 Hello everyone first of all, thank you for being here. I know it's a tough competition of talks at the same time a second day Almost three more talks to go. So you almost did it. Thank you for being here Thank you for the introduction on first of all I will try to explain why we are here and why we are going to be speaking about this Carlos on myself with we work at Microsoft and Microsoft is a big company We work especially in the part that we do the projects to adopt technology a Microsoft technology mainly Almost all the time and how to to get this technology to the final clients Okay, so we are going to be speaking about our experiences and what we have learned that the scale Means to all these clients that we have met. Okay, and the idea is you got a lot of information during these three days Do two days? ML flows machine learning AI we are not trying to throw much more information to you But for you to start asking some questions So ideally the target of this short presentation is for you to end with the right questions to ask When this is finished you go back home and you say how I'm going to apply all these things that I've learned There is a those two days, okay So with that goal in mind the first question and it's a really tricky question is what is scale? Okay, why are you here because this II at the scale? What are you expecting here? Are you expecting a talk about compute power? Are you expecting a talk about how to put more models into production because this is that there is a different answer Every time you ask this question for some companies a scale is having machine learning models and production every two weeks Really ambitious know something like that for their company scale is compute power For their company Maybe a scale is that the company the whole company start understanding what is AI so Having some awareness adoption of AI it could be also because someone said today and I think it was back back on a thumb Like said you cannot compete having more data scientists than Microsoft Google But you have more business analysts than any of us. So how can you transform all these? People that you have in your company to cities in the data scientists people that can use this tool to create some AI Or maybe for you scale is as I said yet people in your company believe in That AI is something possible. Okay, so in this talk, we are going to first Carlos is going to explain time to explain What is a scale for Microsoft is one of the things and then in the last part I will try to give you a framework a model of how to think about the scale for go back to your company Or to your startup or to your daily job and think how I'm going to adapt these technologies to my Company to really really scale. So that's the goal for this 30 minutes left. Okay. Yeah Well indeed I want to just to start this is the a brief summary of the ignite announcement from Randall last week. It's about services that we have delivered to our customers and partners and as part of it was Mentioning I wanted to share with you that notion of scale at Microsoft at of course 10,000 food kids with our Customers and at Microsoft. We talk about a tech intensity. That's a detonation Sanding junior and I like I love and you will see in a second. Why? We talk about the tech intensity. This is a straightforward formula where Women with their first part first thing that you are seeing there is how quickly can you adopt new technology in your Company and your organization that means of course Every new time you start Adopting new technology don't build and really recreate the wheel And it did this part of the work that the product groups at Microsoft work. They build their own product based on Our services and as well an open source solutions and the second part it's a critical and Very interesting for this conversation is about the tech capability. How are you able to? build your own Solutions internally how you are able to use that technology and deliver What really matters to your organization? So for example Which are your the best thing to do which is the most? proud of solution you are of Which is the best thing that you do that really differentiate you from from your competitors that's your take up ability and The trust part is all Really a foundation for us and it's the parallel all day about trust when we call trust We are talking about security. We are talking about Privacy we are talking about compliance and every service that we Built and we deliver is based on that trust and this is not just for us But as well for you you as creators of technology at the end Needs to see that as a key foundation of your solution as you take up abilities But thinking about how take intensity helps and As you may see, this is really extended no matter a AI solutions This is for any type of solution, but here specifically for data and AI solutions Having a low take intensity means that There's a gap. There's a gap between The information that you have the information about your systems the data coming from your customer your systems and how You're able to use Some of your technology so if you have low tech intensity, you don't have maybe the best technology to To run for that use case and that Resonate with the gap that the world I was mentioning so that you will have a gap into the time That that arrives and the insights that you get After working on that data so here that is if I I'm a high Tech intensity. I will reduce the time to be able to gather insights and as well be able to reduce the gap About being able to get insights before we have that data and that Gavive what's we call when we mean take intensity to pull up that Red line to the green lines and you are able to adapt quickly and faster to your Solutions or your use case So at the end when we are Targeting that tech intensity Of course, it needs to be linked to our mission in this case of your mission of your Organization in this case our mission is simply put to enable you To create with our technology at the end is to be able to Empower every organization and people on the planet