Jim Lewis, SEI

As discussed in the first part of our discussion with SEI on the potential of natural language processing (NLP) – click here to read – the company is currently in the early stages of exploring how NLP might enhance the way staff and clients interact with data. The aim of this initiative is to increase the overall level of engagement and grant users the flexibility to choose their preferred method of navigating information. 

To that end, SEI is currently developing a bot framework to expand what Jim Lewis (pictured), Director, Solutions Strategy & Development, Investment Manager Services Division, SEI, refers to as the ability for users to configure their own triggers and actions.

In the first part of the discussion, Lewis explained: “You need some way to trigger an event and some way of delivering it. What channel do you trigger through? Alexa? Google Home? A chatbot? Facebook Messenger? IoT device?”

“The bot framework then performs an action based on the parameters the user sent in the request and can deliver the results through same channel or via another means. i.e. today, the user might ask Alexa to run a specific performance report and send it to someone’s email account” says Lewis. “Then tomorrow, the same user could use IVR (Interactive Voice Response) via their cell phone to send the same report to an FTP site.”

For financial services, a company looking to capitalise on NLP technology, designing a bot framework that sits at the heart of its business model allows for building new channels for triggers without writing additional code. Granted, some work may be required to extend the framework to handle new data sources, but the investment into the core technology is a one-time event.

“The framework knows how Alexa, for example, delivers information and we can create an Alexa specific pattern that can be transformed into a common data set,” explains Lewis. “It is really thinking about how to design these things to future proof SEI, as well as our clients, as new NLP technologies come on line. A well thought out framework design allows us to integrate new tools quickly without affecting existing processes. It will also allow our developers to leverage these technologies without their having to learn how to interact with all of these cutting edge technologies”

That approach is common sense but it really takes discipline to make sure the design works for all possible scenarios, not just for today, but for the future.

“We are hoping to begin rolling out the bot framework to internal users in 2019 and based on the success of that pilot program put together a plan to roll it out to targeted clients,” adds Lewis.

The most exciting part of NLP innovations are the drastic improvements people will see in their personal and professional lives. This is already a reality for those that have incorporated these technologies into their daily habits. Generation Z is leading the charge, enhancing their lives using these technologies, but it also opens the door to Boomers who are not comfortable with GUIs, but would prefer having a conversation. This trend is only going to increase in momentum, to the point that relying on older delivery methods like call centers and web portals will make firms seem dangerously outdated.

One possible use case for NLP is an employee asking the bot ahead of an investor meeting what their client’s asset allocation is across their portfolio, broken out by sector weightings. They could also ask the bot to send a report, such as performance, or sector exposure, in preparation of the meeting.

The bot would act as a virtual assistant, akin to having someone sitting next to you performing these tasks; but without the procrastination or end of day ‘sigh’ that they have to do something just as they were getting ready to leave the office. With a bot framework, the result would be instantaneous, without fuss, and more accurate.

“Here’s a use case I often reference when talking to people about where this could end up,” says Lewis. “Say you are in a conference room with 20 colleagues, deep in a discussion. Halfway through, someone asks a question and one of your colleagues dashes out of the room to print copies of a particular report, then brings them back in. By then, the conversation has inevitably moved on.”

“Now imagine having Alexa (or an equivalent) on the conference table ready to respond to any questions.”

“Rather than physically leaving the room to run a report, you simply ask the question and immediately get the response. You could even ask Alexa to email you the report so you can bring it up on the screen letting the meeting move along seamlessly. That is the potential of NLP.”

Security is an obvious point of concern when thinking about this man/machine interaction, especially when financial data is concerned. As NLP applications, and their underlying machine learning capabilities, advance, the defense barriers to guard against any security risks need to keep pace.

At SEI, the first step in their authentication process is to have the user register their device. Next, when the user initiates the voice assistant to query an application like SEI Trade – a solution that streamlines the end-to-end processing of investor activities – they will be asked for their personal pin.

“Then, we use a third party application that sends an authorisation message to the user’s mobile phone asking to approve the use of SEI Trade; this challenge can be confirmed or rejected using the biometrics (fingerprint or face ID) on the phone. This is an example of true multifactor authentication since it is something you know (personal PIN), something you have (the phone), and something you are (the biometrics.)”

“We’ve also looked at using voice prints, but the technology is not yet fully evolved,” says Lewis.

Device registration and PIN confirmation, coupled with biometric authorisation is already a good basis for maintaining security around NLP devices and once the technology improves to recognise individual voice patterns, it will add another ring of steel.

Moreover, the way we type our passwords, the way we move around the screen with our mouse, or even our verbal cadence can be used as behavioural insights that can all be combined to take data security to a whole new level, and stay ahead of bad actors.

To optimise NLP technology, SEI (or any financial services firm) has to ensure that it provides data in a way that empowers users to help their clients in a secure manner, without compromising the efficiencies gained.

SEI fields many questions from clients and prospects on market trends in the industry. It is increasingly becoming an extension of its clients’ technology roadmap.

“We can help our clients use their data more intelligently with NLP and machine learning solutions. We tell clients ‘Here are some interesting patterns we see’, and give them insights ahead of time before something becomes an issue. The end state we are working towards is being able to alert clients of problems before they occur instead of working through them after the fact.”

“You have to make sure you are training your models to produce repeatable results ensure security is robust, and you have to look to continue extending data sets for more accurate and detailed answers from your analytics, or from the responses you get from your bots.”

“Applying analytics on top of NLP capabilities…to my mind, it’s unbelievable where we might end up, and the value we can extract; it’s going to take time, but boy is it going to be exciting!” concludes Lewis.

Read more about SEI's operating platform for investment managers. Select preferred format: US/UK

Author Profile