Transcription
I know this is the final chapter before a big moment. So, we really appreciate everyone uh sticking around and coming in. Um the the main act will follow. Um I'm Sean from CW8. I was on earlier. Um I would love to introduce my panel. This is like probably the most current topic and theme of today and it really intersects with what are big enterprise doing, what are academic and researchers are doing and what are people actually doing about it. Okay. Um so I'm going to ask my panel very kindly to introduce yourselves nice and loud and clear. Daniel, fantastic. So I'll start. Very nice to meet you. So I'm Daniel. Um I'm from the ETH AI center which is a competent center at ETH Zurich uh where we focus on one hand on uh doing research in a very interdicip disciplinary way um that I can share more of later but also focus on how do we get these research insights um into industry and society and I think that's particularly interesting because as we all know ETH um enjoys a very good reputation but research by itself um is not as useful as if you're able to really transcend and translate that into something useful. for industry and society which we have a strong focus on.
First of all, congrat congratulation to the first anniversary of the internet computer and thank you very much for having me. My name is Gif Yalak. I like I work for Cisco systems here in Switzerland. For those who don't know Cisco, Cisco is the global market leader in the networking infrastructure and enables connectivity at Cisco. I'm part of the management in in Switzerland and lead Cisco's country digital acceleration initiative. This is a corporate co-inovation co-investment fund. We are active in 50 countries uh reach twothird of the world population and the goal is to accelerate innovation and the digitization. I've spent many years at ETA at Harvard University and had the chance to use some of the fastest supercomputer for my research in the past. Looking forward to the discussion. Thank you Dr. Gar.
Sam lovely to meet you. Uh so Sam Burman uh from Hydrickch and struggles. Uh we sound like a law firm. We're not. We're a global executive search firm out of the US. Uh if you don't know what executive search is, um it is effectively we do CEO and board recruitment down to sort of VP level. Um we're most well known for our depth and credibility in the technology sector, which is why I joined the company 9 years ago. Fast forward to today, I'm a partner in our technology practice and globally I lead our frontier tech practices. today that cuts across five critical and core areas AI data analytics as one cyber security crypto and digital assets health tech and government and defense tech as it relates to the topic of the moment AI data analytics that actually predates me so we launched that in 2012 when big data was the buzzword um and over the last 13 years we've evolved with the market uh to where we are today I would sort of categorize ize um or characterize I should say the first 10 years being working with clients around data adoption, data transformation and today of course AI transformation and in the last three years we've continued doing that but we've been doing a lot of work with the AI builders themselves and helping the AI native companies build the right leadership teams to scale and grow. So thank you for having me and I'm looking forward to diving into our panel discussion today. Thank you.
So actually I'm going to stay with you Sam because I I have to say as a as an entrepreneur and as someone working in business this is a people situation and it is is it the biggest conversation you're having with leaders right now like their readiness y to think about it you know it's it is it's there they have to have a plan right talk to us talk to us about what what you're hearing. Yeah. So, you'd hope you'd have a plan for sure. I I think what we've noticed over the last 12 months is this whole topic of alignment or misalignment amongst leadership teams. Um, and if I could provide some context because I think it's critical to answer your question, Sean. So last year um at the direction of our clients, we created a proprietary diagnostic which looks at leadership teams and how aligned they are brackets or misaligned closed brackets around their readiness for AI adoption um and AI transformation. And what we created was this really interesting tool whereby you're effectively we ask questions against 16 key categories for AI readiness. And what it does is it compares the results of the CEO's perception of the company's readiness and the aggregate score of their leadership team. I can tell you it makes for some pretty amusing reading and some quite uncomfortable conversations during the debrief. And what we've noticed in the last six months of that tool being live is irrelevant of the company in terms of sector size, maturity, geography, etc. The same five areas are cropping up as the key areas of misalignment amongst leadership teams. The first one is AI strategy and sponsorship. That is effectively a misalignment between the CEO and the leadership team around do we have an AI strategy? If so, who owns it? Is it any good? Where are react and its execution? And do we have the right sponsorship in cascading it down from the leadership team into the organization itself. So that's the first area. The second area is all around talent. M so do we have the right mix of skills and capabilities on our bench currently to enable the execution of the transformation and once the transformation is done if that is such a thing are we set up for success in a future state. Third area of disagreement or misalignment is around data quality. Do we have accurate and accessible data of which to leverage and build AI capabilities on top of? The fourth area is all around um governance. So do we have the right frameworks, the right guard rails to make sure that when the implementation goes ahead and is quote unquote complete, do we have the right trust and safety and governance in play? And then finally, the fifth one is what we call impact measurement. So how good is we how good are we as a leadership team as a company at measuring the impact of other technology implementations or ways of working and transformation programs that fifth area might feel like a distant fifth to some of you in this room but the reality is as we all know if you're not careful AI can very quickly become very expensive and also not deliver the impact that you need if you've not got the alignment in those four areas listed above it. So, it's super important. And the final thing I'll say, Sean, is our findings map very nicely to a recent McKenzie report, which came out a couple of months ago, which true Mckenzie style, it's about 50 pages long. So, I put it through chat GPT just to get the distillation. And what it effectively was saying is the biggest friction points for AI adoption is not the employee base, it's the leadership team because they're not steering and operating quick enough. So that's some of our learnings in the last 12 months. Thank you.
