Enjoying our podcasts? Don’t miss out on future episodes! Please hit that subscribe button on Apple, Spotify, YouTube, or your favorite podcast platform to stay updated with our latest content. Thank you for your support!
AI is making its way into many areas of banking and fintech but one area where it already has a strong foothold is detecting financial crime. And there are some companies doing new and innovative things with this AI technology here.
My next guest on the Fintech One-on-One podcast is Adam Famularo, the CEO of WorkFusion. His company has created these AI digital workers that are integrated into the workforce of a bank or fintech. These are real digital assistants that have been brought to life by WorkFusion.
In this podcast you will learn:
- What attracted Adam to the CEO role at WorkFusion.
- How he describes the company today.
- How their AI Digital Workers operate and the different job roles they have.
- An explanation of the use cases for these digital workers.
- How they have trained the digital workers.
- How banks are augmenting these digital workers with their workforce.
- What kind of feedback they have received by humanizing their AI models.
- What is involved in the onboarding process.
- Where the AI digital workers fit within their core software.
- The typical organizations using WorkFusion today.
- A couple of case studies from banks using their software.
- How much they are reliant on LLMs and Gen AI.
- How they reduce the risk of errors or hallucinations.
- What this will all look like in five years.
- How this will impact the human workers interacting with the AI workers.
Read a transcription of our conversation below.
Peter Renton 00:01
Welcome to the Fintech One-on-One podcast. This is Peter Renton, Chairman and co-founder of Fintech Nexus. I’ve been doing this show since 2013, which makes this the longest running one-on-one interview show in all of fintech. Thank you so much for joining me on this journey.
Peter Renton 00:27
Before we get started, I want to remind you that Fintech Nexus is now a digital media company. We have sold our events business and are 100% focused on being the leading digital media company for fintech. What does this mean for you? You can now engage with one of the largest fintech communities, over 200,000 people, through a variety of digital products, webinars, in-depth white papers, podcasts, email blasts, advertising, and much more. We can create a custom program designed just for you. If you want to reach a senior fintech audience, then please contact sales at fintech nexus.com today.
Peter Renton 01:04
Today on the show, I’m delighted to welcome Adam Famularo. He is the CEO of WorkFusion. Now WorkFusion is a super interesting company, they are in the financial crime space, but the way they have approached this is through the lens of AI. And they have now AI digital assistants that they bring to banks and fintechs to help them fight financial crime. And we talk about obviously, what this means, how it all works, how they created their AI digital workers. And you know, what Adam believes is going to be kind of the role of these digital workers both today, and moving forward. And the moving forward piece was really, really interesting to hear Adam’s thoughts there. Anyway, it was a fascinating discussion. Hope you enjoy the show.
Peter Renton 02:04
Welcome to the podcast, Adam.
Adam Famularo 02:06
Thanks for having me, Peter.
Peter Renton 02:07
My pleasure. So let’s kick it off by giving the listeners a little bit of background about yourself, if you could just hit on some of the high points of your career before WorkFusion.
Adam Famularo 02:19
Before WorkFusion, I founded a data governance software company named erwin. And, you know, we we got big into data as there’s this huge proliferation of data in the marketplace, from IoT devices and others, and government regulations coming down on that industry or like things like GDPR, required people to figure out how they were going to govern and manage their data. So we built a very successful company over a five year period, until we were purchased by Quest Software back in the end of 2020, literally the last day of 2020. And then before that I was an executive at Verizon for a few years. And then I spent a large part of my career at CA Technologies as a general manager in multiple business lines from cloud computing to storage to security software.
Peter Renton 03:14
Okay, so then so well, what attracted you to the CEO role at WorkFusion? Like you’re not a founder of the company, right?
Adam Famularo 03:20
No, no, I’m not a founder. I know what it’s like to found companies. And it’s not easy. And I invest in founders as well, I do technology investing on the side. But this company, when I was technically taking some time off and enjoying time with the family, I was approached by a good friend and colleague who also went through the company. And when I looked at the technology, it was amazing. And it definitely was one of those Peter your eyes wide open experiences where I was like, boy, what this technology could do to really change the industry in the marketplace is something really exciting, and something that I should take a hard look at. And the further I got into it from the investors and the board, to then peeling back the people, and the employees, and the customers, I got more and more excited to a point where I joined in September of 2021.
