is data that’s accurate, up-to-date and complete, Mr. Hester said. To achieve that, the data team must catalogue all the data that’s available in, for example, the commercial real-estate loan business and understand the data’s provenance, where it is being sourced from, and how it is going to be used. Data provenance is key for ensuring that the insights are useful and trustworthy, he said.

In the dawn of AI-enabled decision making, the notion of data as a strategic asset is rising at companies such as Morgan Stanley. At the same time, managing data has grown increasingly complex, the result of multiple data centers and data storage repositories across disparate systems, multiple copies of the same data and new data privacy laws.

The new data emphasis is changing the CIO focus at some companies from big business applications to the data layer, said Anil Chakravarthy, chief executive of data management firm Informatica.

Poor quality, inaccurate and unreliable customer and business data is preventing companies from leveraging AI, according to recent studies. About 76% of firms are aiming to extract value from data they already have, but only 15% said they currently have the right kind of data needed to achieve that goal, according to PricewaterhouseCoopers. The firm in January polled about 300 executives at U.S. companies in a range of industries with revenue of $500 million or more.

More than one petabyte of new data, equivalent to about 1 million gigabytes, is entering Morgan Stanley’s systems every month, according to the bank.

Morgan Stanley’s new data group has worked with internal AI teams on processing and summarizing unstructured data, a broad category that includes web pages and PDF files and other data not already housed in rows and columns, related to commercial real-estate loans. The data group helped develop an application that allows traders to analyze the risk of a loan without having to do deeper analysis manually.

Another AI system developed with help from the group analyzes transactions that can indicate fraudulent behavior. This represents a major shift in the way AI can help humans process massive amounts of data that otherwise couldn’t be analyzed, Ms. Wetmur said. “In the past, you couldn’t consume the data in the speed that was needed to get these insights,” she said.

Separately, Ms. Wetmur has been working on data projects to speed up the time it takes to test various applications before deployment. To do that, her team sought technology from Redwood City, Calif.-based Delphix Corp., a vendor that allows companies to virtualize, secure and manage data.

Delphix’s technology allows data to exist in one virtual copy for each group that needs it so that changes can be made on a single copy in real time. This significantly increases the speed and delivery of application code, Ms. Wetmur said. Application testers can now have access to accurate, up-to-date data in minutes instead of the 10 to 12 hours it would take through traditional methods of copying data, which can be slow and expensive and require manual processes.

So-called data virtualization is also more cost-efficient because it reduces the need to provision new servers, hardware or databases. In addition, the technology helps drive down infrastructure costs, because the bank isn’t storing as many traditional database copies that take up massive amounts of storage space.

The original article appeared on April 22, 2019 in the Wall Street Journal's CIO Journal.