Come on in and sit back relax, you’re listening to Episode 199 of the WealthTech Today podcast. I’m your host, Craig Iskowitz, founder of Ezra Group Consulting and this podcast features interviews, news and analysis on the trends and best practices all around Wealth Management Technology.
My guest for this episode is Susan Emerson. Susan is Senior Vice President product AI analytics data at Salesforce. This is timely. I was excited to have her on the program, considering all the work we’re doing around AI as a group, all the articles we’re writing, all the webinars we’re conducting and all the town halls and CEO roundtables. We’re running all around AI. It seems like it’s non stop. So having Susan on was great to hear some of the things that are going on in Salesforce.
Susan has been at Salesforce almost 12 years, and she has spent the last 25 years in B2B technology and software. She has been part of three M&As, two successful IPOs and was on her seventh startup when acquired by Salesforce in 2011. So we’re going to get into the podcast, some great stuff we talked about, including Salesforce as AI strategy, some interesting AI, machine learning use cases, and some things that the Salesforce AI team is building on all good stuff.
But before we get started, if you are an executive at a broker-dealer, enterprise RIA, family office or a TAMPs, your tech debt is holding you back. Your old software platforms are rusty and falling apart and they need either a complete overhaul or to be replaced entirely. Your disparate systems don’t communicate with each other and it’s driving your operations staff and advisors crazy with manual processes and other errors.
If this describes your company, you should run, not walk to our website, EzraGroup.com and fill out the Contact Us form on the home page. Our experienced team can evaluate your technology ecosystem, deliver targeted recommendations, optimize your existing systems and operations, or run an RFP or RFI to help you implement new software to help take your firm to the next level. Please subscribe to this show wherever you listen so you don’t miss an episode. Now let’s kick this thing off!
- Analytics, AI, and Data Journey
- Launch of CRM Analytics Product Line at Salesforce
- The role of AI in predicting customer churn and flight risk
- Integrating Analytics and AI
- Statistics on Salesforce’s AI Initiatives: Predictions, Patents, and Research Papers
Craig: I’m excited to introduce our next guest on the program. It is Susan Emerson, SVP product analytics, AI, and Data at Salesforce. Susan, welcome.
Susan: Good morning. Thank you.
Craig: It is a great morning. It’s a great hot morning. It’s sweltering here. How’s the weather up by you?
Susan: Swelteringly hot and looking out the window right now, the infestation of spongy moths that have now stopped eating the trees and are into their next phase of development.
Craig: That’s a whole different podcast. Where are you calling from?
Susan: From Columbia County, New York State.
Craig: And that’s up I don’t know that specifically, is that by Woodstock?
Susan: No, this is up maybe two hours due north in New York City.
Craig: Got it. But beautiful country up there.
Susan: Beautiful rolling farmland right along the Berkshire Mountains. A little bit of my pandemic craziness to get out of New York City.
Craig: Smart. That was smart. All right. So let’s kick things off. Can you please give us a 30 second elevator pitch for your role and what you do at Salesforce?
Susan: All right. Well, I’ve been at Salesforce for 13 years, and for the last 10 years I’ve been part of the analytics AI and data journey. We launched a product line that is now known in the marketplace as CRM analytics about 10 years ago. Along the way we use that as a foundation to bring machine learning and AI to the foreground. Then to use the canvas of the Salesforce different industry clouds to deliver out of the box experiences that are both visual and predictive in nature and provide great user experiences for our Salesforce users.
Analytics, AI, and Data Journey
Craig: One thing I like about what Salesforce is doing is that every enterprise client we work with has, so it’s already in house. It seems like it would be very easy to bring in these AI tools, because I was reading a statistic about Ai and wealth management, and it said, well, most strategy executives in wealth management see opportunities in AI. More than 80% of firms are stuck in the proof of concept phase. They just can’t get out of that. How is Salesforce leveraging AI and how can it help the wealth management vertical?