to chip more and that means that to be able to use that technology to become a digital company and Use that technology to change and turn your organization into a digital company because at the end what you envision is everyone to be really a software company and From those things the challenge here the challenge about the scale and the opportunity that we have is how to be able to Engage and work with a hundred thousand plus employees with 75 million organization and partners in the world to Impact and empower the seven billion people on the planet. That's our key mantra. So When we are able to Empower all the organizations to change and address to change their lives and address their dreams and This needs to be rooted Of course In your values This is an example in the AI space, but it really matters for anything that you may do we Started three four years ago to set up to boot up those six principles in order to Think about everything that you do under a trust mindset We talk about the furnace privacy transparency So that everyone in the organization think and work in with the same principles guiding the and building the AI solutions through that vision and presentables and then we move Into practice because you may know the three four years ago We we deliver a tie bot that was a chatbot deliver that we learned a lot from that We From principle to practice begins to be able to address those principles in a practical way inside the organization the first was internally and if there's an AI and ethics Committee for engineering and research so that they work for sales and services to be able to understand and and design solution that address and follows those principles and outsides which started and There was a lot of traction with the partnership on AI.org that you can visit There are many companies in the tech space, but profit or not profit like Apple Google Facebook many many others to be able to work and Ecosystem through looking through those principles and of course thinking we're mentioning the Empower every people and every organization thinking about how the human AI design should be built and then moving into tools so that you are able to Implement and we are working and delivering sharing with ecosystem tools for example to detect bias or to be able more to be more transparent in the Models or the information that we deliver or we manage through through the services So Here Start what it was? was talking about This is about you so think about questions and what do you think after all the those today's and Yes scale that what are your trade-offs? there are many a Specific attributes based on your mission or your principles on your values that needs to be addressed and Balanced to understand maybe I shouldn't do that project because it does not fit in in in my into my trade-offs and of course being able to understand How should I build that solutions? so I'll hand over to Pablo and See what I've been working with the customers So what happens here is that once you are fully energized and say I'm going to do all these projects I'm going to apply all these things that I learned what coming next Welcome to the disappointment valley Where you are doing a lot of proof of concept or do you are doing a lot of things? but you don't get the real real result real business value and Be honest we are much more better than we were a few years ago But if you speak to the companies out there and it's not everything as pretty at this look sometime I mean, this is the valley that we were speaking about you have a proof of concept you have everyone aligning the business We are going to invest on this after some time Where is the value? How how are we going into production with this? How how how are I am we measuring that this is really working? This is a really good report From last year and one of the sentences that is saying there is that only 4% of the companies are saying that they feel that they are Getting some value out of AI, but 71% of the companies they say yes. Yes. We want to do it Okay, so that's that's that's the mismatch in between I want to do it And I'm really getting that value and this one from also You can see in many of the companies in the world But it's not only in Europe because you might think this is because we are Europe We are different this again another report and I know you're tired of seeing reports But this is also a good gift of what is happening around the world is for example in the states One of the things they say there are many pains but one of the main pains is I'm struggling to move AI initiatives into production Okay, and again just the last one. This is Paco Nathan. I was speaking about this as well The challenges in AI adoption in the companies Three main things culture data skilled people. Okay, we are struggling to go into production That's the reality. They are really good results. They are really good proof of concept They are really good things happening But the challenge is trying to get it as a scale and as I said scale can mean different things from different companies You need to think what is a scale for you, but the goal is the challenges to go to a scale and What is this happening? Why are we crossing this valley of disappointment? And because basically we are looking at this problem from four different lenses The the business that is probably the most important one is the one that is telling us You need to do this. This is where the value is we need to bet on this Initiative, okay, but then we have the tea or that are the scientists the second column it happens that sometimes I'm just fighting to get the best model and I'm speaking about this a lot of time and even with the team the other day in a machine learning project We were where we were working for a few weeks and the results were where we were able to recognize some entities in some documents That were old PDFs and suddenly we were able to recognize who was a notary who was a liar And we spent two weeks doing that when we went to the client. They said yeah for me search was enough We didn't need