Peter Renton 03:38
You didn’t have a long time off, then it sounds like.
Adam Famularo 04:15
No, not too long. I had about a solid six months, I had three additional months that I used to help kind of all the integration work between our two companies.
Peter Renton 04:24
Okay, so then let’s talk about what WorkFusion does exactly. How do you describe it?
Adam Famularo 04:30
So how I describe WorkFusion is we are at our core an AI company that uses IDP technology. And we’re the leader in something called IDP, which stands for intelligent document processing. So we can read and decipher any kind of data input. And then we have our AI engine that allows us to put a workflow on top of it that can then do work. That’s the cornerstone for who we are as a technology. What we’ve done as a company is we’ve focused in on a specific use case, which is financial crimes. So our goal is to stop the bad guys. And that’s what we’ve been doing. So when I joined in September of ’21, I went out there on a roadshow, I met with about 60 customers, I found the best use cases of our technology. And they were just in this fit, which was around financial crimes. And the more I peeled it back, the most successful customers that we had, were the ones that took the time to take our software and treat it as part of their company. So you’ll see that they’ve given them kind of names and roles and responsibilities. And they’ve really treated our software as kind of part of a team. And that’s what kind of gave me the idea of going the next level with our technology. And what we did was we actually took our technology, we gave them all specific job roles and job responsibilities in financial crimes, we gave them names and faces, as you see on our website, we hired actors and actresses to portray them, so that it would make it easier for our customers to onboard our technology and treat them as part of their financial crimes team.
Peter Renton 06:15
Right.
Adam Famularo 06:16
That’s what we did. We launched it back in February 2, 2022. And that was kind of the birth of our digital workers strategy.
Peter Renton 06:24
Right, right. So I want to dig into this, because I gotta tell you when, I get a lot of pitches for people being on my podcast, and most of them don’t make the cut. But when your PR people pitched me on your company and I went to your website, and I looked at these AI digital workers, and I’m like, thinking are these people real? I mean, I couldn’t, I could not figure it out, because it looks real. But then it’s sort of like an AI digital assistant. And they look like humans. So like you said, you hired actors. So maybe these are all based on real people. But they are when you see these AI digital workers talking, who are they?
Adam Famularo 07:05
So great question, Peter. And it does get very interesting for our customers. Because what happens is, our viewpoint is you hire our digital workers to come to work for you to do a job role. So for example, you would hire Tara. Tara does transaction screening. So she looks at all the transactions that come in the bank, and she will let you know which ones are false positives and which ones should be allowed through. That’s her job role. She will work with a team of humans. And when she gets stuck, she’ll be able to reach out to the humans and say, Listen, John, or Jane, I’m stuck with this problem. I believe the answer is this, this, or this? John or Jane will log in, they’ll say Tara, the answer is this and then Tara goes back to doing her job. That construct in itself, Peter, is the reason why we gave them names, and job roles, and specific job responsibilities, because it really is. You’re hiring them to come to work for the bank and play a specific job role. And when you peel back that job role, these are job roles that most humans don’t want to do. It’s very tedious work. It’s very manual, it is very labor intensive. And you have the regulators on you. So very error prone work. So that’s why we kind of honed in on these specific use cases. And we really gave them a persona so that they are viewed as part of the company.
Adam Famularo 07:06
Right. Were they based on someone like an actor that came in and did this, or are they purely AI generated?
Adam Famularo 08:38
No, they are actors and actresses that portrayed those roles.
Peter Renton 08:44
Okay, so they are real people. But when you interact with them, you’re not interacting with obviously, the actor, you’re interacting with the AI.
Adam Famularo 08:52
And we’re getting closer and closer to you interacting them like that, like we’re working on things that will you know, Tara will be able to call you on Teams and you’ll actually see her face and she’ll talk to you about the issue that she’s having. So we are taking this as far as we possibly can to make sure that the teams feel like Tara truly is a part of their team.
