Freedom Through Automation

It's a familiar scene in financial organizations, says Adam Devine: huge rooms of analysts, often well-educated and well-paid, staring mindlessly at one screen and keying data into another.

But these days are over, adds the New York-based vice president for product marketing and strategic partnerships at WorkFusion, as machines learn from humans to perform repetitive tasks with unprecedented accuracy.

Machine learning algorithms are bringing true automation to the "bloated middle" of global banks, says Devine, whose company can apply its solution to a wide range of functions such as corporate actions, customer onboarding and standard settlement instructions (SSI).

Devine believes people have better things to do than key in data all day. And they're also not very good at it-human error is inevitable. Rules-based automation, too, is notoriously error-prone.

But machine learning combines human intelligence with programmatic efficiency.

Workers using WorkFusion's user interface go about day-to-day data entry tasks. After a certain number of repetitions, the software finds patterns in these tasks and makes inferences, getting incrementally better at the task.

Pattern Identification

Let's take corporate actions press releases as an example. The software watches how workers interpret the press releases and within 100 to 150 repetitions, identifies patterns-what a dividend payment looks like, for example, and where in a press release it is typically found. The software can then automate the extraction of information from future press releases and monitor corporate actions news feeds for similar announcements.

Devine stresses that WorkFusion is "built with humans in the loop." The idea is not to eliminate humans from the process, but rather to elevate decision-making to people and let the machines do the boring work.

"We are not just a straight automation solution. Where there is a change in the process or in the format of the website or the document or the newsfeed, and our software needs a new rule, it's automatically escalated to a human to make some sort of a judgement. And it's that judgement that makes the software smarter each time."

One bank deploying WorkFusion for SSIs began with an accuracy of 98.6%.

"This sounds good," says Devine. "But banks are really striving for 100% accuracy. One mistake with settlement instructions means that half a million dollar payment goes to the wrong bank, which would be a big problem."

With WorkFusion, the bank achieved 100% accuracy of process and cut its SSI BPO contract by 80%. The time it took to process an SSI was cut from 15 minutes to 30 seconds.

Devine says solutions like his are eliminating the need for expensive and time-consuming IT projects. "Where it used to be a competitive differentiator for big financial service businesses to be able to access certain data, the trend going forward is that it's going to be a commodity service. It's businesses like ours that are automating this work and fuelling this trend."