"Don't be afraid of AI, and disruptive technologies in general. They are here to stay, they're not going anywhere."
In this episode we are joined by Michael Berns from PwC. Michael is not only an expert in AI & FinTech, but also appears as a keynote speaker at international conferences and guest lecturer at several business schools. In this episode, Michael shares with us three exciting use cases on the different application areas of AI, where you subsequently might ask yourself: Is AI making the world a better place?
If you are also searching for the answer to that question and want to learn more about how AI is being used in the context of financial crime prevention, how it significantly speeds up the process of cancer detection, or even helps to protect people from human trafficking and exploitation, then stay tuned.
Below you will find the full transcript for the episode.
How to make the world a better place with AI
Chris: Welcome back to the Innovation Rockstar interviews. My name is Chris Mühlroth, and today I am very happy to welcome Michael Berns from PwC. Michael is Director AI and FinTech leader at PwC. And there, he is helping both PwC and its clients to benefit from global AI insights and expertise. Also, he is leading the development of AI-enabled products and solutions. And of course, as a well-known expert in his field, he also acts as a keynote speaker at international conferences as well as a guest lecturer at a number of business schools including London Business School with a focus on AI and FinTech. That's quite a lot, Michael thank you so much for joining us today. It is a pleasure to have you here.
Michael: Thank you for the honor to be speaking with you, Chris.
Chris: Let's start right away, I would say let's start with you. Can you tell us a few words about yourself and your career so far?
Michael: Sure. I had the pleasure of working globally out of London for some 16 years. This also included being part of financial technology, FinTech, in the early days, where I was working with some of the innovation companies in that space from 2005. And were some short stints in investment banking with Morgan Stanley. And later I made my move towards artificial intelligence.
Chris: That's interesting. Now, how did you actually get in touch with artificial intelligence, what was kind of the first touchpoint?
Michael: It was actually through my executive MBA at London Business School. Where I started mentoring from the early days like 2011, to see whether I could pass on some of the lectures or the insights on running companies to startups in the London ecosystem. And luckily, for me, London was seen as an entry point to EMIA for Silicon Valley companies. And it was sometime around 2012, where I came across my first AI startup, and I was mentoring them in terms of go-to-market, kind of product insights, working with financial services, and that was a two-way mentoring relationship where I passed on some of my business skills to them, and I learned a lot about artificial intelligence in turn. And that inspired me and changed my role, my future outlook, a lot of my life in the process.
Chris: Now, addressing the topic of artificial intelligence. I guess, one of the most discussed questions still today is about the fundamentals of AI. So, now that I have you here as an expert, what would you say, where does real artificial intelligence actually start?
Michael: Of course, we always have the conundrum of defining intelligence. What does it mean? Is it emotional intelligence? Is it social intelligence? And so on. So, I think the problem with defining artificial intelligence is based on the fluffy term intelligence itself. But for me, AI does include machine learning with all the different flavors from supervised learning, unsupervised reinforcement and so on. But I often also have to explain to different prospects that even though we are not aiming for to replace a human being or to have something human-like, we can still use AI or machine learning to generate insights that they otherwise wouldn't have. So, if I can build a solution that is 80% accurate, that can still fundamentally change the way that they do business today. So, what I don't include typically in my definitions are things like experts or rule-based systems, or robotic process automation, but that's really just my personal take on it.
“We always have the conundrum of defining intelligence, so I often also have to explain to different prospects that we are not aiming for to replace a human being or to have something human-like, we can still use AI or machine learning to generate insights that they otherwise wouldn't have.”
Chris: Now, in the beginning of this episode, I had introduced you by saying that you help PwC and its clients to benefit from global AI insights and expertise. Tell me a bit more about that, how is artificial intelligence actually embedded in your activities at PwC?