Thank you so much for that. And I I really I suppose, you know, obviously understanding how organizations navigate that is is going to define their success really, you know, as a as a follow- on point and and how inaction really possibly isn't an option. Uh and essentially, but let's strip it back just one layer. Uh Daniel um just speaking specifically from your perspective at the ETH um AI center for innovation talk to us a little bit about the foundational piece that you're working on and what your innovation center is focused on and some of the the areas you're tackling at the moment. So yeah absolutely and I mean I I can just say you know as well building that bridge um in terms of the collaboration we do with industry partners we do see that very much you know kind of the struggles that at the end it's not about the the tool but really on how you implement how do you advance this change um changes in these organizations but still you have that need of having you know better large language models for instance um um but first and foremost you have to imagine the ETHI center has a very wide area of of the research performance when it comes to AI plus another topic. We have 123 professors linked to the AI center which makes us one of the um largest in Europe um and um through that we have I think it's a quarter almost a third of all ETH professors that are linked to the center. So that means we have we cover so many different aspects. Yeah. When it comes to the research bit through our industry partnership programs and and um support for entrepreneurs we have I think at the moment about 60 62 startups. Yeah. So as well very strong focus of really kickstarting as well um from these top tech entrepreneurs um that want to endeavor on on on kind of really application based um um or deep tech AI startups. Um so just to give you kind of a sense of of what that ecosystem uh looks like but we have besides that focus of on this research bit and the translation or transport into industry and society we have many different um initiatives that we put forward and I think particularly I mean there's different aspects you know think of uh we have a red teaming network which is um we team up I think it's about 10 12 different companies across the value chain of large language models that typically the teaming heads the the the security and technical responsibility sit in a room. We exchange in terms of what are defense and attack mechanisms when it comes to um large language models or as well applications you build on large language models. But that brings another of the initiatives um um at the forefront which any kind of application you think of um at the end whenever you use a large language model as a base you have to choose one and at the moment you have basically a choice between the very commercial models and then so to say open weights model you know think of the deepseek think of of the llama and so on but you have with all of these the same problem that a you're fully dependent on the commercial terms and with neither of the models you have full transparency to understand what are the security risks, what are you know what what are the tra what's the trading data behind and so that means an initiative initiative that we started and now is kind of um u packaged under the what we call the Swiss National AI institute is um an initiative that focuses on building a Swiss made large language model. We assume in the in the next um month or so um we will have the first version of that as well um available. there is an um uh there is a strong focus on and it will be the first model that includes the most different set diverse languages um and um I think it's over 70 different research um um labs that are involved so besides ETH Zurich and APFL loan we have many of the other Swiss universities and um can use and maybe this is something that not everyone is aware of we have in Lugano through the Swiss national supercomputing center the seventh largest supercomputing uh power available globally and we can use that for research but also for building these large language models for the advancement and use for industry and society.
Amazing. So collaboration is key. Just following on from uh Sam's point. Um Dr. Gareth Yalik and I'm going to call you Gareth after that. Okay, Gar, please share with us a little bit um more about um we we've touched on some of the parameters that Sam highlighted in the study and again Daniel has touched on as well from an academic perspective, but talk to us about I suppose the scalability, resilience, security and you know what's keeping Cisco awake at night at the moment? What problems are you solving? First I think it's important to look into Cisco's history where Cisco is coming from initially. We have been founded exactly 40 years ago and ever since we have been contributing to power the internet as market leader in the networking infrastructure. I would say the internet as we know it today would have not been possible without Cisco's four decades contribution and I know this sounds like a ridiculous claim but as we speak more than 85% of the global internet traffic is running through Cisco infrastructure and the success why have been so successful in the past is one thing our commitment to deliver state-of-the-art open transparent reliable and secure solution this is a commitment that we are continuing Contining we are honoring our commitment to enable our customer to build an AI stack that they can fully own, fully run on their side. And one example, we've recently introduced our so-called Cisco AI ports. This is a box that includes the entire AI stack that you can install on prem. You own it, you run it. All the governance is on your side. It's scalable. It's possible to do it with a oneman show. It's also possible to scale it to a multinational organization to implement very concrete reference cases, AI cases that are relevant for your institutions and for your vertical and I think this has been Cisco success and we're looking forward to enable our customer to be ready for the second phase of the internet.
