25Minutes: Insights. Expertise. Impact.
In just 25 minutes, I deliver concise and thought-provoking conversations with top minds in technology, cybersecurity, business, culture and enterpreneurship. Whether you’re a technologist, executive, culture-enthusiast or someone passionate about growth, each episode explores trends, strategies and ideas that shape success.
For those with limited time but unlimited ambition, 25Minutes offers actionable insights and fresh perspectives where they matter most. Your time is valuable. Your 25 minutes. Your advantage.
Contact: eliel.mulumba25@gmail.com
X (Twitter): Eliel Mulumba (@ElielCyber) / X
25Minutes: Insights. Expertise. Impact.
4 - Adrian Ott: AI Disruption how every industry will change now - AI, GPT, Switzerland, Pharma
Ten years ago, would you have imagined companies appointing a Chief Artificial Intelligence Officer? AI is no longer a futuristic concept - it’s actively reshaping industries, automating processes and redefining business strategies. In this episode, we sit down with Adrian, a forensic expert with over 18 years of experience, specializing in forensic services for financial organizations across both the public and private sectors. We explore AI’s transformative power across industries, where it’s making the biggest impact and why Pharma is emerging as a key player in AI-driven innovation. Which AI use cases are truly effective for businesses? Where is AI proving to be the most disruptive? Adrian shares his insights on the biggest opportunities for automation, the common mistakes companies make with AI adoption- and how to fix them and how AI is revolutionizing first- and second-level service desk tasks. He also reveals his personal strategies for staying ahead in an AI-driven world, along with recommendations for individuals and organizations looking to fully leverage AI’s potential. Plus, we uncover his personal approach to staying ahead in a world of constant AI breakthroughs - and his top recommendations for individuals and organizations looking to deep dive into AI.
Our Guest:
Linkedin: https://www.linkedin.com/in/adrian-ott-517759143/
https://www.moneycab.com/it/adrian-ott-chief-ai-officer-ey-schweiz-im-interview/
25 Minutes Podcast
Hostey by: Eliel Mulumba
Audio editing & mastering: Michael Lauderez
Join conversation on LinkedIn: www.linkedin.com/in/eliel-mulumba-133919147
Adrian, welcome to 25 minutes. It's a pleasure to have you as a guest on our show. Adrian, your chief artificial intelligence officer have more than 18 years of professional experience with regards to forensic services. In the public sector, but also in the private sector, you have been working with large clients and actually switch from a more technical role to also strategic role when it comes to everything around for forensic services.
And now, since you also chief artificial intelligence officer, you're also working. In helping clients to have the right use cases and to build up their platform. I'm very happy to have you here as a guest on our show. And I'm wondering, you imagined 20 years ago to have a position of a chief artificial intelligence officer?
No, actually not really. By the way, hello, Elial. Very nice to be here. No, you're right. I, I didn't even know that this role will exist in the future. And I never had any plans to be a chief or the fifth intelligence officer. I still have a hard time to pronounce it. But I was always very.
much interested in programming in artificial intelligence. Back then it was video games. So something that fascinated me was always to program an enemy that moves and behaves like a human. So I had some nerdy friends and we we were deep into programming our own video games. It was very basic back then.
But there was always an artificial intelligence parts that that catch at me, but I never pursued that from professional level.
I mean, before going into the details about artificial intelligence, and I know that people are quite curious to have your take on that and how you see actually the current market situation in Switzerland, I would like to talk a bit about your career journey. So we talked a bit about the different paths that you had and the different employers you have been working for. What are the three, four key milestones in your career that you are most proud of?
Yeah. So I think as I had only, I think three employee year employers. So I think that is maybe my three main milestones that I had. So after I finished my studies. In economics and IT, I joined Credit Suisse so I was there part of an IT team. So we were building an application around trade finance.
So trade finance is something where you can do all your deals when it comes to shipping large amount of goods. And we developed a tool at Credit Suisse that handled the whole trade finance business over a web interface And initially I was more on the programming side. But then I also had quite a lot of exposure to clients.
So they called me, we had a hotline back then everything was done by the same team. So the programming answering to hotlines, but then also the selling. And I, I went out quite often to clients. I wrote them our tool they had then feedback and said, Oh, it would be great if you could do this. And if you can click and that, and then it would do something automatically.
