Deriving Layers of Value from GenAI Applications


Insights on how one financial services company implemented and is using GenAI to increase employee productivity and better serve its clients.

Roger Burkhardt, Broadridge’s CTO for AI and Capital Markets goes beneath the surface of the most common GenAI use case and explores how the same technology can benefit different user groups and different workflows with human-centric adaptations.

We caught up with him after an executive panel at the Open Data Science Conference East 2024 and discussed how Broadridge uses GenAI across its businesses to increase productivity, accuracy, and allow associates to spend more time on valuable and satisfying work.

(See Roger’s bio below.)

RTInsights: To gauge the impact of GenAI on enterprises, we’d like to hear your observations from customer conversations before and after GenAI’s appearance. What’s your assessment?

Roger: It’s been transformational at all levels. We’ve observed it in our personal lives. Our younger relatives are showing us how to use new technology to do things. There’s tremendous awareness of AI given all the coverage it’s received. But even in the world of sophisticated financial institutions, it’s really been a game-changer. And I think this is a mass democratization event akin to the Netscape browser.

RT Insights: Say more about the browser analogy and its role in a mass democratization event.

Roger: Before the first browser, you needed to go to the library to look at a lot of information. Now, all of a sudden, we had access to all this information online, instantly, and over time, access to the whole world. Now, some of that information was fatally wrong. But it was a mass democratization event, and that has highlighted the power of AI, too. That affects a C-level executive at a financial service institution as much as it does anywhere else. There’s been enormous interest in the very powerful new large language models that are available, and they enable us to operate on forms of data far more efficiently than in the past.

I first started working with AI back in 2001, mostly with numeric data and analyzing trading patents for manipulative behavior. But we were also looking at it in chat rooms–we were actually doing social media analysis. But that was not the norm. The norm was that most of AI was predictive and based on numbers. Now, AI allows us to text with music and video, right? It’s very powerful, and it’s very general purpose. The other thing about these models is the P in GPT. When I’ve been talking to groups and executives, I like to point out that P stands for pre-trained. The fact that OpenAI was pre-trained means that you don’t have to spend months and many dollars to get a model for your job. The other side is that it’s general and might not be the perfect thing for a particular job. It reminds us of the ultimate Swiss Army penknife. But sometimes, if you’re actually looking to tighten up the screws, you don’t use your Swiss Army knife; you use a screwdriver.

The impact of GenAI was a galvanizing effect on financial institutions that have large investments already in AI. And this technology had the potential to really change the world–something that financial institutions can’t ignore. If we ignore this, we may get run over.

See also: Beyond the Buzzwords: A Deeper Look into GenAI

RTInsights: Would you say financial services companies are set to adopt GenAI on a large scale?

Roger: It did create some level of concern and fear about some of the issues and risks, such as hallucinations and the potential for harmful bias. Is my data going to be mined by some of these vendors and taken right out of my control? Where does my data go, and what do people do with it?

It’s interesting how the different firms reacted. Some banned access to these technologies, but other firms said that they had to move forward and learn how to do this safely. At Broadridge, we decided we’re going to learn by doing, we’re going to do it safely, but we’re not going to sit back and wait for a new leader. So, let’s learn by doing. Let’s do it in a way that has broad-based access, but let’s do it in a way where access has some controls, so we know that if Roger starts doing something inappropriate, we can track it, we can warn him, we can take action, we can apply it–and in a consistent way across the whole firm. Brokerages have many different businesses, some of which we acquired over the years. In our case, we centralized the service data and decided to have one AI platform. It’s constantly evolving, with new models coming in all the time. With one platform, we have one set of controls, which has allowed us to move really quickly and get what we believe are very safe products in the hands of our clients. One element of the safe path is that the data that we serve up with ChatGPT is all the data we know and have curated.

See also: Use Case Milestones on the Path from Data Foundations to GenAI Applications

RTInsights: How is the GPT LLM being used at Broadridge?

