Morgan Stanley is rolling out an AI chatbot to research clients early next year. Wall Street thinks the tech could save $8 billion annually.
- Morgan Stanley has developed an AI-powered chatbot to help its workers sift through the bank's research.
- AskResearch was originally created for internal use by analysts and sales employees, but the bank plans to offer it to clients in early 2020 via a mobile app.
- The chatbot is the latest effort by Wall Street to create AI-powered virtual assistants to cut down on menial work for employees and clients.
- D'Arcy Carr, global head of research, editorial and publishing at Morgan Stanley, told Business Insider that the chatbot helps find research efficiently and can also uncover new data.
- Click here for more BI Prime stories.
Wall Street loves to speculate about the drastic ways artificial intelligence could change how it does business, but the main goal of the technology is a simple one: to help people work more efficiently.
Research is one area ripe for automation, given the sheer volume of data that banks produce and how time-consuming internal and external client requests for information can be. Morgan Stanley sees a solution in AskResearch, its AI-powered chatbot.
Since an internal launch last year, the bot has evolved from answering simple questions ("Which analyst covers Apple?") to more complex ones ("Which stocks did Morgan Stanley upgrade last week?"). Morgan Stanley hopes the chatbot will be offered to external research clients early next year through a mobile app.
"Our vision is really, how do we get the most relevant content out to the people that will get the most value from it?" Eden Kidner, head of research technology at Morgan Stanley, told Business Insider. "We decided the most natural way for us is to adopt the flexibility of a virtual assistant or chatbot where we are not restricted by the traditional user interface."
What is your bull case for Apple?
Often pitched as a virtual assistant - like Apple's Siri or Amazon's Alexa - chatbots are seen as a way to help companies cut down on mundane tasks. Bank of America's "Erica" is already helping Main Street manage its accounts, while PayPal is hoping to reduce costs by automating customer service requests. A report by Juniper Research in 2017 estimated chatbots could deliver over $8 billion a year in cost savings by 2022.
Kidner said Morgan Stanley started by developing a list of the most common questions its analysts and sales employees ask when searching for information from internal research content.
Natural language processing was applied to questions like "What is your bull case for Apple?" and "What is our US GDP forecast?" to allow the bot to recognize the prompts and provide the correct information. The team could then build from there after going live by analyzing other commonly asked questions submitted by employees, Kidner said.
The bank spent the past year improving the bot to answer more complex questions ("What is Morgan Stanley's view on gold?"), which involve pulling unstructured data that sits within analyst reports. Kidner said the key is for the bot to understand the intent of the question, making it easier to track down information.
D'Arcy Carr, global head of research, editorial and publishing at Morgan Stanley, told Business Insider the new feature also opens up research that might have previously been missed.
"You have this historical archive sitting in a library and there is so much value embedded in it, but traditionally it has been hard to unlock that value because insights and data are fixed in monolithic PDFs," Carr said. "We have seen our internal readership spike pretty substantially over the course of the last year."
Solving big-ticket problems
Kidner and Carr didn't give more specific statistics on improvements tied to the chatbot. Kidner did say, anecdotally, that employees are spending less time sending emails, since they can use AskResearch directly in Outlook.
Longer term, the chatbot could lead to the bank reallocating resources currently used for menial work, Carr said.
"Can we get rid of thousands of small inefficiencies? Can we hire another analyst to cover this? Can we get out on the road to meet more people? I think there is a big opportunity here," Carr said.
"There are tons of inefficiencies in our business still, and the more skills that we add to AskResearch to help analysts manage their day, the more time that can be spent doing the fun parts of their job, like talking to clients and companies and doing research."
Carr also sees the chatbot benefiting research clients outside the bank, such as large investors. In late 2017, the bank revamped its research portal to improve how users searched for information. The change resulted in a significant uptick in readership, Carr said.
But not all investors are optimistic that chatbots will help them save time. Matt Daly, head of credit and investment research at $141 billion asset manager Conning, said AI does not totally solve the problem of being inundated with too much data, and that firms should focus on better curating what they send out in the first place.
"If I go out and look for the information, I feel like I can navigate pretty well to find it," Daly said. "Say I have 10 research portals available to me, how do I know what to go out and find unless they are pushing that stuff to me? And then how do you scrape through all the stuff to actually add value? Just culling through and managing that information is a challenge."
Still, Morgan Stanley sees AskResearch as a huge potential help inside and outside the firm.
"The way we think about it is an expert looking over your shoulder. Whatever you are doing, whatever workflow you are currently in, you have the opportunity to get the contextual information you need," Kidner said. "The idea was to remove the friction in getting data from our traditional monolithic applications, and expose it through a more natural interface embedded into the analysts' day-to-day workflow."