The promise of artificial intelligence in asset servicing is in what it can do with its lifeblood — data, sources said.
To develop new business in the investment operations industry, "the largest firms identify areas with the greatest opportunity to structure and reorder and cleanse data, and then identify specific use cases," said Lesley Keefe, executive director, Americas asset management advisory leader at EY in Boston. "The business' success defines the need, first and foremost."
It's how artificial intelligence in its myriad forms can be applied to processing that data that's most important, said Wayne Riches, director of strategy and solution management at financial technology developer Fidelity National Information Services Inc., London. Those processes include reconciliation of unmatched transactions, cash balances, foreign exchange and trades with custodians.
"Going forward, you can use AI to continually refine the reconciliation process and create rules to make it work better," Mr. Riches said. "It can be used by a money manager as well as a custodian, or also a pension fund that insources. We've seen trends where plans shadow their managers and custodians to understand what's going on. AI gives them an element of control. Ultimately, the asset owner is the last port of call."
One kind of AI that's seen as having direct applications to back- and middle-office functions is machine learning, which tracks and then mimics repetitive tasks that have previously been done manually, said Josh Sutton, CEO of Agorai, a Singapore-based artificial intelligence open-access software marketplace. Other forms of AI, Mr. Sutton said, include natural language understanding, in which computers can listen to or read words and make sense of them, much like applications like Google Home or Apple's iHome. Another is machine vision, imaging-based automatic inspection and analysis for such applications as process control and robot-driven guidance.
All of those can be applied to areas in which contract and trade reconciliation are done, Mr. Sutton said. "In trading, reconciliation boils down to figuring out why trades were incorrectly reported," he said. "These technologies help people determine where errors happened."
Liz Blake, global head of managed services at Eagle Investment Systems LLC, Boston, said another component of AI, robotic process automation, can be applied to multiple functions related to trading and settlements.
That automation can be used for "anything that's a repeatable task," Ms. Blake said. "In the middle and back office, that's reconciliation, gathering statistics, trade settlement, fund administration. That can be applied as a component of AI because robotics, by nature, is a standardization of an automated task."
Ms. Blake said robotic process automation is the "foundation" for other forms of AI, such as machine learning. "Once I do that, I can do machine learning predictive tools, then look at statistical machine learning. Millions of data points can be gathered and included to determine a predictive trend," Ms. Blake said.
Once robotics is applied, Ms. Blake said, deep learning can be applied. "The next step — and this is even more fun — speech, images, all sorts of unstructured stuff becomes digitized," she said.