Michael: Sure. So, my team helps the organization as a whole to become more digital to fully leverage innovative technologies, AI being one of them. This is to empower employees to help create resilience to handle some of the routine tasks, to make life and work more interesting. I also believe that as a company, we need to fully understand and get involved ourselves and make more digital before we can consult clients. So, kind of really to put our money where our mouth is, to really do these things and then go out to the market.
Chris: Right, and from a helicopter view, like from the very top level, what are some of the hot topics, especially in AI or connect with AI that you are currently dealing with?
Michael: So I've been in this field now, kind of as a mentor since 2012 and then professionally from 2013. I think we're now seeing the second wave where it's not just about having an AI system or implementing AI. It's more about how exactly it's being implemented. So, the new topics now for a year and a half, two years, on the international ground, in the US is ‘responsible AI’, the ethics that are related to that. And when I look at my own kind of publications and debates over the last year, even doing COVID, it was interesting that there's a lot more interest for AI for ESG. So really, looking at the environment, looking at the data, what can AI contribute to have a higher awareness in terms of the products we buy the companies we deal with, the reporting and everything around that. So in short, it moved on from a buzzword to a much more granular and sophisticated discussion around how it should be applied or used, it's more in terms of we have our own view from an ethical perspective.
AI in the financial sector - RegTech & SupTech
Chris: Okay, that's great. Now, let's dive into some of these hot topics and also a few other exciting application examples you brought us for this episode. Let's start maybe with the financial sector in general, where you have been working for many years. I did some research prior to our episode and I did find an interesting quote from you. You said that "AI is going to be a key competitive factor for financial institutions in the future. But it also offers other application areas far beyond process automation.” What would you say, how widespread is the use of artificial intelligence in the financial sector today? Would you maybe argue that AI is already a part of the standard toolbox?
"AI is going to be a key competitive factor for financial institutions in the future. But it also offers other application areas far beyond process automation.”
Michael: Sure, I would. I think the quote that you took is from the AI in financial services study that we published in May last year. And that was a reflection on my discussions with about 150 experts in the DACH regions of Germany, Switzerland, and Austria, to see where they are at in terms of AI. And while everybody seems to be interested in the topic, when you talk to the executives and the expert, It was very clear that only a very small part is actually ready for fully doing that. So, for the majority of them, they would say interested but not ready. And when you look, let's say to the US or Asia where I also had some years of working there, and the situation is entirely different it becomes more in terms of what are the new use cases, of course, we are doing this JP Morgan and other large institutions spending billions of dollars every year to build that out. And that is something that we're not seeing yet so much in Germany and the DACH region. So the spendings on AI are very different depending on where you look around the world.
“While everybody seems to be interested in the topic of AI, it becomes very clear that only a very small part is actually ready for fully doing that.”
Chris: That's true. And I guess especially for your GSA, Germany, Switzerland, Austria and related countries, maybe entire Europe, actually spendings are considerably lower than in China or US. So that's totally true. And when people talk about AI and FinTech, I oftentimes hear about a few other boards like a RegTech which means regulatory technology, and also SupTech which means applying artificial intelligence for purposes of supervision. Now, can you explain to us in one or two sentences what exactly this is all about?
Michael: We heard about FinTech earlier. So, you could say both are in a way sub-areas of FinTech. When you go back to my background going to my time at Moody's, you would say that, yes, we worked in regulatory technology, because we were building systems that generated insights that could be used for regulatory reporting, such as Basel II, Solvency II at the time and so on. On the other hand, once you've built solutions for banks and other financial institutions to help them with their regulatory reporting and other compliance areas, you can also see that the same solution could be applied to the supervision to the regulatory authorities to kind of look at the market holistically to gain insights from unstructured data and really see where the market is going in terms of trends.
Chris: Now, let's make an example, here, how is, let's say we talk about financial crime prevention, for example, how would one apply artificial intelligence to help actually preventing financial crime, do you have an example of this?