Susan: I’m actually going to take a quick step back before I answer the AI question specifically. Maybe just address in terms of Salesforce’s utility in general to the wealth management space with financial services cloud. The foundation of that is a great set of capabilities for wealth managers to organize their book of business, to document all their interaction with customers and foundationally to make sure that they’re driving engagement with customers, so that customers are well acquainted with all the different financial services and products that are available to help them meet their investment and other financial goals. And to drive those interactions essentially to deepen the relationship and for the wealth management perspective to help drive any net new assets and AUM to the firm.
When you think about it, the foundation like AI is going to be one of the tools to do that. My step back to that question before going direct to AI is to also think about the visual insights that we can bring to the foreground with a book of business. For example, if someone has a large book of business and the day in day out question is who do I call and why? What is my cost to serve? How are customers doing against their financial goals that would warrant a call and a discussion about their goals? A lot of those questions can be answered very quickly and easily, which is a visual way to look at the book of business. I start there because you can get a lot of value off the table just by having very directed and curated understanding of that book before you even get to the AI questions. I usually start there, that’s where a lot of people start their journey is just using these visual capabilities to explore that.
Susan: Then the next path is like a lot of people could say in words what might drive their engagement strategy. We want to work with customers and make sure everyone gets a touch every 90 days. Or we want to work with customers and we want to make sure that we’re engaging with people that maybe they have open cases a sentiment that’s trending the wrong way. Maybe assets under management or trending down. These are all just things that you could describe that a nice visual way to get with that list of customers that would drive that engagement. There are all sorts of things that you can do to help drive the main goals of like interacting with customers, driving the financial health and wealth of your customers and bringing new assets to the firm before you even get to the big AI questions.
Craig: Now, the AI capabilities we’ve been in market since mid-last decade with machine learning type predictions in our ecosystem that help drive those engagement decisions. Things like, are there customers that are flight risk? Are there customers whose sentiment is trending the wrong way? Are there customers with a CSAT score that need to come to the fore of my list? Are there customers that have a high propensity to what engage with us around a new product or asset class? All these types of predictions that have been in market with us for quite some time that help the advisor answer those questions, who needs to call and why.
When we get on the phone or have the meeting, what are the types of things that we should be discussing around their financial health? That’s all stuff that’s we’ve been up to for a while, and then with the introduction of all the large language models, there’s a whole new set of capabilities that we are working on to drive even increased productivity around those things. I don’t know if we’ll get into conversations about GPT and generative AI as well, but from the predictive side, we’ve been doing that since 2016 at Salesforce when we launched our whole Einstein product line.
Craig: I am aware. We were very excited by that when it came out. We follow the industry closely and Einstein was something we keep an eye on, but you covered so many areas there with just one question.
Craig: Let me break it out.
Susan: Let’s unpack it.
Launch of CRM Analytics Product Line at Salesforce
Craig: Let’s unpack it. Let’s decompose this. Driving engagement strategy. You talked about sentiments that are trending the wrong way. I think these types of capabilities are the superpower of AI, and that once these permeate much of the advisor’s systems, it will greatly increase their productivity, as you mentioned and the way they operate. Can you talk more about how the AI and Salesforce develops these sentiments, like you talk about flight risk, how does that work? How does it do that? What kind of data does it need, and how does this data, the results surface themselves?
Susan: It could be a variety of things. Many organizations do surveys or have different ways where they actually collect voice of customer details. In some cases, it’s just reading the screen. Customers respond with a survey and then you’re able to quickly calculate that, like, here is the CSAT, or here is the NPS, it’s reported by the customer. We can also, if we’ve got data that is predictive in nature, also use those things to generate potential predictions based on those facts. That’s one thing, just reported numbers and reported scores or maybe using some machine learning to derive that.
Then in CRMs you get a lot of unstructured data in terms of meeting notes and call notes and things like that. Some of the machine learning capabilities that we have there, from a technical perspective it’s things like using text clustering or sentiment analysis to take those interactions and then report the sentiment score. I think what is most helpful is to actually see the sentiment score change over time. Those types of things on sentiment.
Craig: Can you give me an example of sentiment? What is a sentiment?
Susan: Positive, negative, neutral, that type of thing.
Craig: Positive, negative, neutral.