Fantastic. And specifically um can you just shed a little bit of light on the types of public and private enter you know projects and initiatives that you're focused on in fulfilling that mission. What what could you share some of those? So we are working in in in the entire domain from cyber security in healthcare in energy in upscaling with institutions such as ETH EPFL. There are many aspects in in in the projects that we implement. I can share a few examples in in healthcare for example. One of the project we implemented here in Switzerland is with the university hospital in Balist where we combine our forces Cisco 20 other major corporation who are joining it in the efforts and we bring together the different components that is really needing. No single actor will be able to to implement this use cases. ETH by the way is part of this initiative and the focus of this project is to enable education of f future surgeries surgeons without uh putting patient at risk. They are able to collect practical experience with virtual reality and for that you need a stable AI ready infrastructure. This is one example. Thank you G.
Uh Sam with leaders. All right. Um let's think about it. It's almost like a therapy session. It must be where is their head at? What's what's the situation with terms of the leadership attitudes towards this? Like you know I'm sure you meet almost evangelists who are like all in on the tech and then you meet others who are fearful. Where is it at the moment? What would you say? How would you describe it based on your research? I think there's a healthy balance of optimism and pessimism. I I'll certainly sit here and say I'm very much an optimist uh and hope that I don't get replaced otherwise my children won't be very happy with me being at home all the time. Um I think in the context of what what we're seeing if I could maybe give you a different angle is a lot of not just our clients but organizations that we spend time with are treating this sort of topic of AI adoption and transformation purely in the lens of it's a technology program and we think that's a limiting factor because ultimately if you treat it as purely a technology program you're going to put the majority of the onus of responsibility for delivering an impact at the door of a CTO or a CIO and that's going to have limiting impact across the enterprise because again in our view we see AI as very much a team sport okay where used the word collaboration earlier on the leadership team of any organization as long as you're not an AI native company you need to look around the table and see what can everyone bring to the party in the context of driving AI adoption and transformation Now the CTO CIO is going to have a pretty important role. However, if we go around the table, I'll pick a couple of them. You know, the chief people officer or your HR director, hugely important role because this this technology that you know um Gareth's working on for example, that is going to have significant implications across people, talent, teams, organization, culture. That's at the heartland of the people officer. You take the general counsel and chief legal officers of your respective organizations, they've got to be hot on compliance and making sure that if you work in a regulated regulated industry, any AI technology you're using has the suitable compliance uh to uh to sort of move forward with. And then you pick a final one and a chief financial officer. They've got to free up suitable capital to allocate to spending lots of money with Cisco. Um, but equally they've got to make sure that they're tracking the ROI to make sure that the cost is not getting out of control and the impact is actually driving genuine impact in the organization. So that's some of the things that we're seeing in terms of what what they're wrestling with behind the scenes. And I think there's a healthy balance of um optimism and pessimism.
Very interesting. And actually Gareth I might just ask you just given um the enterprise capabilities of Cisco there's a lot of options to choose technology right you know it's very there's a lot of options out there and actually equally for those leaders they're being almost asked to take a bet on is that the right pro is that the right protocol is that the right application to use within the organizations how how would you view just from a technology perspective and this is a slightly different conversation we're just following on what Sam said, how would you see how organizations should think about the range of options they should be considering? So, so first of all, what what we see at our customer base is that we have an fragmented IT infrastructure. We have an outdated IT infrastructure that is currently not ready to implement AI use cases. And one of the reason actually is the the skill shortage. We have at the moment we have the challenge of more than 12,000 open position in Switzerland in it and AI cannot be filled. So first the companies need to have the right skill sets inside to have the knowledge inside. And we see many companies are struggling to build that up. If you don't have the skill set inside, you need to partner with major corporations, consulting companies who are really deep dive into the technology to help you uh establish a strong data and AI sec AI strategy that is integrated into the business that you're running. This is mostly missing because most of the management teams that I'm talking to look at AI as a technology project which should not be the case. It should be a strategic project at the top level of the management and then you can you can come up with the right solutions once you have the the knowledge inside the companies.