And the good thing was I then could go back and then just like get it done within the day or two. And that was good times back then. I think it was around 2008, 2006. So there was much less fragmentation when it comes to I. T. I think that was my starting point in, I would say, requirements engineering, programming, working in a team.
Mm hmm.
years at Credit Suisse, I moved to the office of the general attorney in Switzerland, so the Bundesanwaltschaft, there I was tasked to build up the IT forensic department. So you have to imagine the office of the general attorney and the federal police.
Whenever you have like white collar crime or even threat actors within Switzerland, you often do a house raid. So the police will get all the data. Early days, it was often physical folders, but already at that 2012. A lot of the data was obviously on, on hard drives, on laptops, on, on, on this.
And We had to find the evidence that somebody did something wrong or maybe didn't do something wrong in a very short period of time because people often end up in the prison. But then you need to like find enough evidence in a short period of time to actually keep them in prison. And for that it's also called e discovery.
We built a workflow. to process masses of information through a forensic pipeline and then start with keywords trying to find the smoking guns that somebody did something illegal, for example, or was involved in a white collar crime. And yeah, this was a, this was a very big milestone where I really learned a because it basically the premise was we have Tons of data, terabytes of information in all different kind of formats, sometimes mobile phones.
How do you find the smoking gun and the relevant information in a short period of time? And this is also where I came very close to artificial intelligence. Back then, it was machine learning trying to figure out algorithms, natural language processing and systems to, to find just the smoking guns very quickly.
And I don't know is it already boring or should I just go to the last step?
I mean, it's not boring at all. I think it's really spot on because the next element is actually, when was the moment in your career where you realized, Hey, I, this is for me,
Yeah,
grab the opportunity and embrace it.
yeah, this came actually much later because in the beginning, let's say in in business terms, when you work for a large corporation, AI has an impact or already had an impact also in forensic, but it was. To a much lesser extent, because all these NLP methods, they didn't work that well to find smoking guns, for example.
So there's still a lot of manual work that you have to do. So we experimented there with graph databases, communication map or just like by training an algorithm that is able to. Based on older smoking guns, find the newer smoking guns quicker when you have a new house rate and new data, and it kind of helped, but it was just not enough to get me too deep into an artificial intelligence only topic.
So before I did that, actually, I moved after five years to consulting. I built up there, the. digital forensic practice. We did a lot of cyber incident response, but we also do for companies like internal investigations or where we have a large litigation against other parties. Basically, we set up the same workflow.
How can we handle a large amount of data in a litigation or in investigation for a company and find the smoking guns as quickly as possible. possible and all the facts that is required to, to solve a case. And there recently especially with the early stages of large language model we started to experiment much more to have like an AI analyzed.
It's massive of emails and documents because they were much better in reasoning and in understanding context. And when JetCPT came out and there was an API we could use from Microsoft Azure, we basically created like the first productive solution worldwide. And it was our forensic investigation solution that was able to scan.
Masses of emails and documents and look for potentially relevant information in the investigation or litigation.
Mm.
That's where it started and where I felt, okay, this has now really a business impact. It's more than just a more niche topic. It's something that creates direct value to clients if used properly.
And that's how I somehow ended up becoming the chief AI officer. Because the top management said you cannot only do that for forensic. We want to do that for all the different service line in all the different sectors and help you develop these solutions and tools. That's where I am now.
mean, that's, that's really spot on and thank you for elaborating on it. You were referring to many topics we would like to cover also as part of this episode, Chachapiti, language model, business value and also making sure that it's really helping your clients. And I was wondering, based on your experience, right?
You have been working in the private sector and the public sector, which industries. From your perspective will be most disrupted by AI in the next years.
So I think it's not that easy to predict because I feel every industry. Will be disrupted to a certain extent. I see especially pharma is an industry where there goes a lot of effort in research and development, let's say drug development, protein folding. So these are fields that can benefit a lot from artificial intelligence because I'm finding maybe customized medicine.
