I'd like to talk today about how to operationalize I why is this such a big topic first of all because it's only the five to ten percent let's maybe let's be positive in a positive world here yeah so maybe it's ten percent I think most what I've seen so far and I come a little bit to my life history I have seen a lot of companies working with AI I was a little bit puzzled when Fabian this morning opened the conference and he said hey the I think is only five years old I was doing this already the last eight or nine years at least and I think I'm not in an early adopter when I spend a lot of my time the last couple of years with a big company with three letters IBM i switched to PwC a couple of years ago to have more business focus on the theme of AI and then you know what what I'm interested in it's not only the technical stuff I will come to that but
also what is it all about and how can you make an enterprise successful by using it so the journey between concepts to outcomes is a big big thing for a lot of companies let's try this this way so in the end it's about flying a new technology and concepts to what we have in our existing ecosystem and bringing new ideas from a level that is entrepreneurial prototype ish to something that really makes a difference in the enterprise and answering this question this is why I like your segue and what's not kind of plant is it's really a kind of big effort to do something like that so in the end if you want to start with the concept and want to have an outcome that's not only irrelevant for a couple of nerds data scientists and people working with data is a big issue and the whole thing is in the end very complex so let me start to tell a story a little bit in terms of car of clients I worked with and
I'm mixing the things up so you cannot identify a single company out of that but I think that's a kind of kind of very common way in terms of how these things work most of the ìiî ventures I've seen in the last couple of years were all going in the way that you start with an idea and normally it's one a business guy together with a data scientist sitting together and starting to brainstorm to have an idea that they have taken out there and with all the good things that we are seeing in the internet and what we're seeing what the big internet forums are doing we have an impression that you can solve the real big questions of our time by using this new thing called AI machine learning deep learning you all noted differences so starting with that one as a data scientist in a certain company together with a marketing manager for instance they came up with the idea to think about customer segmentation in a certain market and to build something
like a recommendation engine a couple of you may have heard that so what happens then it all starts with this little thing here in the end it's the data scientist starting together with some input from the business to build a model so let's look at that model a little bit deeper so what are they doing in the end they start with collecting data do some engineering cleaning about the data train the model pray and if they're lucky have some outcomes that you then can put into production in a certain way they luding normally you're using a lot of technologies stacks we're currently deviating more to an open source world some of us are also working commercial tools I don't I come to this a little bit later but in the end that's the process you come up with a result and that's coming back to our data scientists in the marketing and the marketing guy they came up with a new kind of segmentation your recommendation engine they think hey super that is really great
yeah and then next step is they think hey we need to implement that as a process in systems so what are they doing then we have this little model the model weight stands for itself and then they start to approach to take guys in the company IT guys and say hey guys I'd like to operationalize my model and then the start it starts next discussion which kind of platform I'll be using maybe using a cloud service provider Emerson whatever Google Microsoft IBM whatever you need to mention all of them at the moment and then maybe they say hey I'm not operating that kind of platform so what's happening on the other hand we may use tools that are not company standards so we have no clue if these tools are you know under our software licenses and stuff like that are restricted or do we know what these tools are doing to our data we don't know lots of discussion so what happens they're trying to use this model and do a lot of so-called
POCs in terms of proving that this kind of model with the technology they're using works with the existing technology set up in the firm what you get then is the so-called POC hell yeah I've been really through that that I've seen companies producing 20 30 proof-of-concept some of them call them pilots some of them think that it's a minimal Viable Product but it is not it's just something like a proof of concept or technology proof spend hundred thousands of dollars into pocs without having one thing that is really implemented in the enterprise so AI adoption in this case doesn't mean that I I adoption is really creating outcomes it just means you're using the technology and that's just one aspect let's assume that the whole thing went well and these guys have found a way the marketing manager did a scientist started with the tech department we have something now let's work with it and let's try to operationalize that we have now a platform let's work so next thing that comes into place is
the whole thing around data you know this conference is kind of also about data and we have this kind of discussions all over the place so in the end it's about using the right data let's assume the guy has done customer segmentation do we know which kind of data has used to do this kind of segmentation things as a data scientist you know that you know you I think a couple of you are data scientists in the room you know that that you may think sometimes maybe I'm not waiting until I get an official data I said let's try something with and infuse some data external data whatever you know the easiest compliance we are talking about it mainly here in this country GDP are everybody knows about it don't use personal data that is much more complex a lot of data sets can use can be used as soon as you're crossing data set with others even the legal aspects on compliance change a lot next thing is about data quality and data sources