Michael: Sure. In my last role, we worked with about five of the top 10 banks in the world, and it was all around preventing financial crime. We worked with them closely around regulatory fines that were 200 million to a billion dollars in the areas of ethics collusion, LIBOR or front running and also anti-money laundering. So, given the amount of fines that were issued from maybe 2013 onwards in these areas, the banks had tremendous pressure to look outside for solutions and to use innovative technology. What we built for them and that was with American Banks were solutions that could help them to see the communication that was happening within the institution and then look for signs of breaches for those collusions, front running, and so on. And it was very often that even during PoC, where you just build a test system that the institutions found things that they had overlooked years ago and immediately wanted to investigate. It's kind of using AI to make sense from vast amounts of data. The communication within the firm across all channels and then deriving insights that help you find the needle in the haystack. And to protect the firm from another billion-dollar fine or any kind of legal prosecution.
“It's kind of using AI to make sense from vast amounts of data and then deriving insights that help you find the needle in the haystack.”
Chris: I just was about to say or to ask the exact same thing. So it sounds like you have been applying, or you're still applying artificial intelligence for preventing financial crime by looking for a needle in a haystack. So, collecting large amounts of data, processing large amounts of data, of course, that's a given. But in that case, you try to really spot or help make the AI spot certain things that look unusual, that seem to be outliers that may talk about certain topics or textual information that may be a problem and can give you a hint for "Hey, there's something wrong, you need to look at this." So that's what I understood.
Michael: The fact that let's say if the banks were sampling 1% of the communication before, and they were having people listening into that by the time the day is over, and that person has listened to eight hours of conversations, it's very difficult to make connections for them or see the patterns. The AI solution can really compare a discussion on one day to the one next week when the traders are using some code words, or they use a takeaway menu, just to avoid detections. We've seen all of this in the data. So they do everything that they can to avoid detection. And without a solution that can kind of predict with a certain accuracy, what it might mean that they're talking about, it's very difficult to detect this and see the patterns in the data also in terms of the time of day and kind of getting involved in a chat room to collude on certain aspects.
AI driven health care automation
Chris: Right. Okay. Now, that's one application case. Thanks much for bringing this up. This is really powerful, I guess. Right. And I guess another area where you have also been successfully implementing AI is in the healthcare sector. And then one specific example that we touched on in our prior conversation, is the development of a model that speeds up cancer detection or the process of cancer detection. Can you also tell me more about this, you know, what was the driving force for starting the development of such a model?
“Usually, the data procuration is what takes a lot of time in these cases.”
Michael: Sure, I mean, when you compare the healthcare system in spaces like the US where about 18% of the GDP is spent on health care, that's a massive amount. And it's been ever-increasing. And they're always under struggle to really deliver on this. And then, at the same time, in the UK, where it's constantly overfunded. So those two places were the basis for some previous work in this area, where just to try to alleviate the pressures on the system in a way you could compare it to these massive regulatory fines and financial services, but different pressures. And you build a case where you try to speed up patient care. So helping the individual patient, you're trying to kind of make it more efficient and faster, by also creating something that generates value for the insurer, by having more accurate prediction in terms of what experts somebody should be referred to. And of course, it also works for the hospital in terms of business, and the way that they can have more efficient use of their staff and their experts. So it's one of those rare scenarios where you create a setup that works for all three parties, you could say, Oh, it's a cliche, win, win-win. But it took a lot of work to get there over several years to kind of really talk to them and get permission, given the circumstances to use the data for that. So usually, the data procuration is what takes a lot of time in these cases.
Chris: Right, and let's dive into that. So how does this model actually work in practice, and maybe also combined with a question: At what point in the overall disease diagnosis process does AI actually help? Right? So where is maybe detection accelerated? What's the role of artificial intelligence in that process?
