how to tackle AI and if you sit down and and just scribble down a couple of things where are we right if you imagine to be a dodgy telecom we actually have tons of repetitive processes across the globe we have approximately 160 million mobile phone customers alone and you can imagine that that creates a number of repetitive things that are to be done with those customers and internally so we really have a lot to to autumn eyes if you like then secondly all those customers and all the processes the internal processes create tons of data we are not Google and we are not Facebook we have different kinds of data partly but we do have a lot of data so a lot of stuff to have AI work on if you find the right place and then obviously this is why we're here at rise of AI it's the right time for any organization to start looking deeply into AI for a number of reasons which have been quoted back and forth I don't have
to have to repeat that right and there are oh my my little earmark then secondly there is a lot of enthusiasm on AI tic but there's also a lot of doubt and a lot of fear and that is across the country but also within the organization's obviously because people fear that jobs are being automated away so there's a very strange mixture of feelings towards AI and in that situation if you are convinced coming back to the air to to the Y that you want to work with AI to improve your organization what is your tactics what is your strategy to go into the organization and there could be different games you could play number one could be the classic big surprise game you you could like build up AI skunkworks somewhere develop something until you think that's the right product the right process support and then go to your Coggins agent say haha hello we've got something new how do you like it and that might not always work perfectly so we didn't choose that
path second one would be grassroot approach like you you just let AI happen somehow with the scattered and few and far between resources a normal organization has in terms of deep AI knowledge so you hope that people just grow by themselves because it's a hype topic we've feel you stretch out existing resources too far if you do that and nothing really will happen there is no no focus a third approach would be the classic big organization brute force from a buff approach so the CEO is saying come on everybody now has to do something with AI across the organization there are some interesting examples with interesting players like for instance Amazon I understand the process at some point says each new system has an F API so we can plug into the data that's a good approach but for the use of AI in an organization like ours I don't think that works just having our CEO say look wherever you want to use AI wherever you feel you can use AI please do it
why is that because in each and every corner completely unguided people would feel pressure to do it even if it's not useful in that spot or if we again would use our resources in the wrong spot what we feel is the best approach and you might have guessed because in the title of my talk is the trust game have enough resources in one spot and then create trust across the organization that it's right to work with you and right to work on AI right so the trust game this is what I will focus on the next few minutes how do you generate trust in organization with as I said earlier where we are with a lot of enthusiasm a lot of fear a lot pipe and a lot of misunderstanding what a I really is and can do how do you create trust for what you are doing in AI and such an organization so this is what we need number one is rigorous focus sounds easy mundane always said but what we have done
as we've started more than two and a half years ago to work on AI and we have more or less exclusively worked on building BOTS may many text BOTS but also BOTS you can talk to which are not yet out on the market but we work on it but we we have completely focus on it with quite a not quite a lot of resource and a number of people I met and the company said oh now you're building that AI grade they can yeah you can do this and that and that and that and that was it if you've got that AI and they thought of a general artificial intelligence kind of brain because they didn't know enough about it and thought it could do anything and kind of had that expectation that we would go into all of their problem fields and we always have to had to say no we were close to going into new directions already you want to have years ago and we said no no no we don't do
it we just keep that focus which was pretty pretty helpful for us because that was a key driver because we could actually show something then by focusing that created trust number two trust through fast delivery to a point where we actually when we started to an hour of years back we took a lot of time better part of a year even more than a year from some perspectives working in this field of BOTS without ever touching an operative assistance AI right because we wanted to be really fast in the market with this new dialogue logic how to interact with people and there was a lot to be gained by actually learning how to use dialogue systems that are rule-based and we had a bot out on the market we iterated in front of customers all the time fast iterations and learned a lot in doing so and obviously while we did that we also scouted the market intensively endured a lot of pocs with with AI but it was more than a year later that
we actually introduced the first relay I salute in front of customers so we went that fast to deliver and we went that that that far to actually deliver fast in front of customers iterating with you both users also that had a lot to do with the trust we gained internally because people really saw we just do things and don't create moonshots right I'll think about moonshots so small victories small and fast victories and not going for the AI moonshots that many people in a large organization might want you to do if you go into AI because there is so much talk about it and then thirdly and maybe that's even the most important thing from day one even from second one this was not an R&D project done somewhere but we worked closely as one team and one project with the business side so I'm working in R&D no more D than R and that was the success ingredient for us one team booth for instance in our case we started in Austria with our
Austrian subsidiary with the e business team down there who really wanted it we saw the glowing in their eyes and that persuaded us these are our partners we can really have success with right and and that was really not only at the start but all through a journey with new customers who onboard every time from day one we work with them okay so these are the key three drivers we felt really helped us I feel that that really helped us to be successful on that journey to build trust and then there's one more thing we hardly we almost forgot trust with our customers and in a larger sense with society in this field of fear and doubt which are also encircling AI a little bit and and just I I think one or two weeks back Deutsche Telekom has published an AI guidelines so a number of guidelines how we want to work with AI that we feel should go a long way in generating trust in us and how we deal with that topic