all the things that you created and for me was like Yes, sometimes we are just trying to get the best model the best technology doing this and we don't think about what the business really requested I know that seems simple, but stop and think about it We don't want to end up being doing waterfall projects again in that assignment. We want to be a child No, that's that's the thing and so and the third the risk people they say no, this is dangerous This I don't know this I don't want to move the data to the cloud. This is dangerous Most of the times they think that this is dangerous because they don't know what it is about and of course the architects Everyone has an architect inside of his heart. Let's try to create the best platform possible Let's try to create something that is going to work for all the projects that we have now and in the future The so you need to give all these people what they want So you need to create a team where all these people are involved from the beginning and that's that's the key that's the key to go to scale and From our point of view this translates into three x's you need to think about the people part You need to think about the technology part and you need to think about the process part What does he mean to think about the people but you need to think about how to create how to Get the people involved into creating new AI initiatives So they need to believe that this is possible. You need also to have the technology But you know that the technology is there You know that DevOps is really important the MLOps is the key You know those things already But you need to be able to Scatter than that knowledge, you know in our around the company and of course the process the part of thing I want to do this with less effort every time. I want to do this in an automatic way Someone told me once that a data science project is like putting a rocket in the moon I said well not always but the goal is putting a rocket in the moon every two weeks and one after another One after another that's what we want to get that's the scale being able to do that in a way that is automated and is working Fine properly or almost Perfect, and it's not important how you call this you can call it center of excellence You can call it AI factory. You can call it control tower of AI It's not important the name, but you need to have something that put all these people together. Okay, and This is just an example. This is a framework for you I mean, this is something for you to think about I'm not saying that this is the solution But you need to check if you are checking all these boxes Imagine in the people part and can you see that? Well, yes, you have continuous innovation You have AI innovation you have to be able to in your company Everyone has to be able to create a come up with ideas or what they want to implement Of course, you have the DevOps for AI and data governance. There's been tons of Talks during these two days about these two things. These are important I'm really glad because actually three years ago. We were not speaking about DevOps for AI Three years ago asking wanted a scientist. Are you able to tell me which data have you used for this? Experiment and exactly reproduce that and all those things and retrain was like a I'm still not there Now we know that that is possible But that's not the only thing what we are saying is you need to put all these things together And of course the last part you need to build as Carlos said a platform that everyone trusts You need to create a community of people that believe that this is possible Okay, and all these parts are important and in the beginning, of course, you are going to need some help This is not possible to do it from yourself from scratch But the end you will get the velocity at the end if you check all these boxes You are going to be able to put a rocket in the moon every two weeks straight What happens if you don't do it? What happened? That's a question. This is like testing for wrong What happened if you remove the continuous innovation part of this equation as again? Think of this about like a framework for you to check to think about if I have all these boxes If you remove the continuous innovation, there is no alignment. You will end up doing projects That are not relevant for the business. You will end up doing maybe a natural language processing algorithm that For what? Okay What happened if you remove the AI driven culture? Maybe there is no adoption No one believes that you are building or no one uses that is even worse what you are building What happens if you remove the box? Of course, that's Disaster, you are not able to retrain. Okay, you know you are not in a good position and you remove the box from the equation What happens if you remove data governance? No one trust the data. There is no quality the whole thing doesn't doesn't hold What happened if you remove the center of excellence? You can do that, but you are going to reinvent the wheel time after time. You are going to build the same project End times and you're not going to reduce the knowledge and what happened if you don't have a strong AI community What happens is that you don't scale and again you need to think here What is the scale for you and I'm not saying that these are the boxes? I'm not saying that I'm saying that you need to think where you are and in which in this axis where you want to be Okay, and to move to yet another example Let's imagine that you have initiate stabilize optimize and scale maturity levels Okay, and if for each of these streams, you need to take some little steps to move to the next level Okay, for example in the first level continuous innovation one thing is doing right right doing right things That's correct, but doing things right is the next level Okay, and for you my my question will be try to identify yourself where you are in there Try to identify if you are doing things in all these streams And then think what is the next thing that you need to do to move forward to the next level? Ideally we will want to be in the