The role of AI in predicting customer churn and flight risk
Susan: Positive, negative, neutral, and then being able to trend that over time, just so you get a sense of how that relationship is going. Then on the churn, because you also asked about churn, this could be a lot of different data points and maybe a comment about the nature of the machine learning environment that we have to generate that. One of the things that it’s very good at is rapidly telling you what data that you’re using is actually predictive in nature. A lot of times you might have a lot of rows and columns of data and you think it’s all valuable, but not all of it might have predictive signal. In doing something like churn, you might have a whole bunch of different data hunches or hunches about data that you think will be helpful, like interactions that you’ve had, like activities calls, like the performance of the portfolio. There might be a lot of different data that could be sourced both from the CRM and your transaction systems.
The way it works is, like you basically would tell the system in with a couple of check boxes. I would like to minimize churn, and then it will tell you what data elements are actually predictive in nature in that. Then give you a lot of additional support in terms of how to engineer your data for an even better prediction. In terms of helping you do things like protect against leakage or looking at multicellularity or, looking at data elements that just aren’t even impactful for a prediction, so get them out of the way. There’s a lot of guardrails and support for improving these predictions. Churn could be generated by many different data elements coming both from Salesforce and other back office systems to do that.
Craig: Churn is such an important statistic in wealth management. As with as would probably be in many industries, you want to keep that churn to a minimum.
Craig: But I find it fascinating that when I check off the box, I want to minimize churn that the system then tells me which data’s important.
Craig: I imagine people have lots of hunches. They think what’s important and then they find out, what’s not important? How does that change the way they operate when they realize the data they felt was important isn’t?
Susan: Well, it allows them to go faster because you can get the stuff out of the way that’s not predictive in value. It just becomes more of a rapid iteration cycle would be one thing. The second thing I would say is that in a CRM, you can populate a CRM with systems or record from your back offices in terms of transactions and accounts and that’s often very complete data. It’s very binary. There’s account with a number, you move it in and go. A lot of times the CRM other components are human driven. I met with the customer, we talked about this.
When you have a large organization of people all interacting with the CRM potentially in different ways just in terms of style and level of completeness, one of the things that is often said is, “I’m not sure I know fully the health of that data, the completeness of this data,” because it is in many cases human generated. One of the byproducts of having this machine learning that tells you what data has signal and is very useful to curate and to look after, is that you can start to get more organized around. You should fill in these data facts because when we have these data facts. Your predictions are going to be even more important to you in terms of potential outcomes. There’s that benefit as well. When it’s the human and the loop that’s actually is updating the CRM.
Craig: Most CRMs rely on humans to update them.
Craig: That’s probably where we used to call that many years ago when I worked in tech support. When someone would have a problem, we’d say we’ve isolated the problem. It’s between the keyboard and the chair.
Susan: Yes, that’s right. That’s a good one.
Craig: We can talk about what is —
Susan: Then the other thing I would just say, just in terms of AI in financial services. I mean, many years ago I sold OTC derivative trading systems to Wall Street, and I learned quickly that the customer’s data science team is always smarter than yours. I’ve carried that throughout life, and so when I think of like a large financial institution and the resources, the quantitative and the data science resources. That’s a rich pool of very talented resources and investments in product areas that people have for their corporate AI stacks.
What I often see with the work that we do is that there might be things that are already being predicted by other tools inside the four walls of a bank or a financial institution. If we can get those into our environment, we can help be the endpoint for them to get them in front of the wealth manager or the banker. One of the classic examples would be like on a company’s website, there might be a lot of interactivity. People on fun fact pages all sorts of data about time on page, what they’ve looked at. How long they’ve looked at it, and maybe the bank’s corporate AI team has built like a next best product or a propensity to buy. I would like openly embrace that and say, let’s get that into Salesforce, because we can help be the endpoint for that because it’s most useful if we give it to the advisors.
Susan: But then what I would go on to say is that the machine learning that we have, and I already talked about some of the nice attributes in terms of its speed and its interactivity and it’s sort of no code vibe, but it also is outstanding. If you’ve got a process that begins and ends in Salesforce like a lead, that we open a lead that we close. An opportunity that we start an opportunity that is won or lost. An onboarding process that begins an onboarding process that it ends, all that interaction is in Salesforce.