Michael: It's in a way similar to financial services where you try to pick the needle in the haystack, or where you have a lot of unstructured data, and you try to derive insights from that. So what really happens is the patient goes to the hospital for a scan, so an X-ray, and in the patient's file you have a lot of doctor's notes, from previous visits, from SGPT, and so on. So having all this background information is certainly helpful. But as a human, you will probably struggle to keep that all in your brain to make sense for now, the patient going for the X-ray. The X-ray might show some things that we're not looking for that we're not even being aware that they might be there, but based on the doctor's note from previous conversations, based on all the context that is being analyzed by the machine the solution then tells the specialist, we didn't really look for this particular thing here on this x-ray but could it be that there's another thing present here, and we should really refer the patient to another specialist while he is at the hospital. So from a patient's side, they get a text or an SMS, to say, "Unfortunately, we found some other things in your X-ray scan, would you want to see a specialist today?" So much faster than going back home to wait three months for another appointment. And also in terms of proactive care and proactive prevention, a much better and safer route from that perspective. So it's very much an integrated solution that saves double-digit millions a year.
Chris: So, it's not only for good, it's also having a massive economic impact, obviously. But now, especially in the health care and health sector there are long processes for safety tests for approvals. How can I imagine such a process when you are about to introduce that new technology or artificial intelligence for cancer detection? How long does it take to actually get the approval to apply this new technique, given all the data and also the privacy protection considerations around the data and the processing of the data?
Michael: Now that I'm back in Germany for about two and a half years, yes, I would imagine it would take a lot longer here than in the US to procure that data. Even though we built this case before the arrival of GDPR, I believe if there is enough pressure on the system that people would be willing to share their data more freely. I think what you see in Germany right now, for example, with Corona is that the adoption rate of the app in terms of sharing your location, sharing some bits of personal information, there's been a good uptake on this, I think that as long as people see the benefit, they will eventually change their mind and come around. And I think we had a luxury situation here, where we haven't had the same pressures on the health care system as we had in the UK where I previously worked by underfunding or the increasing costs like the US, where it's all about the legal side where people can be sued and so on. Where there are a lot of side costs involved that are not even healthcare itself.
Chris: Okay. That makes total sense, especially in GSA - Germany, Switzerland, Austria - the end of the European Union as a whole, whether it's good or bad. But, we have a very special take on how to work with personal data or personally identifiable data. But I think as soon as it is applied for good purposes in the health care automation or healthcare sector which definitely is a good thing there are fewer hurdles to overcome. But on the flip side I'm pretty sure the discussions will be around for some time. It's a good application case and maybe, we need to think over our own policies of working with the data. I totally agree with what you just said before.
Michael: I mean, we see the challenges on the healthcare system here, when it comes to the COVID vaccinations and other areas where if we could automate or streamline or make this more efficient by having more data insights, that would certainly be helpful and society as a whole would benefit. It's about getting critical mass in terms of opt-in and really showing the benefits clearly. And I feel that the people that went through the cancer treatment and the automation around this would be strong advocates for more efficient healthcare and referrals.
“If we could automate or streamline the COVID vaccinations or even make this more efficient by having more data insights, that would certainly be helpful and society as a whole would benefit. So it's about getting critical mass in terms of opt-in and really showing the benefits clearly.”
AI in the context of human trafficking intervention
Chris: I mean, that's also what we see with innovation happening in general. As soon as there is enough pressure for a specific topic, things get moving pretty fast. But before that actual tipping point is reached, there are tons of discussions, but after you have hit that tipping point oftentimes things move pretty fast as mentioned before, and you see change happening, sometimes from the outside, not in the European Union obviously, but I know that both, the European Union itself and also the countries inside are working hard on finding a good working model for dealing with data. Let's see what comes out of it. When addressing data and sensible data let's talk about another application case that we have briefly discussed before. It is about the intervention of human trafficking using AI. That's a very sensitive topic, I'm being curious now. Tell me, at what point in your professional career did you actually get into contact with this sensitive topic?