obviously that includes data towards our customers and and society at large I'm not going into details you find them online they are on our on our corporate website and I think they are great they really tell you a lot about the way we work ok and now for the few remaining minutes I've got a counter here just so that's great seven more minutes a little bit about the world what are we actually doing so I said we build BOTS or have been building BOTS until end of last year that was our exclusive focus part of it looks like this we started here that is what it what I told you about that was the rule based bot we've launched that actually had been launched and we started the project in a pretty similar simple fashion in Austria we improved on it in a number of dimensions we introduced new media formats right in the chat flow we've integrated a hand over to a live agent if the bot comes to the end of its knowledge
by the way it's called tinker the bottom and that's why you see it up there so you within the chat flow it you can switch from the from the system to a live agent just continue seamlessly your chat it's got backend integration and so forth so it's far beyond a simple but you you you built in a couple of weeks but it was still on a rule based on on a rule based system and then we had our new AI system which shows our external partner we work with and then we thought do we now wait until we have feature parity for that system and then are courageous enough to switch to the new AI system no and that was again matter of trust on internal partner side we decided to launch it very early on far Biff when we didn't have even been even near to feature parity to the existing bot and what we did was we we let tinker introduce the new board first thing you saw us so you open our
rule base button it says hey do you want to meet the new tinker it's early it's early on but you can try it out now so we got feedback for our new bot early on that was a bit more than a year ago and now just approximately four weeks ago we switched completely to the AI bots so now we've got a fully advanced and scalable system in front of customers it looks looks like this in Austria and with a little bit of delay but they caught up quite nicely we did more or less the same for our largest subsidiary or largest which is taja dodge telecom Germany and this is what you see here on the right so we have a pretty similar chat pod out here all of these chat bots are for customer service purely right so that's again matter of focus we decided to start very much focused on customer service issues obviously it was a starting point that also has tons of potential because in customer service there's a lot to
be gained if you automate away the highly repetitive questions and and dialogues that happen each and every day in our call centres so that's if you like the business case obviously behind it we are doing a lot more we are as I said we are also working on on on voice BOTS that you can call and talk to we are building systems that are not customer facing but agent facing so our agents have more immediate and more clever access to information when you are on the line with them and answer more stuff so that was our main focus not the title of the talk was not only trust but also scaling and all the trust we've built internally also enable us now to go beyond BOTS and this is what we've started earlier this year and it looks like this everything I've told you what we've done in the what so far we call smart dialogues and if this is a chain of IT systems here then we've been focusing on the talking front end
all right and obviously we continue on that work the journey is not yet done so there's a lot more to do a lot more back and integration into systems a lot more dialogue breath and so forth but we now go into two new fields one field is these systems these IT systems if there's a process running across the IT systems for instance the bot is telling you something if we go into sales and you wanted to be implemented more automatically than it's done today what is happening is that you have manual breaks between the systems so an effect an operator is reading screen a of system a and typing into screen B of system B not all across the system but that happens in all kinds of companies that has created a new industry of its own which is called robotic process automation many of you know and robotic process automation is also a rule-based world so you have simple algorithms that weed one screen and decide what to click on the next screen and
we feel that there is a world behind or beyond robotic process automation driven by AI and this is exactly what we are looking at sorry what we're looking at here so automating away these manual breaks between individual IT systems beyond classic robotic process automation and the third bit two more minutes the third bit here is what we call smart decisioning so that is really going deep into one system look at the data and derive using machine learning better faster more efficient analytics of the data and actually then also prescriptions or actually action advice what to do based on the data in the number of fields and what we find so far that there is tons of finance use cases for instance that really can help a company a lot of use cases we are finding we actually have a very long backlog of stuff we would love to do if we had my more capacity for it so this is our journey we just started actually many people have told me we started a little
bit at the tough end with natural language and we are now going into the into the easier fields not so much true here this is pretty complicated too because in a cooperative pretty complicated systems and system landscapes legacy systems old systems new systems that speaker don't speak to each other so that that is complicated this is nimbler you can find smaller use cases and can be pretty fast here so already in the first month of the year we've been pretty successful in this field with actual output that's it and we have half a minute of questions for questions if you like other any yes well beyond the three fields so that is already an expansion that is ongoing and we will most likely continue to work in all three fields there are new ideas new phases within these three blocks there could be other blocks which we are thinking about now not yet done so I couldn't really tell you but we see a lot of potential to continue the journey yes as I said
in the beginning so we're coming back to the why there is a lot to do other questions that's a clear no so we're in time thank you very much