data quality is a main issue in the industry since 15 years 20 years we're talking about data quality of the main source for all analytical problems and if you just take that as an in as a kind of idea and the problems behind ain't equality if you have not solved it so far the sources in terms of how can I adopt a external data how can I buy for it how can I get a license for these kind of things it's also something that is really driving a lot of energy around these kind of data discussions having the data will give you definitely the way to better outcomes but also understanding the data and having the whole thing together is still a major problem in the enterprise and you all know that because in the end the biggest problems I've seen so far is to provide the right data sources to the data projects and AI projects in the enterprise's I've seen in my life multi-million projects that have not been successful in the end
only because the data that everybody assumed the somewhere existent in a certain enterprise company was not been able to discover or presented or to be used in a way that the AI systems could work with it so if you extend the model even more then we come to I think the most important pieces beside technology and data it's all about people people who can work with AI are not only the people that are the AI specialists and we have heard it I think in several kind of talks today in terms of understanding AI is - first I would say the first kind of foundation of really being clear in terms of what can I do with AI and like how can i anchor it in an organization so that means for everything of these kind of projects and I coming back to my data scientists in my marketing manager in a lot of cases there is some someone either a sponsor or something like that missing who understands what is all about let me extend
this story here a little bit we have done our customer segmentation recommendation engine and you know I think the marketing manager was kind of happy because he was assuming that he can have a better something like that 10 6 10% sales increase in certain products in a certain market if he apply certain ideas he got out of the model the problem then was that his boss the guy who was asking for the budget in this kind of cases and was responsible for the budget promised something like a kind of miracle to his board with the great things a I can do and 10 percent to be honest was not enough so I was pushing and pushing and pushing and pushing to have something like these miracles that we always see in Silicon Valley you know I always think how impressed I was when I cutting is three or four years ago some data scientists predicted the Oscars by using just like open source data from from from the internet and could predict who would period
that academia Award winners something like that is in the mind of a lot of people in upper management in terms of how you use AI and some of these things in the end lead to the point that sponsors may overestimate what you can do and also have an kind of not real expectation for the outcome of AI so that means it this guy is not happy there may not be the next project skills is also a thing I'd like to talk about in roles as well you may know that I don't want to know what I don't want to ask at the moment who is with the data scientists room but in the end I think we heard that in the last I would say four to five years on a regular basis coolest job on earth sexiest job on earth I tend to agree partly it may be the sexiest job in earth but without this guy he has no chance to integrate anything into an ecosystem an enterprise organ organizations now indeed in
the end the data engineer may be the most important drop on earth and you know I'm also leading a team so I'm hiring people for PricewaterhouseCoopers and we're looking for data scientists looking for data engineers at the moment my ratio is that I get 20 applications for data scientists to work with us which I'm happy about and if you have a certain ambition please contact me on that and out of the 28 kind of applications I'm getting one for a data engineer and if you see what we want to do with AI to implement it deeply into our ecosystems it's so much it's so much more important at the moment not to have only the ideas and having the models but also to make them real and to make them operational in the companies we're working in or the organizations we're working in so you need this data engineers and I think to be honest these are not the superstars because they don't have this kind of epiphany moments where you find a great kind
of model that is solving something nobody has thought about before these are the poor guys we have to arrange the technology the data and all the stuff and need to bring things together and it may be a little bit more boring way so that's a that's a big one as well adding the next thing is culture culture and a lot of enterprises I'm coming we talk about Germany a lot here but I think that applies a lot of kind of European countries as well culture is something that is not only giving us the ability to think about how to use technology in a certain way it's also kind of the basic and foundation to to cope with this whole change that we were in if you think that 50% or more of the jobs that we are seeing in the in our in our society at the moment will be somehow infected by AI and we have heard that all over the place it's 50% the other ones at 45 I don't care it's almost
every second drop description will be affected by by artificial intelligence if you don't have a culture that is currently embracing that and supporting that in your organization you may have a problem and a couple of these thing I'd like to go a little bit deeper first of all the whole thing around being data-driven I have the point as always in terms of if I look at Germany that's my little cube being in data for the last 20 years is looking at Germany and the European market is most of the companies I'm working with and I'm talking brick and mortar I'm not talking about the new internet startups and so on and so forth they all think about process first so it's all about how can I build about the process improve a process whatever building us ap system or something like that or ERP system to be mutual here and and and and just do something to make my organization more optimized on my processes isn't this a failure in thinking don't we need to