scale phase, but again, then is the question what is the scale for you a scale for you is Having all the business analysts in your company working as data scientists So maybe for you the most important part is the community What is the scale for you a scale for you is being able to retrain models? Automatically then maybe DevOps is the most important part for you The maybe the company doesn't believe that AI is there or machine learning is the important thing for the future of the company Maybe you need to work on the innovation part So the thing here is try to create your own road map Try to find the boxes and create something similar to this and the first part that you need to do is identify where you are And in this stage you have seen too many maturity levels I mean this thing of where you in the first stage foundational approaching aspirational I'm sure you are tired of seeing all these kind of graphics, but for me there isn't one important thing here How do you know that you are in the in the highest maturity level possible for us? This is visible AI. This is where AI is not in the laboratory AI is not something that you are doing in IT AI is everywhere is visible is tangible is in your application is in your reports is everywhere so You can use these Triggers you can know you can use these pains to identify where you are But at the end this is just yet another maturity level that probably you have seen many many of this Let me drink some water And because I knew that you are not going to remember this Then I we I borrow a metaphor from one of our colleagues. Okay, and probably this you will remember Let's think about the age of AI What is this? It's the same but with some cartoons Imagine that you are in the stone age People haven't still haven't discovered the fire. They don't know what AI can can can do can what is possible with AI What is the bronze age? This is where you are creating tools. No, you're starting to use DevOps. You're starting to use something you are Using tools to create things What is iron age? Maybe you have a battle of houses data against IT you have some other Things that are coming into into into the equation then but you are still not using the full power What is the renaissance the renaissance is where finally you manage to merge all these things You have IT people working with UX by people working with business people or all of them are enriching the others and The final final stage. What is the final stage when you are? Industrialized when you have a factory of AI you have a process when idea to production not idea to MVP idea to proof of concept To idea to production you have a streamline. Okay, I think this is much more Easier to remember know that the maturity levels that we are used to see but for me again This is the same question try to think where you are and what you need to move to the next level Have you discovered the fire? I don't know. Maybe there are some companies that said and we are going to put as an example imagine that your company is a 20,000 people company and They don't believe that they are something that is for them that is too complicated is it's too far Okay, how do you do you help them to discover the fire and here is an example? But just an example They are imagine that you have an application like this and this is called power apps and it has something that is called AI builder That this helps anyone to create an application where they can recognize objects Okay, I'm showing this is just an example But you need to think about your own example if you want to convince your people that AI is something real Just scale for you imagine that you say a scale for me is convincing 100,000 employees that AI is possible Think about something that you need to do in the area and for example This is an example get an application that is easy for them to use easy for them to build and they will start thinking Okay, this is possible. Okay, I have that just to talk about this part Think about the citizen data science When we talk about Citizenship at the end is about to a scale to the all the employees in the organization so imagine and sometimes indeed you have the The capability to deliver that to your end users But maybe your tech capabilities they have you appeal you can think about how to deliver to all the employees I know that not just thinking in one people but thinking about 10,000 all the employees in the organization and and this is about that about democratizing in the organization the tech capabilities you have done or even the tech adoption for others platform other Providers Carlo jumping because he knows the next slide is he's so you return to the next one this Well, just for examples that we were certain. This is a Solution for doctors and Hospitals, this is a This is called inner eye and is Solution that they build based on our platform to be able to reduce the time to train algorithms to detect in this case cancer The point here is that doctor instead of spending two three four times a In the daily activities Reviewing the images that are big to analyze complex to position so We're able to use that capabilities in a simple and and productive form Think about inner eye view Look it in that space Or for example, and indeed I think that it's just today was mentioned and the reps all we were working With the results at the end is how to scale In the organization the awareness and the capabilities that you are able to Do to the internally so at the end this focus on the awareness and drive in all Your organization that knowledge and that solutions that you have built internally and this is one of the I like the most you can search For it as as wild me. This is a nonprofit organization that Many years ago you could say this is a crowd-sorting a scenario, but what they push is gathering information from Social networks Analyze that information to be able to understand and track specific not yet species, but a specific animals Around the world so be able to understand which are the immigration paths And even where are the species around the world