So there was this example that was also mentioned in the US that in the future, you will have a blood test and an AI will be able to see small traces of cancer cells before they really manifest. And within 24 days, you will have your own info. Sorry. Injection that you can use that is personalized for your exact cancer cell, um, to heal you and to, to, to kill these, uh, cancer cells.
And today, obviously all the cells are different. It's very hard to have a one approach for cancer, but if you have like a, something where an AI develops your personal vaccination that is a game changer, obviously. And I think pharma is very much exposed to that angle. But I think in other industries, wherever R& D plays a big role, I see that AI is a big disruptor.
Um, and often I also see margin pressure for just the classic industries where maybe you have less R& D focus, but there is still operational excellence that you need to cover and that you have like the best processes that you optimize, maybe tasks that are currently very time consuming. And maybe to maybe make the other the other step because we thought a lot about it when you look at banks and the financial sector there.
I feel a lot of money is also done by trust. So you trust your relationship manager and you trust that the bank is safe and your money is safe. And there, I think it's not the best idea for an established big bank to say we outsource now the relationship manager to a chatbot ai, but I think they will be augmented a lot by AI to to help them find the right product for their clients.
And I mean, based on what you just mentioned, I can imagine that you're talking to a lot of clients and executives out there from different industries when it comes to AI based on your view here in Switzerland, how does AI already influence the business strategy and decision making within companies?
And
see, I think AI is a big topic in Switzerland in for most of the companies. I think there was a UI has done some research across all the European countries and Switzerland is the one that most companies said that AI is one of the top investment fields for the coming two years.
Much higher than the rest of Europe, for example. So I think every company tries to find the use cases and how they will be disrupted at the moment. And placed a lot of focus on, on, on these use cases. And we see a lot of activity, but in the end how much it already now is. It's impacting the strategic topics.
I think it already impacted because even sea level, they maybe let check GPT sound their ideas, or maybe provide a prompt with information and ask, what could I do with that information? How should I prioritize? And so on. I think there's already day to day use that influences what companies do, but I think the companies that already have integrated it.
So deeply into their structure and into their processes, I only see maybe a handful and then it's mostly the very innovative ones and also like the midsize smaller ones and large corporation often have a lot of problems to solve before they are that far.
I think you just mentioned something that catch my attention when it comes to integration of use cases within the company. We just had a guest here I think for the second episode, it was related to blockchain who said that there are some use cases, but he just see a few companies that are really applying those use cases and also understand the value.
So I would be curious to understand which use cases. With regards to AI already working well based on your observations. Mm-hmm.
Yeah. I think the first use case that most companies have is they have their internal chat GPT. Or like some form of it. It doesn't need to be open AI. It can also be Gemini. It can be an offline open source model or, or Claude. But I think most company first, what they needed to do is, okay, I don't want our employees to go to the official chat GPT website with their private account.
And then they ask questions and they put in client information or sensitive information. So the easiest, what you can do is You build your own chat, GPT within your company's domain. And you have tech providers like AWS or or Azure that allow you basically to access the brain of chat GPT within your company's domain.
So no data leaves your, your company.
Mm-hmm
and I think that is a use case that already helps a lot in day-to-day business. So people have a place where they can ask these questions. And that has maybe the biggest impact where, where I also see then there are things like co pilot, for example, I think that's a famous one from Microsoft.
So Microsoft basically says, if you have office 365 and you already have word and, and email outlook in the cloud, why don't you get co pilot? And with Copilot, you have basically today's GPT 4. 0. So like it's an open AI brain that, that it uses, but it can also analyze your internal data. So like your emails, your calendar, and all the files that you have access to.
And it gives you actually not only answers that that GPT would give you, it actually gives you answer that incorporates the data that you have access to in your company into the answer.
Mm-hmm
And I mean, there is good things and bad things. It works sometimes very well, but sometimes it also doesn't work very well.
I think the copilot is still a little bit overwhelmed with the amount of information that it gets access to. Um, but, but this is the two things I, I have seen a lot in companies. But now what's more and more is coming out that you integrate AI in their processes. For example when you, when you do marketing or when you do a campaign, you can optimize a lot of steps by chaining different large language model together that will basically work together to come up with, with marketing material or.