Gosh, I might just bring Sam back in there very very quickly on on your view of of how it should be tackled within the organization in terms of deployment you know what you know especially if it's coming to procurement and the CFO weighing up how much it's going to cost him on his bottom line. Yeah, I I think well we've talked about this, you know, behind the scenes. You know, I I mentioned earlier on that I think too many organizations see AI adoption purely through the technology lens. You should really see it as a people change program, a very complicated people change program where if you think about all the areas of an organization which AI in theory could impact. I mean everything from org design and or structures, job families, role displacement, role creation, role augmentation, pay structures and most importantly culture. That is a big big gnarly hairy topic to unpick which not one person should do but typically you know a chief people officer would would lead that. And if you put that in the melting pot, you can think about your aspirational goals for what you want AI to deliver, but you've got to root it back in this topic that we're talking about, which is what does it mean for the people and for the organization in question. And I suppose the final thing I'll say is we are not uh strategy experts, right? We work with companies to build their leadership teams, to build their boardrooms, but from what we see from our purview, from what we hear from others and what we read and all that kind of good stuff, what it seems like the best AI strategies are the ones that are in service of the overall business strategy. So if you think about your average business strategy, there's going to be three, four, five top priorities for that company. The best AI strategies that we've seen in play are the ones which pick two or three of those priorities and quickly figure out okay well how can AI turbocharge the delivery of those priorities that's where the focus area of a successful strategy should be versus a whole host of disperate disconnected proof of concepts running around addressing multiple user cases which aren't really moving the needle.
Okay. And so another way of putting that is focus on the path to enterprise value versus tinkering around the edges. It's um and and I just want to bring Daniel back in because things are moving so fast like it's exhausting. Genuinely it is for me anyway. I can't speak for anyone else. We love it. It's fantastic high velocity. But you know you're working with some of the academics and the developers who are coming up with all the iterations here. there's a great sense of collaboration on a national level within Switzerland. Do you invite private enterprise to come and talk to you? You know, maybe is it is it time that the leaders of these companies need to come back to the universities and have conversations? How can they get involved? How can they support the academics and the research? You know, what can they do? Especially with your center here in Zurich. Yeah, very good question. And I think very important as well to understand that the center although it focuses on on breakthrough AI research it is very closely connected to the industry bit right. So in a sense we have uh on one side we have the latest research on AI related that um converges at the center we have through um different uh industry collaboration programs that we have. So everyone that is interested can plug into the AI center with different in different ways. We see as well what the pain points are. we see the problems they struggle with. Um so take that together with the entrepreneurs and the startups that we that we um um help u move forward. We see kind of these these whole like 360 circle of the challenge of those next things that are coming around the corner. M we see the problem sets of the companies that they at the moment have and struggle with and very often you know I mean this is exactly it's not just the technology it's like so many other layers of complexity but we do see kind of where their today's pain points are and we see what kind of applications different startups kind of work with and I think if you take that together that gives a very interesting insight to understand or maybe as well to estimate a bit kind of where will be when you think a little bit ahead what will be the big breaking point. So the things that we if we don't get them right, we will have even bigger issues. Right? So that's where the security aspect of these large language models comes into place. The fully openness, transparency, so kind of trustworthy um large language models comes into place. Um it's you know it's kind of sometimes you we forget that whenever we talk about these applications on top of large language models, it's like if you think of building a house on top of a piece of land that you do not know what's underneath nor you have fixed commercial terms. Yeah, no one would build a house, but at the moment there's just no alternative, right? And I think that just shows how early we are in that process, but also how important it will be to have something that you can more reliably build and use as a foundation to then build this application on top.
Um, as a follow on point and obviously our wonderful hosts here have countless PhD developers working um, with on on ICP and within Definity and and and Zurich is an unexpected home of research and innovation. I know a lot of praise is given to the West Coast. It sounds like you guys in Switzerland have really got it together in terms of how you work to together and and collaborate. Do you think that um Europe has an opportunity here to actually be a center of of greatness, let's say, in this new chapter of the world? I know this is a we didn't discuss this, but I I just wanted to put you on the spot slightly. No, definitely. I think the answer is quite simple. Yes. Because um we are I mean Europe as a broader, you know, there's more complexities within Europe and and more fragmented. But if you just look at Switzerland for instance, we have some of the top-notch researcher in AI as well. We have this top um infrastructure with the Swiss National Supercomputing Center, Alps, the supercomputing logano, but we also have an amazing density of large companies from across the globe in Zurich. So within kind of a 10 kilometer um hor distance, we have these top companies, you know, with their AI labs and research as well. when you think of um um you know the the obvious one from Meta um Google and so on but you have as more recently as maybe some of you heard entropic open AI that open office here we really have we think we have one of the most dense AI researcher ecosystems in Zurich but we have the pieces as well in Zurich to play and to contribute the significant role as well within the European context and that means and this is also not predis but I think it's very important to mention companies like Definity that also work with the ETHI center, they are basically at the forefront of showing how you can leverage these resources that are here. Um and and I think that is something that every company that wants for every company where AI becomes more and more of a central strategic piece that they need to focus on. they you know because the space moves so fast and you said it earlier it will be ever more important to understand where is that space going and what are even if you don't understand exactly where what are the things you now need to start get right talking about change management talking about um getting your your the workforce as well open um figuring out how you can enable the individuals to come up with solutions and and challenges that are not just the technology bit but we also need that technology bit in order to really be on the forefront of it.