Michael: That was also in my last role and came up during the interview process, actually. When I think back to that it's like I was surprised that you could use AI for that use case. But I was thrilled when I heard that the company was doing this on a pro bono basis that they were working with law enforcement but to provide their systems for free and create something that can help in that area also. What happened there was that the famous actor Ashton Kutcher started a charity called Thorn. And that was intending to protect children online and then also protect from human trafficking and exploitation and all this. What they didn't have initially was a system that can help them to scale, to empower the police and other law enforcement agencies to track down on this. That was some very interesting discussion as part of my interviews. And then once I joined the firm that was something that some colleagues and I got a lot of exposure, and we admired as it was being presented in the Senate and got more funding as we went along.
Chris: Okay. So again, how is AI involved from that context, how can it be used to prevent human trafficking? Is it again about finding the needle in the haystack pinpointing to certain outliers or clusters of data that can be found?
Michael: Yes. It's like we heard from financial services and healthcare. It's dealing with the unstructured data, again, we're talking about communication for human trafficking. A lot of this is happening in some chat forums on online forums and adverts. And the solution kind of tried to use some of the IP that was initially developed more than 10 years before to identify terrorism on public forums and chat groups. Using different intents and trying to detect the age of people posting on those forums. Whether they try to present themselves as somebody that they are not, and then whether other people were offered in this process. And it's been very effective for the police force, they also have a struggle for resources and time allowing them to pinpoint cases to investigate where people were offered. And yeah, over the last five years or so more than 7000 people were freed from human traffickers by using this solution or this solution being used by FBI and police agencies around the world.
Chris: Oh! Wow, okay. That's clear evidence that the use of AI is significantly improving the detection rate of that. But I'm pretty sure this is not only using public information data. There needs to be some good and best practice data sources that can actually be attached to those systems, but that's kind of a confidential topic to touch on?
Michael: I think what I can say is that when the company was founded, they were founded past 9/11 to detect intent and communication. They have a lot of the very early days patterns and natural language processing, deriving insights from language. And then it was a case of training those models to detect different intense of collusion, boasting, harassment within financial services, but then also detect online in certain forums, different intents when it comes to human trafficking. Always take in a lot of data, train a model and do something like profiling. When I think back about some of my work, it was in some ways a bit like a secret agent, where your profile, how somebody would behave in a certain way, how they communicate on different channels, what terminology, what kind of expressions they use. How frequent would they communicate, who would they want to connect to. And some of this communication is in the public space and others is within the individual institutions, such as the banks. And as far as let's say the US data privacy is concerned at the time, all the communication within the bank belongs to the bank. So if the employee is using bank systems or bank mobiles whatever, this is part of the employment contract that the bank can protect themselves and look into this communication. When it's online, of course, in a public forum, it's a different thing. And yeah, healthcare, a lot more protection around private data.
“When I think back about some of my work, it was in some ways a bit like a secret agent, where your profile, how somebody would behave in a certain way, how they communicate on different channels, what terminology, what kind of expressions they use.”
Chris: Right, a lot of collaboration at first certainly are required to, get this all together. That is interesting. Now, when you look at all these application areas that we have had, in the financial sector, in healthcare, in preventing human trafficking. Would you argue that artificial intelligence is indeed making the world a better place?
Michael: Yes. I strongly believe it does but as with all new technology arrivals, it's currently not really fairly distributed around the globe. It's a case of some countries are currently behind and probably need to catch up. Otherwise, there's a risk of others leveraging AI more. I compared to the rival of the internet as such, is a neutral tool, but it could be used for good and bad, and the same goes with AI. It depends on how we use it and what users will make out of it. I'm on the side to believe that if I probably create as many jobs as will be migrated or destroyed, it's just creating different roles and replacing all those, that's for sure. And if you get involved early as a country or company, then you will probably have a chance to be part of the disruption to create new roles rather than just being disrupted.
“If you get involved early in the topic of AI as a country or company, then you will probably have a chance to be part of the disruption to create new roles rather than just being disrupted.”
The future of AI
Chris: That's true. And in your personal opinion, what do you think, where will we be with AI, maybe in 10 years or so from now?