Peter Renton 09:15
Okay, so then, obviously, they’re digital, they can, they work 24/7, right, they don’t have to break, how are people using them? Are they using them in that way where there’s a million documents that need to be analyzed, go off and do that. I mean, how, maybe give us some examples of how they’re using different types of digital works, maybe describe the different types that you have and how they’re being used.
Adam Famularo 09:38
And that’s a really important part of giving each one a specific job role and job responsibility, because where you’re going, Peter is that you could literally do anything with these digital workers, right? You just say, Okay, go focus on now reading and deciphering this data, and then you can create a learning algorithm behind it. What we’ve done is with intent, so with intent we’ve said okay, Evelyn, you’re going to focus on adverse media. So you’re gonna look for all the current bad people and new bad people and make sure that we stopped doing business with them. And that’s your one specific job, and Tara you’re doing transaction screening. Kendrick, you’re gonna look at all the customer IDs and ensure that it’s a valid ID. Darryl, you’re going to read the documents associated with the KYC process. So each one has a specific job role and responsibility. And they will do just that job role. So what we find is that other companies that tried to do something like this, they go the more kind of broad approach where we could do anything for anybody. But then it takes a lot of time, money and effort in doing all the training. Our biggest differentiation is our models come pre-trained. So this is a, you know, one, two, sometimes three years of training that’s gone into these models. So our viewpoint is when you’re hiring a Tara or an Evelyn to come to work for you, you’re hiring somebody as if they’ve been doing the job for three years. This isn’t somebody that is, you know, out of college, and you have to train them from scratch. That’s why we’ve given them kind of the names, the faces, and specific job roles and responsibilities that they’re tasked to do.
Peter Renton 11:18
How have you trained these assistants then?
Adam Famularo 11:21
So we’re very thankful that our existing customer base has allowed us to take the training models that we’ve done with them, and bring it up and productize. So from there, so that was the initial onset. From there, we actually have all of the additional Taras and Evelyns that are working out there, their feedback is going into the models, the future models, we’re taking all of the model improvements, putting them into new releases, and then bringing them out to the marketplace. So that’s what we do. Now we do have a way that you can opt-in as well, this is up to each bank, where they can actually get real time updates with Tara and Evelyn as we make the model changes. But what we find is most banks want to be a little bit more structured behind the changes that they bring into their business. So from that standpoint, you know, we can usually do upgrades every six months to 12 months.
Peter Renton 12:16
Right. Okay. So are the banks that you’re working with, do they have, I mean, obviously, all the things you’ve mentioned, are things that banks have been doing for decades. When you bring in these digital workers, right now, are they just augmenting an existing team? How are the banks finding it?
Adam Famularo 12:34
That’s the word. So augmenting the team is really the right word to be used. I was just with the CEO and his leadership team from a bank last week. And he used that same word as well, which was, look, these are augmenting our team members. We have so many other higher end jobs that we can promote, train and develop people to take, that this is augmenting the workforce that we have today. And we see that for the most part is that most of these banks are having a hard time expanding their abilities to handle the pure number of sanctions that are coming in. If you look at from the Russia, Ukraine war, sanctions are up over 100% since you know, the war broke out in February of 2022. And they’re planning for more sanctions coming from other countries. So it’s like, how do we handle all this growing demand with a current labor force that barely wants to even do this job? So that’s really where we come in, and we’re augmenting the jobs that are there today, and allowing others to do more complex higher end work.
Peter Renton 13:39
Right, right. And then did your clients embrace this kind of, I don’t know what you’d call it, humanizing way of kind of implementing your models, or what was the feedback like there?
