Artificial Intelligence Done Right. How Panda Incorporated AI into its Business Intelligence Solutions

An Interview with Omri Nir,  Sales And Business Development

Artificial Intelligence has been one of the biggest stories of 2023, but Panda has been working on A.I-based tools for the brokerage segment for quite some time now. What led you to focus so heavily on A.I before there was a demand for it and have you found it to be a fruitful endeavor? 

At Panda we like to joke that we were into A.I before it was cool! Nowadays you’ll find random companies paying lip service to A.I in order to make the business seem more relevant, but our journey with A.I started purely from a functional point of view. We observed all the progress that was being made in machine learning, particularly in the open source space and we started realizing that these tools could be powerful game-changers for our own industry.  

When you know an area so well, for instance the internal structure and workflow of online brokerage businesses, then you know exactly how new technologies can be utilized to create efficiencies that were not previously possible. I think our advantage here was the combination of being so deeply embedded in brokerage culture, combined with a genuine fascination with new technologies. Our engineers are true geeks in this regard and have been invaluable when it comes to thinking outside of the box and bringing new innovations to market. 

As far as whether the focus has been fruitful, I would say it definitely has. Our flagship Panda CRM product was already one of the most flexible business intelligence suites for online brokers. Our continued focus on A.I has only made this product more powerful and allowed brokers using it to compete with much larger teams. 

What does the process of experimenting with a new technology such as A.I and then bringing it into your core line of products look like? 

In business development there’s always the question of whether you’re straying from your core competencies by trying something new. I would say the key is to start small. Don’t be too ambitious from the jump. Try and see whether there’s a glaring pain point among your customer base that new technologies can go some way to solving. Have a small team work on a proof-of-concept and see whether there’s anything there to get excited about.  

For our customers the obvious pain point was document verification, which is a huge drain on human resources at the best of times, but can quickly become unmanageable during times of surging interest in online trading. Understanding that image detection technology had become sufficiently advanced some years back to really take a big bite out of the problem, we set to work on harnessing Google’s Vision AI algorithm to provide back office and compliance teams with a non-human helping hand. 

The result is a document verification module in Panda CRM that can not only reject document submissions that fail to meet the grade and automatically request resubmission, it can also “read” the supplied documents and compare them to the information the client has already provided via the registration process. In this way, incoming KYC/AML documents can be graded according to how likely they are to pass, allowing back office and compliance teams to batch submissions most likely to do so and thus vastly improve onboarding efficiency. The success of this module showed us that we were onto something, and that further resources should be devoted to bringing A.I into other areas of brokerage tech. 

What is it about the online trading segment specifically that lends itself so well to the incorporation of A.I and machine learning elements? 

If you look at your average online brokerage, it’s really a big data-generating machine. You’re dealing with vast amounts of traffic and browsing data from incoming leads prior to registration, and then you have a myriad of data pertaining to platform and client area activity, as well as any touchpoints the customer may have had with various brokerage departments. Previously this data mostly went to waste as brokerages didn’t have the systems in place to do anything with it. Nowadays, though, we’re able to use this data to inform how brokerage team members organise their activities. 

An example of how we’ve been able to do this within Panda CRM is our Next Call AI system, which provides sales and retention teams with top priority next calls to make based explicitly on customer actions. In this way a new client registration, login, deposit attempt, card decline, withdrawal request, and much more besides can be used as an opportunity to touch base with the client and provide pre-emptive customer support. 

These event-based triggers are completely customizable so as to reflect each individual business’s priorities. Not only does Next Call AI ease customer support workloads and improve customer waiting times, it has also been shown to drastically increase customer re-deposits as well as the rate of successful outgoing calls made by brokerage team members. 

Are there any other areas of the online brokerage business that you feel are particularly ripe for A.I optimization?  

Definitely. As you can see from my description of Panda’s forays into A.I above, we started with the low hanging fruit of image and optical character recognition for use in document verification, we then moved into the more dynamic area of real-time client activity with Next Call AI, and have since moved on to more technically challenging areas. 

I think call center activity is another highly rich source of actionable data that is currently being wasted. Perhaps this has to do with the nature of audio. It is fleeting and, even when recorded, it’s costly to store and difficult to manage, unlike images, text, or event triggers as in the case of Next Call. 

This is what led us to working on Call Control, another A.I-based module within Panda CRM that focuses specifically on call sentiment. The way the module works is by analyzing call recordings that are more than two minutes long. These recordings are transcribed and processed by a natural language processing algorithm, providing each interaction with a sentiment score. The system has proven to be extremely useful to call center managers who can use the data it provides to improve the performance of their respective teams.  

Beyond sentiment, Call Control also provides a variety of highly useful data regarding each customer interaction, such as the operator’s talk-to-listen ratio, a “patience average,” as well as flagging when the operator talks for too long, or when a client has shared a great deal of uninterrupted information to the rep in question. Beyond its obvious usefulness in call center training, the system can be used to “listen” for specific keywords, which can then be used to determine the effectiveness of marketing campaigns, or even to conduct market research on client preferences.  

Further, in order to help call center agents stay on track and communicate more effectively with clients, we’ve recently introduced Call Script in Panda CRM, a feature that provides agents with scripted prompts that are individualized to the client, and thus help call center operators stay on track and provide a more tailored service.  

Considering that you have been actively involved in bringing A.I technology to online brokers for some years now, how do you view the overall A.I landscape as well as the future of the space? 

As far as what we’re currently doing at Panda, I think we’re only getting started, and I anticipate that brokerages in the future will rely much heavier on non-human intelligences in order to guide the efforts of human teams. It’s not so much about replacing humans with software because the human element is a crucial part of doing business, rather, it’s more about making human teams smarter by giving them access to the collective knowledge of the business itself. 

As far as the broader landscape is concerned, I think there’s a parallel to be drawn between the early days of the Internet and the current situation in A.I. The interest is there, the technology has proven itself with the recent release of several large language models that have taken the world by storm, and there’s a tremendous amount of investment currently taking place. Not every A.I investment will lead to the killer app that everyone seems to be striving towards now, but this acceleration in interest and investment will definitely usher in a new era of progress for humanity, just like the world wide web did in a previous generation.