Super. So, we will get to some closing comments, but Sam, just one, I suppose, to build on your point about that readiness and getting the leadership team on the same page. Is that literally getting everyone in the room and saying, "Guys, what's our plan or is it is it more is it more complicated than that?" Yeah. Yeah. Well, like I mentioned of this uh this tool we rolled out in December. Uh it's called AIQ, which just simply stands for AI questionnaire. And we've had about 35 companies go through that. And I've sat in I think three or four debriefs. So the the leadership team does the survey. We then get the results. We then look at the results and we look at the sort of pockets of misalignment. We then go back in and we sit down with everyone who completed that survey. So the CEO and all of their leadership team in a room with you know me plus at least one if not two others and you basically hold up the mirror and say right this is how you all feel and you put the scores on a big screen and it basically looks at as I said what the CEO thinks versus the aggregate score of the leadership team and there is a oh moment like we are totally misaligned and then it forces a very honest conversation. Yeah, I I'll give you one very quick anecdote. I know we're time. Please do. Um the first debrief that I did was with a very large media company uh in the tens of billions of dollars and one of their areas of misalignment was their definition of the AI strategy. And what I mean by that is half of the leadership team associated AI strategy with purely gen AI and the other half saw it in the true definition which is the holistic definition of AI where gen AI is but a component. So you had half the team who thought, "Oh, we're doing a pretty good job on our AI strategy because they thought about it in purely generative AI terms and the other half were like, we're nowhere." And it was a very simple I'm giving you a very simple anecdote, but that's how basic some of the areas of misalignment are amongst leadership teams because it's either being treated as a tech project, it's not being treated at a leadership level, it's in silos, it's in the organization. Um, so yeah, it makes for interesting conversation. It needs there needs to be a culture of curiosity uh to get on the right page.
So this is the uh closing remarks before the hottest moment of the day. Um and I'm not going to ask you guys to do you can choose. You can say because it's really fastm moving. You can either come out with a prediction or something that's important to take note of. So predict or tell us what is a priority that we need to think about. Daniel, I'll start with you if you don't mind. Um I think a priority is to to think of these changes that we are seeing really not from a just technological perspective or tool perspective but really from a change management perspective and I think what kind of exemplifies is some of you might heard you know some people say it's not AI that takes over the jobs it's people using AI but I think the same applies to organizations it's not AI that will destroy an organization it will be other organizations that will use AI better okay and I think that that encapsulates that super thank you
Garrett What would you say is the priority or a prediction? Which would you I would go for the priorities what you can do right now. I would say invest in getting your AI infrastructure AI ready. Second one invest in the upskilling of your workforce to enable that to use those technologies. And the f the last one for the management look at AI as a strategic project and not a tech project only.
Fantastic. Sam I'm kind of disappointed that this wasn't the hottest moment of the entire I know you know conference. Um, I think I've only talked about priorities, so I'll go with a prediction. So, full disclosure, as I said, I'm huge optimist. I'm very excited about what the future holds. However, my prediction would be we will see more failures from a adoption perspective in the near term versus success stories. I think the end result will see more, but in the near term, the next 12 plus months, I think we'll see more failures. And Cler is a great example of that. Last year, Cler got all sorts of press and plaudits about building a customer agent with open AI, disbanding all of their customer service teams. Great promise on, you know, uh, time to resolve, time to serve, consistency of customer experience, lower costs, and what, two, three weeks ago, they said, uh, we got that slightly wrong, and please can some of you come back to work for us as customer support. So, I think we'll see more of that in the short term, but long term, optimist.
Fantastic. And my final prediction in 12 months time I will be around 40. Um gentlemen thank you very very much uh for your time. Um thank you so much. Thank you everyone. Thank you. [Music]