So to any kind of material that you need, it doesn't need to be marketing. But the key there is that you analyze what are the process steps needed where currently human work together and which portion can be taken over by, for example, a large language model. Or sometimes it's not a large language model.
It can also be a written model of it can be just like a script and so on. And then you start actually automizing core process in, in your company. And maybe one example that I, that, that I feel works very well is to port first level support, maybe second level support. So people have a problem. Instead of reaching out to a human in the beginning, they, they ask their question to a chatbot.
The chat bot has access to like the pro problem shooting documents. And if the chatbot cannot help you, it'll actually create a second level ticket.
Mm-hmm
And there's more and more tasks that even the first level chatbot can already do. It can access your computer, it can see if there's certain settings you need to change.
So it gets already quite advanced, and I think it'll get much more advanced in the future.
Yeah. And I mean, this is definitely something that I can confirm based on what we also see in the security market. There are a lot of tools out there that, uh, clients can use. where AI functionalities have been previously more an option that could be enabled or disabled. But I think when we're looking into the product roadmaps of those tools, it seems that those functionalities are becoming more and more core and a part of the actual functionalities as businesses and clients can also use. you have been talking also about things that do work well. Use cases that are already integrated. I would like to understand a bit what are still common mistakes. Definitely.
not understanding it enough to do the right decisions. So for example, I see often that they, they try to come up with use cases and they do it in the team. And usually there's always the wish to make something better and easier. And then what I have seen a lot is that.
You start with the business case that has then like fancy return on investments. So somebody is a graphic, how much revenue that will bring you or how much relief it will bring you from your day to day work, but then the components that they use then to go into that and business case is often is technically, it's, it's just not feasible either there is still.
The risk is too big that the model does something wrong or doesn't understand something. And also, very often, it's much easier to think about how it should look like in a perfect world than when you actually do it. And when you really want to implement AI use cases on a company, on a corporate structure, you have to think about a lot of topics like risk.
Like what data can we actually use? Are we allowed to use client data, even if it is an internal use case, because sometimes you contractually cannot use data from clients for, for AI use cases, or you, you just, you build something and you see it's much, much harder to make it good. It's a little bit like.
Self driving cars so you can get self driving cars quite well for maybe a highway with with no No real traffic and it starts very quickly But then there is so many different things that can happen and then your self driving car doesn't work anymore So the whole journey to make these solutions really good that they actually can be working independently and without human oversight or minimal human oversight.
I think they're it needs good people and, and good understanding before you start developing something that will then fail in a year and people lose motivation to do AI use cases.
Thank you very much for elaborating on that. I mean, we're also approaching a bit the end of, uh, this episode and, I was wondering what book, tool, or any model has influenced your work with AI the most, and if listeners can take away just one key insight. What should it be from this conversation?
So look, I, I read a lot about artificial intelligence. I think there's a lot of great books, but Maybe something that is a little bit my secret knowledge pot is actually a YouTube channel. So there is one called AI explained so that the YouTuber that reads all the papers that are created tests all the models and brings really good thought leadership.
In a short period of time. So when I drive to work every morning, I actually check his YouTube channel and see if there is a new piece that I can listen to in 20 minutes, because it is very good in like breaking down complex information in something digestible and get like the right thought leadership.
So I really liked that. Maybe. Yeah, some key points. I think we initially also talked a little bit about career and what you can do and, and, and how you maybe can work for me to AI. I think it's important. And I think you asked me before, did I had a plan to become chief AI officer? And I can definitely say, no, I had absolutely not the plan.
I didn't work towards like a goal. So I, I would almost say I went a bit with the flow. I always liked technical work and deep dive into systems and understanding the logic behind it. And if you are very curious in what you do, and if you. Try to educate yourself and use these tools, even if you maybe have no time, you learn a lot just by doing it and by and by being very curious.
And then you can go and propose maybe to your boss or to your committee. Things that you have heard show them what you feel is the right way to do. This is how I, it is kind of a lot and that's also how I ended up being, uh, now here where I am. So maybe that is a takeaway. Stay curious, educate yourself, don't wait for others to do it and, uh, yeah, get some sources that you can trust.
Thank you very much for sharing all those insights with us. It was really a pleasure. Adrian.