In the case of that propensity to buy coming from the bank’s AI system, we went and would do propensity to close because then we own all the interaction steps. It’s just a way of I think embracing like best in class and embracing, fundamentally what we’re trying to do is help the wealth manager engage with customers and to help the customers with their financial wealth goals and to drive AUM net new assets to the bank. I find that a nice continuum of embracing things that might be already in place and adding additional value add on top in terms of the whole Salesforce journey plus our machine learning journey on the Salesforce side.
Integrating Analytics and AI
Craig: Talking about driving net new assets there’s a number of out the box applications for Salesforce SFC. Can you talk about how those can help advisors, and are any of those AI driven?
Susan: Yes. I’ll talk about that use case in terms of driving that new assets and then the out of the box stuff. One of the one of the consistent use cases that we see is that firms will have a protocol that every customer is owed a call and a touch, an interaction every 90 days. We’ll measure that and we’ll help people understand who those customers are that need the call, but then if we add some predictions to it, right?
We might add predictions in terms of propensity to grow. What is this customer’s propensity to grow net new assets? That might be a prediction. It also could be something as straightforward as what is held away with another financial institution. Sometimes you don’t have to go as far as the machine learning component, you might have data elements that would help you drive that focus. Then adding a prediction on top of that, like what are customers that would have a high propensity to go through a full financial goal planning, like whatever the term would be at the institution. Who’s going to put the time and attention into doing a full wealth goal retirement investment plan and then grabbing all those predictions together.
From a product perspective within, I’m with the CRM analytics team when I make this statement, we started building out of the box applications back in 2016. What they serve as is the following; technical accelerators. You can pick these templated packages and you can install them. It knows the Salesforce data model, the user experience, the predictions, and you can install them and go. Now more likely in large financial institutions people like to make it their own, whether that’s branding or whether it’s different use cases, so we fully expect that people will take these things and modify them. But in addition to doing the technical acceleration in terms of creating that analytics data set or a predictive model, the other acceleration is the focus on the best practices in the use cases.
Because if you don’t have to sit there and what I call, stare at the tyranny of the blank screen, it’s a huge accelerator to align a team of people in terms of what they should build first, second, and third. It can serve as a blueprint for how to engage. We’ve got a very rich set of things in financial services cloud. Everything that focuses on the wealth advisor and their book of business to other applications around appointments and scheduling to other applications that help the back office around complaint management. Both in terms of how triage and engage with those things and then rip off a sheet of paper and give it off to a regulator. All these things become validation points, best practices, technical accelerators and then user experiences that you might want to mimic and make your own.
Craig: You said something that I just highlighted.
Susan: What was that?
Craig: Staring at the tyranny of the blank screen?
Craig: Can you explain that?
Susan: Well, with analytics and data and AI, you’ve got a pile of data and you’ve got some experience that you need to deliver. And that experience that you deliver needs to be done in a very curated way that drives activity, right? It needs to be in the path of the advisor because you’re inspiring them to do something with their book of business and their customers. The path between I’ve got all this data and I’ve got outcomes I’m trying to drive, that is a blank screen. Anything that we can do to accelerate that is value. Does that make sense? It’s always better to edit than to start from scratch, that kind of thing.
Craig: No, it’s the blank space in between, as you mentioned, data and the outcomes you want. That’s something we work on with a lot of companies is they have a lot of data.
Craig: They know what their outcomes should be, but they don’t know how to go from A to C that they’re missing the B in the middle.
Susan: Yes. I mean this isn’t an AI thing, but it’s more a I’ll just call it a design thinking thing, and particular in the realm of analytics. Analytics is a category that’s been around literally for decades. So you can find a lot of experience in the market of people who know how to work with data and build dashboards and visual things that count things and show it very nicely visually. The skillset of building dashboards for performance reporting is different from the skillset of building a designed, curated visual, predictive contextual, actual actionable experience in Salesforce. Because what you’re doing there is you’re more an application designer. I mean, yes, you have to know data and you have to know the outcomes that you’re trying to drive.