Michael: I would hope that it will become a commodity, where it's underlying a lot of other things, working in the background, that a lot more people have access to it. And then also, I would really hope that Kai-Fu Lee the author of “AI Superpowers”, that his quote becomes true to say that, "the global GDP impact will be somewhere around $15.7 trillion by 2030." That's a tremendous amount. And yeah, he's quoting the PwC study on this. I can't tell you what the exact amount this will be. But at the moment, unless the setup changes, a lot of this benefit will be driven out of Asia, and we need to catch up on making sure that we also see some of those benefits in Europe and come to terms with our role in this.
“I would hope that AI will become a commodity, where it's underlying a lot of other things, working in the background, that a lot more people have access to it.”
Chris: And, after all that, we have heard from you in this episode. If I were to call you an AI missionary, what would be your message to the people out there when dealing with artificial intelligence?
Michael: Yeah. I mean, personally, I've been kind of branded as an AI thought leader for a long time or evangelist. But my message on this is very clear regardless. Don't be afraid of AI and disruptive technologies in general. They are here to stay they're not going anywhere. I feel that to fully make use of it and understand it you have to get involved on a personal level. I look back at my time in mentoring, and my first exposure, and I can only recommend that if you haven't started that journey yet, start it as a mentor and see what it's like, and see what the opportunities are for mankind to work on AI for good cases that will bring mankind further, whether it's helped to stem the disruption for COVID or other things that will come down the way.
Chris: Yeah, and to be honest, I experienced exact the same thing. So I also did my, you know, my PhD studies in the intersection between managing innovation and artificial intelligence. And I've seen the exact same thing when discussing the topic. So oftentimes, it is just the same thing as it is with all new kinds of technologies or technology areas. So in the beginning, there is a buzz around that, and there is fear. And at a certain point in time people and the economy actually does understand what is possible, what is not. And how can it be applied, of course for good and bad, as with most of the large and big technological changes that we have in the past decades. So, it will be very interesting to see which side of the good or the bad the artificial intelligence techniques will be applied. But certainly, at some point in time there may be regulations, even laws, and ethical questions around that. I mean, the discussions are heating up these days, at least, that's what we see. But it's great to hear some really, real-world use cases from you today in this episode, rather than just having a high-level discussion on what it could be. So that's very helpful. Thanks so much for that. Please allow me one last question that I ask all my guests. Now, when you look back on your career so far, what would you say were your most favorite Innovation Rockstar moments so far?
Michael: So, I certainly enjoyed my time working for a number of Silicon Valley firms and doing innovation. But I feel that my first exposure to artificial intelligence, the one that came from mentoring, that changed my career, my life, my future outlook some time around 2012, that's kind of like my favorite moment. Because otherwise, I would still be in compliance and risk management and very different financial services areas. Now I'm kind of looking at those areas with new glasses on or a different view. And I apply some of the things that I learned helping startups to innovate, to my own line of work and to things that we can do for us as a firm but also for helping our clients in the market.
Chris: That's a solid reply. Thanks for that. And then Michael, thank you so much for all the insights into your work and the exciting use cases that you brought to us today. And yeah, to everybody listening or watching. If you're into the topic of AI and want to address questions to Michael or discuss some of the use cases you just heard, then feel free to leave us a comment on this episode or drop us an email at email@example.com. Michael again, thanks so much for your time, it was really inspiring listening to your stories.
Michael: Thank you for having me, Chris. It's been a pleasure.
Chris: And that is it, we are at the end of our episode. Thanks for listening, take care and see you in the next episode. Bye-bye!
About the authors
Dr. Christian Mühlroth is the host of the Innovation Rockstars podcast and CEO of ITONICS. Michael Berns is Director, AI & FinTech at PwC.
The Innovation Rockstars podcast is a production of ITONICS, provider of the world’s leading Operating System for Innovation. Do you also have an inspiring story to tell about innovation, foresight, strategy or growth? Then shoot us a note!
See the ITONICS Innovation OS in action