The company’s Toronto-based Lab of Forward Thinking (LOFT) is collaborating with Silicon Valley-based Nervana Systems on a new AI-based application that could help portfolio managers analyze the high volume of online information, financial news, emails, and documents they receive when researching investments.
“If you think about all of the things that a portfolio manager or researcher might look at when making decisions… a lot of it is social data – emails, industry reports, stock and economic data,” Ace Moghimi, Manulife’s Boston-based head of innovation, North America, told ITBusiness.ca.
“The system we’re putting in place essentially uses natural language processing, supported by an underlying deep learning neural network, to go through vast stores of unstructured data, allowing researchers to analyze the information much faster than they would be able to on their own,” he said.
The project’s genesis lies within the LOFT division itself, which researches ways that Manulife can incorporate emerging technology and unconventional business processes into its products and services – and its Toronto-based data scientists were particularly taken with AI and deep learning, Moghimi said. (The company also has labs in Boston and Singapore.)
In researching potential uses for AI, the researchers settled on Manulife’s asset management business, which frequently forces portfolio managers and researchers to process a mountain of written data that could benefit from machine-based analysis, he said.
Using a deep learning-based natural language processing engine co-developed by Nervana, the Toronto team is training a computer system to decipher communication at every level of the financial services industry, from entry to executive, Moghimi said, before analyzing the resulting data and presenting advisors and other professionals with appropriate recommendations.
Though Moghimi admits the project isn’t likely to result in changes to Manulife’s customer service anytime soon, he thinks it could be applied elsewhere sooner than later.
“We’re in the discovery phase right now… but I think eventually as we go down this path… we’ll find other areas of the business that we can apply it to,” he said.
For now, the application is being developed for an internal client, and while Moghimi cannot disclose any names, he does say the Toronto LOFT team, which is based in the city’s MaRS Discovery District, will continue leading the project.
“I think Toronto’s always been a lot more forward-thinking in its deep learning space… the students that come out of university in that space are top-tier – so it was kind of a natural fit in bringing this mission to life,” he said.