Adam Famularo 13:51
They love it. So we’ve had multiple customers that have been doing kind of their own variants of it, because what they portray to me, using our software is culture changing. So when something is culture changing, you need to make it easier for people to adapt to it. They themselves were, before we took this tact, you know, I had one company was creating little robot figures and naming them. I had another company has literally a cardboard cutout, named her Sarah, and sat her at a desk. And that was the person that was doing transaction screening. So I had all kinds of different variances. So they actually loved the fact that I made it really simple to understand what the job role each one was going to do, that I gave them a specific name and a face that actually made it easier for them to plug them into their teams. And they were able to address them as though we’re now hiring Tara to take on this specific work. And oh, John, you’re gonna have to work with Tara to help her when she gets stuck. That made everything very easy for them to be able to digest and use, and then the other big thing that we did outside of just giving them names and faces, is that we built all these partnerships. So we built in all the integrations. So it made it really easy for you to onboard Tara or Evelyn, where back in the day, it might have taken you six months to deploy it. Now we can go live in weeks, three weeks, four weeks, and you can be up and running and start using Tara in production.
Peter Renton 15:19
Okay, let’s talk about the onboarding process there, what’s involved like, particularly, you get a new client, right, that you’ve never worked with before, they want to bring your software in house. What’s involved in the onboarding process?
Adam Famularo 15:31
So now, as I mentioned, the integration part is number one, right? So we look at what financial systems are they using around sanction screenings, like things like Firco, or Clear, different types of companies out there that we need to integrate with. So think of the tools that the humans were using, our AI software needs to interoperate with. So we have built all the partnerships with all those tools, and we have out of the box integrations built. If there’s any others that we need to kind of integrate into, we can do that. That’s part of the kind of onboarding process. And outside of that, our machine model is ready to go. And usually, you know, from going from just doing proof of concept we’ll usually be about 50/55% automation rates. And then when we plug them into the organization, so now we assign a person of the team to work with Tara or Evelyn. And now that person will log into the system once a day, twice a day and make any updates or, you know, address any of the issues where Tara or Evelyn got stuck. And then from there, it will continue to improve and improve and improve until it hits 60%, then 70%, then 80% automation, and Tara and Evelyn is just doing more and more of the work.
Peter Renton 16:45
You know, I’m curious about how they actually do the work. Are they, I mean your software must go out to like sanctions lists, lists are being updated all the time, there’s more bad guys that get found out. I mean, are they just going out and monitoring different databases in real time? How are they doing this work?
Adam Famularo 17:06
Yeah, that’s one of the jobs like Evelyn that’s doing adverse media. She’s tapping into all of those different databases, both external and internal, using websites, and understanding kind of how that is changing. And then as it changes, then she’ll roll that back through to her human team to then implement any of the changes within the organization from what she’s discovering.
Peter Renton 17:29
So then, is this just part of what WorkFusion provides now? Or is this really what we’ve been talking about for the last 15 minutes, is this really the product now?
Adam Famularo 17:38
This is the core function. So if I just categorize it, we have what we call our Work.ai platform. Our platform is the industry leading IDP technology with our AI engine, right. That’s the core construct of who we are as a company. Then on top of that, we built out our instantiation, which is these eight digital workers that just do financial crimes. That’s it. And then what we’re doing with partnerships, like we inked a partnership with a company called Epic, we inked another partnership with a company called Emphasis. But we’re enabling those companies to build and develop their own digital workers for different industries. Like we’re never by ourselves going to go into the legal industry. Epic is one of the leaders in legal software and services. So we’re going to empower them to take our software, build digital workers for legal, and bring those to market, and then just give us a small royalty.
Peter Renton 18:36
Interesting, interesting. So I mean, you’re focused on financial services, though, right?
Adam Famularo 18:41
Yeah.
Peter Renton 18:41
As far as the core of what you’ve been doing, who is the target market, exactly? Do you need to be a certain size before you can bring WorkFusion in? Who are you working with?
Adam Famularo 18:53
Yeah, that’s a good question. So we do cut across from fintechs to regional banks to large banks. We do cover banks, you know, all throughout North America, from US to Canada, as well as through Europe. Those are kind of our primary markets. The who we sell to, what we sell primarily to your chief risk officers, or chief compliance officers. So the people that are responsible for looking out for and protecting the companies against the bad guys, those are, you know, our core users and use cases that we work with.