But it’s a different approach, and so we do a lot of work with customers, educating them around design thinking in terms of, who is the persona? What are they doing in Salesforce? What are the questions that they ask and answer every day? What makes answering those questions hard? What is all the compensating behavior? What is all the friction in the process? Then all of that around an understanding of what we’re trying to get them to do. Because if we don’t know the outcome, our user experience will be cluttered. And so using the design thinking techniques that just come from design thinking or the jobs to be done techniques of designing things. We try to bring that into the foreground around data and analytics. So these user experience is quite frankly just fade into the background as part of their Salesforce journey. Does that make sense?
Craig: That’s exactly the way you want technology to be. You want it to fade into the background. You don’t want it to be in your face having to deal with it and having to interact with it. But you’re constantly interacting with, but you don’t want it to be something you have to physically manage. It’s should just be there, deriving what you need to do, as you mentioned, if it helps you, you got to reduce friction in the process.
Craig: It’s all about what their outcomes are and that the technology needs to facilitate it.
Susan: What is the behavior in the outcomes are trying to drive and just take it all out. It’s a big focus and I always like the David Foster Wallace story of the old fish in the water and the young fish in the water. I might not tell it exactly right. But the whole thing is like the old fish ask the young fish, how’s the water boys? They’re like, what water?
We want the visual and predictive insights when you’re in a Salesforce journey to be a what water story. Because we just want it to be so obvious you wake up in the morning and walk from the living room to the kitchen and starts your day. Who do I call and why, and when I get there, what do I say? Just to make it part of like that engagement model and not about AI, not about inspection, but engagement with customers.
Craig: That’s exactly right. I love that story. I use that a lot about when a particular technology is going to become mainstream. It’s when you don’t even notice it.
Susan: Exactly. Yes.
Craig: That’s the way people talk about AI is great. I can use it. I can just go to Chat GPT and do this. But you’re still going somewhere. You have to know how to prompt to use. It’s way better. It’s still going to be something that’s going to directly drive a lot of advisor productivity until it’s built in to the underlying fabric of what they’re doing, and it just appears in front of them without then having to do anything differently.
Craig: That’s when the true benefits in productivity and engagement and driving better outcomes will start to appear.
Craig: There is another area I wanted to talk about before we run out of time. Einstein is your product name for the AI parts of Salesforce. I’d like to learn more about Einstein recipes.
Susan: Oh, okay.
Craig: How can wealth matter firms leverage? For me, a recipe is, here’s a preset group of instructions you can follow to get a certain outcome without having to go and have to spend as much time putting it together yourself.
Craig: Is that what they are or are there’s something else?
Susan: Just maybe a comment on the word Einstein. Einstein at Salesforce just represents our AI brand. There are products called Einstein. There are features branded Einstein. It just means that something in our ecosystem has been AI enabled. So it’s very pervasive and its across sales service, marketing, commerce, everywhere.
Craig: We’ve noticed you’ve gone a little crazy with that, I got a little Einstein crazy at Salesforce. I think everything’s now called Einstein.
Susan: It is our AI brand. Now in the case of recipes and this is an interesting altitude question. All these other things have been about driving behavior and net new assets and focus for a wealth advisor. A recipe technically for us within the product line is a data transformation where we take data and we process it and derive new things. In financial service cloud, it might be things like taking a whole bunch of accounts and aggregating it into a household, this is the AUM of that household. It might be more aggressively complex transformations. Maybe we’re doing things like invoking one of our machine learning models and calculating sentiment along the way, or maybe we’re doing other types of math and transform. It’s just a technical data transformation where we’re working with Salesforce data or the firm’s data to derive, calculate, predict, transform, and then get it into an environment that we’re either visualizing or servicing up to the user in some way.
Craig: I’ve never heard that those two adjectives put together before. I love it.
Susan: Which two adjectives?
Craig: Aggressively complex.
