Wall Street Will Be Assimilated
The financial services industry’s broad adoption of artificial intelligence and robotic process automation is a matter when rather than if, according to experts.
The technologies have reached a tipping point just in the past few years as the business world generated the necessary amount of data needed to train AIs and automated robotic systems, said Roger Park, partner/principal, strategy and innovation leader, financial services at EY and who spoke during the DTCC’s Fintech Symposium.
“There’s a pervasiveness of data that we haven’t seen before,” noted Park. “Nearly 90% of data that has ever been produced has been produced in the past two years.”
Wall Street’s continuing elimination of manual interactions and reliance of physical artifacts also has lowered the threshold for adopting AI and robotic process automation, which Park defined as a continuum with rote-based transformations like a Microsoft Excel macro on one end and the discipline of cognitive computing on the other.
Unlike legacy process automation projects, which required firms to open their existing systems and might have taken nine months to implement, robotic systems run on top of the existing systems and might take 45 days to deploy.
Many of the robotic systems also can be configured or taught by watching workers performing their tasks.
This flexibility also has altered the return-on-investment calculations on whether a process is worth automating.
“In the past, if you had a function with fewer than 100 people, it might not have been worth doing the system integration,” said Park. “You might have outsourced that. Today, functions that are equivalent to eight to ten FTE are worth automating because of the cost is low.”
In terms of overall headcount reduction, the World Economic Forum estimated in February that AI and automation will eliminate 40% of current jobs in the next five years.
The loss of these positions will bring new risks to organizations as business knowledge starts concentrating among fewer and fewer employees, according to Park.
“The nightmare scenario that none of us want to run into is that you have implemented automation, received great productivity gains but then something unexpected changes,” he said. “In many cases, it is an underlying system that changed, and now you have to explain to someone that the entire operation shut down because you had issues and did not have the proper governance in place.”
The consequent reduction in force will also reverse the 25-year trend of offshoring productivity, added Christopher Surdak, program director, Institute for Robotic Process Automation and fellow speaker.
“Business processor out-sourcers are absolutely out of their minds terrified of the implications of this technology,” he said. “Using this automation, I can bring that back those seats. It might be only 10% of them, but domesticating those positions might not be a bad thing in this political environment.”
Deep learning may do to HFT what HFT did to statistical arbitrage.
AIs may flag "unambiguous" rule violations in the future.
A difficult, but not impossible, task
Machines beat a conference room full of lawyers manually reviewing docs.
The firm completes natural language processing PoC.