Peter Renton 19:28
Could you provide maybe a case study on one or two of the banks that you’ve worked with that are engaging with these AI digital workers?
Adam Famularo 19:40
Yeah, so outside of this, Jess Casssdy, our head of PR, can actually provide you with case studies that we can put along with the interview.
Peter Renton 19:49
Okay.
Adam Famularo 19:50
We also have stuff on our website that we can use. I always have to be a little bit careful because a lot of the banks don’t want their names being used and I don’t like to step on their toes, but we could definitely provide you with multiple…
Peter Renton 20:03
That’d be great. If you could provide that when we publish the podcast, I’ll link to it in the show notes.
Adam Famularo 20:07
Perfect.
Peter Renton 20:08
So then, let’s talk generative AI for a minute, because it’s been a hot topic now for over a year with Open AI and kind of ChatGPT journey that that’s on, I’m curious about how much of what you’ve done is taking some of these elements of the generative AI technology, and large language models. I mean, how much of what your you have produced today is reliant on some of this new technology?
Adam Famularo 20:37
Plenty, actually. So we are gone pretty deep with LLMs. And we started very early on both with ChatGPT as well as with Google, with Bard, and now the new new product off of Bard. We believe that the LLMs are this great kind of horizontal play on understanding data and information. But what they don’t have is that deep vertical presence, and that’s where we come in. So we have, the technology I was alluding to before, we call it human in the loop, where we plug in humans into our technology to help enhance our models. Well, we have the same thing with LLMs, where we plug LLMs into our models to help enhance our models and do work that is better served by them, and then feed it into ours. And at the end of the day, we’ve been able to see automation rate improvements from I mean, shoot, we took Tara within one company from 80% automation to 92% automation by integrating LLMs in. So there’s a nice fit in there, where we use LLMs along with humans in the loop to help create a better worker that we deliver into the marketplace.
Peter Renton 21:54
Okay, and you’re in an area that is where you really can’t make mistakes, right? You’ve got a, this is serious business that you’re in, you’re not talking about, you know, movies or TV or anything like that. You’re talking about finance and highly, highly regulated companies. And with you bringing some of that in like the large language models, how are you reducing the risk of hallucinations or errors of any kind?
Adam Famularo 22:21
Yeah, well, there’s a couple of good points in there, Peter. The first point is LLMs are not welcome at every bank, right. And that is duly noted by each bank. And the ones that are jumping in are jumping in with with bated breath, right. So they’re putting a box around it, they’re watching and monitoring and making sure that they’re not harming themselves. From our standpoint, we view it in a very similar way, which is we have core underpinning information and data that we’re taking out of the LLMs, we’re documenting any decisions that we use the LLMs to make. We are not a black box, we’re very much a white box, and to your point, regulators need to come in and like what they see with us. And they do, right. Today regulators come in and they see that our product’s installed and being used, and they can easily click on and be able to see all the decisions that we made and why we made them. That’s a very, very important piece to us is that we’re very much a white box. It’s open, right? It’s an open ecosystem that people can understand what decision we made, why we made it, and can prove that this decision was the right decision at that point in time. So from our standpoint, we enable our banks to be a little bit better served and better protected, because our software is riding on top of the LLMs, not just using the LLMs to make decisions, but using their data. And then we’re the ones that are ultimately making the decisions at the end of the day.
Peter Renton 23:49
So if there is a hallucination from the LLM, you’re catching it before it really gets gets out there into the wild.
Adam Famularo 23:57
Correct.
Peter Renton 23:58
Okay. Okay. So the last part of our interview here, I want to talk about the future because I think you’re in an interesting position, sitting where you are with these, you’ve really gone all in, shall we say, on having kind of AI digital workers be a part of the future of work, shall we say. So my first question here, then is when you look at the state of where we are today, I mean, I imagine we’ll look back in 10 years and think this was 1.0. And we were pretty rudimentary, right?
Adam Famularo 24:31
Rudimentary. Very much so.
Peter Renton 24:33
Well, where are we going? How’s this gonna look in just in your niche of like financial crime and all the things we talked about? What’s it going to look inside a bank in maybe, even in five years?