Susan: Well there can be simple stuff like, take these numbers and aggregate it up because I want a total AUM for this household. There can be, like in other parts of our portfolio, we do a lot of complex things. This is outside of the domain of financial services, but that same recipe component does all the mathematics for us for rebate programs. You can imagine rebates programs can be very complex because there can be volume tiers and there can be product like this product is in, this product is out. At this tier of volume something else happens and what if there’s a return? Those can be very complex. We do similar things with loyalty point management for members who accrue points with usage of products and services also can be complex. Aggregating household accounts could be straightforward, but the math that we do can be more aggressively complex.
Craig: I’m going to steal that. When I talk to clients, and say, well, is this a big problem? Not only is it a complex problem, it’s aggressively complex.
Susan: It’s a lot of math.
Craig: It’s a lot of math.
Susan: It’s a lot of math and you want to do it nicely and visually. Yes.
Craig: I was reading about a new press release just came out about Einstein GPT, where you’re partnering with OpenAI and it’s going to be going through all of your different products. I was most interested in Einstein GPT for Slack.
Susan: Well, Einstein, as you mentioned, we launched officially Salesforce, Einstein GPT, I think three weeks ago at a Salesforce AI day in New York. Our plans for those types of experience are literally across the entire portfolio at Salesforce around sales, service, analytics, marketing, all of these areas and Slack as well. Slack is obviously a rich environment of unstructured data, and so like internally, I’ve been using capabilities with Slack that summarize the channels that I participate in and just give me the quick summaries of what’s happening. That’s a very rich environment for that. As is on the servicing side of things with Salesforce Service Cloud, just a great environment both to help operate service agents and operators with proactive prompt response. It’s much easier and quicker to respond to customers with large language models suggesting responses all with the human in the loop that an agent can then edit or just use. Then, the case summary and the action steps that are required after a case, all of that sort of productivity gain for operators. Then on more of the traditional sales side doing things like generating a customer interaction and doing it in a way that leverages what we know about a customer for the CRM, but without exposing or risking our data. If you followed us on AI day, some of the things that we talk about is the Einstein Trust gateway, where we do all sorts of things in terms of prompt engineering, grounding, masking, no prompt retention. You can leverage all this rich customer interaction that you have, but without exposing data in ways that were unintended. Sales, service, marketing, commerce, Slack, like yes, all of our product teams throughout Salesforce are looking at all their product plans in terms of how we can bring these large language models and experiences to the foreground,
Craig: At some point it’s just going to be AI, talking to AI. We’re going to be — I saw a great comic cartoon in the office and the guy was saying I’m going to take this large block of text; I don’t have time to read it. I’m going to use AI to summarize it in bullets so I can send an email. The other side personally, I got this email, these bullets, I’m just going to give it to AI to expand it so I don’t have to do that work. It’s just one AI talking to another AI being funneled through humans.
Susan: But I don’t know if you like watched Silicon Valley, but there’s that one great thing where I forget his name. They’re always arguing with each other and they created both their little AI bots and it basically melts down there. Their production servers because they — kind of silly stuff like that, but yeah.
Statistics on Salesforce’s AI Initiatives: Predictions, Patents, and Research Papers
Craig: That could happen. Before we wrap up, are there any AI related statistics or any statistics we love, that you can share about Salesforce?
Susan: I would a couple things I could share. We’ve been at AI since 2016, so it’s way predates the GPT stuff that we’re doing. We’ve baked wherever we can, machine learning both as tools people use can build their own machine learning models within Salesforce. But also we’ve got all sorts of out of the box products that just do AI. Across both of those things together, we’ve got over a trillion predictions a week that are happening across sales, service, marketing and commerce at Salesforce. We’ve got a rich team of RND that also, other facts are things like our AI patents, over 210 AI patents. Many, many AI research papers in the research community. Across usage and adoption, a trillion plus a week, and then in terms of our participation in the ecosystem with research and, and scientists also very prominent.
Craig: Susan, that was fantastic. Can you tell our listeners where they can find more information about Salesforce? I need to tell them that, but tell them.
Susan: Oh, Salesforce.com.
Craig: Salesforce.com of course we all know. Thank you so very, very much for being here. I know we had rescheduled a bunch of times, because you’re so busy, you’re in demand. Thanks again for being on the program.
Susan: My pleasure. Thanks for having me this morning.