talk and to think about enterprises that are based on assets assets mean I think it was mentioned in another speech beforehand it's the capital of the enterprise what is the capital of an enterprise its production facilities it's necessarily the people around it it's maybe also the financial capital you have depending on your business model and it's if you really want to embrace AI its data it's an asset and all these things together may work together through a process and the process may produce data but the process is just something that connects all these things to make something out of it if we think about the classical thinking of it enterprise in most of the brick-and-mortar industries they think about data just being a product out of the process and not the other way around and I think that is something I think just think about it if you don't have that thinking in your enterprise it's really hard to embrace the new technology because that's all the basics that's the fuel we're all working with
is the data where we can place our AI models and our apply AI applications on next thing and I think you know that as well it's also something that's pretty rare is the acceptation you need to accept failure not only as a single person as a group as a team if you try out coming back to my story about the data scientists the marketing manager if you don't have to write results in the first place you may need to reconsider the approach and try it again you may need to reconsider that again that's not part of the DNA of most of the companies we're working in we have a lot of CEOs attending astreus we accept failure one time yeah that's reality and if you think about it from in terms of how do you make these kind of things successful and I think about remember entrepreneurial spirits that we see in other countries lagging me like in in the Silicon Valley which is all about that you know how many startups have failed there
there's only one Facebook there's only one Google there's no Yahoo anymore yeah so that sees also there is failure is part of the process and if you think about that also in a small way in terms of just getting applying ai to one of the single pro of one or the other process accepts a accept of failure ISM is one of the most important cultural kind of factors to make AI wheel in the enterprise the last thing I like to talk about is governance Y governance if you start AI projects I think it's important to select ideas wisely I said yes we need to figure kind of culture as well but also the I did the point around that one is it's never I've not seen one company I've working it was where we'd elect ideas you know seeing all these kind of gems that are happening in a lot of enterprises where ideas are gathers and his idea of competitions and stuff like that that's all important super important you need to harvest these
ideas but you will soon leave soon find out if you look in your organization that there are enough ideas the problem is to come from the idea to reality and this means also that you have to select ideas that you want to drive further by applying something like a metric in the beginning you need to think about the value I'm not talking arrow eye here well you can be everything it doesn't need to be a commercial value doesn't even a financial value but you Newton you need to define what is the value you want to get out of it and if you define the value out of an idea and you want what you want to achieve you can measure your idea if you after you have implemented and you look at the outcome and to think about it in terms of has it deliver the value and I think that process in terms of governance is so important because we are sometimes missing the point in terms of also taking the stakeholders with us
and to make sure why we are doing this and sometimes we're doing this kind of AI stuff I would say we're trying it at least in the some of the experiments I've seen just because it's cool yeah and just because it's cool maybe not enough so in the end it's about the value case in the beginning it's about working about ideation it's about the measurement of the outcomes so you really have something that really goes from concept to outcomes I call this thing kind of the AI molecule that is something I put together because I think molecules are also complex and you know ad reality and if you're really working in this space you will see all these things affect you and if you're not taking care all of the all of these balls in parallel and get these things in there's a high likelihood that the eye eye initiative to drive AI in an enterprise may fail so what's the conclusion in terms of coming from concepts to outcome is something I think it's
feasible we're seeing it we see more and more success but in enterprise has to live it it's not something you can just do by having this poor data scientists together with poor marketing manager and then some around and then rescue my company with a I so learnings from my perspective that I like to to hand over to you is first of all we really see organizations massively failing to adopt AI for the reasons I've just shown think holistically it's not a tech game it's not only culture game it's not only people gave me sort of a dinner game it's not only take game it's not only governance it's the combination of all of it and if you think holistically there's a good chance that you move your organization third thing everybody needs to be involved if we think that AI may be successful in adoption in enterprises organization we need to be aware that the culture and the knowledge about artificial intelligence starting with the definition what it is is an important factor and it's
one of the key factors to drive the successor so everybody needs to be involved with involved I mean really involved more as a pig than a chicken so it's complex and it's also something I would say we're all tackling I wish you in your organization's all the success with these kind of things because I think we needed to take also the opportunity to compete with the other parts of the world as we have seen the main stage to drive artificial intelligence to success in Europe and in Germany thank you for your attention you