Adam Famularo 24:45
It’s going to be so much more advanced than it is now. I started giving you a sneak peek of it earlier in the conversation. But like right now the interaction is, you know, within a portal that you log into, and you see all these data feeds, and then you make, you help enhance the models based upon those data feeds. You know, we’re looking at taking, you know, the faces from the actors and actresses, and enable them to be able to converse with you, just like I’m conversing with you, Peter right now. You’re, gonna be able to converse with Tara, just like this. We’re gonna be able to have a conversation, Tara’s going to be able to learn from you in this conversation mode, and be able to go back to work. That isn’t too far out either. That’s like 12 months, 24 months, you’re going to be able to converse with AI as if you’re conversing with a colleague. And and that’s going to kind of change everything, because the closer that the AI technology will come to feeling and acting like a human, the more work that’s going to be prone to go over to the AI side, and make it that much easier and faster for others to learn from them as well.
Peter Renton 25:54
What about like, are you concerned about like deep fakes and different ways that the talk about the future, and it’s already now you can have, you can basically create a video of anyone saying anything.
Adam Famularo 26:04
Correct.
Peter Renton 26:05
Today that technology exists. So how is that sort of, how are you thinking about that going forward?
Adam Famularo 26:11
We’re actually looking at using some of that technology so that we can actually portray our AI as humans, right. So that’s actually how we’re working on doing that. And making it so that it’s more conversational AI, that’s kind of the next, the next world in this is getting taken to a conversational point, right? You know, I’m not as worried about it, I kind of feel like governments and governments around the world are looking very hard on this, I think that there will be things in there to make sure that we know that it is an AI versus it is a human, I think that’s what needs to be kind of part of the dialogue. But I do feel like there will be the right rules to and protections in place to help stop the impersonation. The impersonation part is really where it gets dangerous, if it’s impersonating, you know, a president, and it’s saying something, and then people are acting upon it. But it wasn’t the president that said it, right. That’s where it gets really dangerous. And I’m hopeful that the right rules and regulations will be put in place to help us, you know, stop that from getting out of control.
Peter Renton 27:21
So then what about when, as banks get larger, and your technology gets better? You did say earlier that this is really augmenting workers today. Five to 10 years time, if you want to, if you want to really ramp up your Financial Crimes Unit in your bank, are you going to be hiring people? Or are you going to be hiring, you know, AI, digital workers from WorkFusion?
Adam Famularo 27:44
Yeah, you know what, Peter, that’s the best question of the day, because this is going to be in every industry. So at the end of the day, every industry is going to change, where people are going to wind up doing more higher end work, you’re still going to need people in the process, some way, shape, or form, but they’re not doing the base level work that they’re doing today. So a lot of those like L1 remedial work are going to disappear, that will absolutely happen. But the higher end work will keep on just getting higher. And people will always need to converse with AI in some way, shape, or form, that will always be there. But I do feel like it’s just like when, you know, automation came to the manufacturing industry, right? So people move from doing you know, one job, which, you know, the automated robots were better off doing to other work where, you know, humans were better doing, you’re gonna see that kind of ebb and flow within enterprises over the next decade.
Peter Renton 28:42
Well, it’s really going to be interesting to see how this all shakes out. And you’re in a very interesting space. And I think you know, what’s going to happen over the next three to five years, I think, like, I sometimes think that we can’t even imagine some of the things, and some of the ways it’s going to change, like we couldn’t imagine an iPhone in the 90s. It’s just those sorts of technologies, just were science fiction, but anyway, Adam, really appreciate you coming on the show today. Thank you so much.
Adam Famularo 29:11
Peter, thanks for having me. It was a great conversation. And I look forward to meeting up again with you one day soon.
Peter Renton 29:18
Well, I hope you enjoyed the show. Thank you so much for listening. Please go ahead and give the show a review on the podcast platform of your choice and go tell your friends and colleagues about it. Anyway, on that note, I will sign off. I very much appreciate you listening, and I’ll catch you next time. Bye.