 The model is used to help users navigate through complex data. We do not get the data from the internet. We have curated the bond data ourselves. After all, we’re in the business of securely managing complex data, whether it’s post-trade transactions, which are very sensitive to who traded what for which price, or whether it’s free trade. Which bonds are in liquidity today, which bonds are similar to other bonds, etc? So basically, we discovered that we have the safe path that we can now double down on, right? Because we know the safe path for data, we can give relatively rapid approval with light governance because we know the data has the right access controls.

See also: Smart Talk Episode 2: The Rise of GenAI Applications and Data-in-Motion

RTInsights: We’re seeing that many companies never paid much attention to getting their governance practice in shape. They’re left at a disadvantage. The GenAI use case you described is internal. Can you describe the process for a client-facing use case?

Roger: We characterize two separate areas: client-facing products and services, and the internal tools. Internal tools are very straightforward, and we give all of our associates access to LLMs for their personal productivity. So for example, the ability to take this meeting recording, and then transcribe and summarize it. Within a few weeks of launching we had 2,000 meetings summarized–quite a few hours of work were saved.

We keep a human in the loop where you’ll get a rapid first draft that’s good to work with, but you have to be ready to make sure it represents the sense of the meeting. But it’s five times faster to edit and review than write. We’ve increasingly found that providing an associate with prompts that they can just use out of the box is encouraging. Increasingly, we’re going to have a kind of inner-source model where people say, “Oh, I’ve got a better version of the prompts,” that they can then share.

One of the areas that we invested in is emails for our managed services business. We get tens of thousands of emails and now automatically extract the key information that is going to instantly cause the right system to take action. Instead of wading through mundane details, most of which can be ignored, the operations staff can just zero in on where they provide real value. GenAI is enabling that now. So that’s an example of a very, very fruitful area to improve productivity, improve quality, and also give people a better experience.

Another example would be in software development. The one that people tend to gravitate to first is generating code. But we found that there are many other areas, which are really powerful such as creating documentation which is a great way of addressing legacy assets that weren’t perfectly documented.

We’ve used GenAI to create behavior-development test cases. Instead of writing one long test case, we broke the process into steps, and GenAI was able to really accelerate five of them. [When we deployed] we gave the users 5 “best-practice” prompts for sections of the workflow. it’s like prompt engineering on steroids, if you will.

Users are more inquisitive since the pain of asking your first set of questions is minimal. It encourages you to dive deeper. That ability to interrogate the chat is a good fit for the way human brains work.

RTInsights: You can think of it as a puzzle, working out how to get the information you want. You’ve mentioned several ways that companies can take GenAI capabilities further within well-known use cases. These are great examples that others can try out for themselves, not by increasing the technical investment but by spreading the use across the organization and working on aligning them to existing workflows. You called these “power tools” at one point.

One more question. What advice would you give to a company that is struggling with what’s a good way to use Gen AI to amp up business? 

Roger: Yes, I think the key learning from our experience is you learn by doing and iterating. And so you need to find an environment, which is safe–you’re not going to dive into something risky. Start small, and then scale up. You could even start with one of your employees. To scale, find five things that are hard to do. Use that success to get traction and go forward in a focused way to get the maximum benefit from GenAI.

Roger Burkhardt is a seasoned CTO, CDAO, and CEO with twenty years of experience leading technology transformations. He has a track record in mission-critical trading systems technology and accelerating value capture from AI. Prior to joining Broadridge in November 2021, he spent four years as the co-lead of McKinsey’s AI practice for financial services, driving large-scale programs in the top 4 financial institutions in the US and Europe. Before McKinsey, Roger was the CTO of OpenLink Technologies (now part of ION), the largest provider of energy and derivatives processing platforms. Prior to this, he founded the CDO role at the NY Federal Reserve Bank and was CEO of Ingres Corporation, a data and analytics company in Silicon Valley. Roger was CTO and EVP at the NYSE, where he designed and led the historic digital transformation from floor-based to electronic trading.

Elisabeth Strenger

About Elisabeth Strenger

Elisabeth Strenger is a senior technology writer at

Leave a Reply

Your email address will not be published. Required fields are marked *