and they are able to develop to deploy Solutions and model for everyone and the tech for some that that whale is called Margarita and it's different from others. So to understand where she is and and how to track that Okay, so at the end what we were trying to get here is for you the question that what is a scale for you? Maybe a scale for you is to change the world and find the whales and and help the The sea to give you a better place. I don't know but if you're in your company first of all stop wasting money in proof of concepts and MVPs, okay Let's stop doing that. Let's think about Big production going to production as I say for example an ambition can be I want to have a new in artificial intelligence Case in production every four weeks. That's where I want to go. That's my scale Let there's you need to have all the variables You need to look everywhere in your company to get there, but it's possible. That's where we want to go these days So build capabilities build call it center of excellence control tower AI factory the AI smart people That's not important It was the name but create something that is able to help and and help people to be more agile and help people to Understand how to get there and again this and I think everyone said this during these two days this at the end Technology we are getting there Technology more or less is solve every day is easier to use is about the people It's about how you convince people To use these tools and to use these tools for something that is important and relevant for the business And again, how do you how you do that in a way? That is not the biggest effort every time you want to put something into production. Okay for me That's are the three main takeaways For you to think how to apply this and we we want to end with a video It's a great place to put a video with the sound and the screen So also for you to think what are you going to do next with these technologies? Today right now you have more power at your fingertips than entire generations that came before you Think about that. That's what technology really is It's possibility. It's adaptability. It's capability, but in the end. It's only a tool What's a hammer without a person who swings it? It's not about what technology can do It's about what you can do with it. You're the voice and it's the microphone When you're the artist, it's the paintbrush We are living in the future. We always dreamed of we have mixed reality that changes how we see the world and AI Empowering us to change the world we see you have more power at your fingertips than entire Generations that came before you. So here's the question. What will you do with it? Thank you very much. If you want more information and more technical information There is another talk from our colleagues in 410 a power to multi-data quality and data lakes I really recommend you to go there if you want to know more about the technology now We still have few minutes for questions in case you have any questions not very difficult. We are here to answer those Thank you very much Rock question a microphone or the one thing you answer this I Have my opinion, but first I want to listen to Carlos That's great. Well, you were mentioning Tools and services that we have right now. For example, one of the things that we have derivatives after machine studio As we're machine learning Services that's to reduce indeed that complexity. That's something that they will go over time that's to reduce the end-to-end in the sense of from MVP to just get out, but that's a product that we were talking is more involved as a technology There's something that we are working on is more for transparent Modeling the sense of time to understand what's the model model that you are building and how that affects and it is moving This to everyone. So for example, there are power bi that you were you may know with AI builder Yes, to reduce the complexity to be able to everyone to deploy models, but in a business environment As an I so that's thinking something that it's moving as well And the last thing that they would say it's more of what we call inclusive design That's when you may start designing for just one people. That's your ideal persona and Then the things that we we are working on is how to build for seven billion people. So sometimes you just not really think about Instead of disability Thinking about mismatch with your environment. So at the end we do we talk and that for example about Detecting bias in your models about the AI but as well for any design design mindset that you have When building your solutions, so you're thinking about how to scale and that part but for any type of people That's a good answer. Nothing to add Okay, great I wouldn't take this one and for me one of the biggest challenges to get Everyone in line for example the data of the people were are able to give you the data and having a paper line a pipeline that is working To put the same data into production and the test environment and the deaf environment IT people security people so having everyone online is for me is the Most challenging part having everyone looking into the same place because it's sometimes you have the models and that's great But you have the data so it's like having pipes with a water and you put those into production But there is no good water going into that So I might sound simple but getting the data into the models that is the real data and having that in an automated way It's a big challenge. So it's you need a cross team to be able to do that Okay, I will think it about Iot solution that you sometimes you need to gather data is Usually you start off with data and juice and many people said well, how can I What can I ask with that data and maybe you are not able to answer any question because Usually you need to put the question first and understand you have enough data So sometimes in the Iot space when do you need to gather information? Sometimes you start thinking about the model and be able to deploy but you need to review that Your production data matches the the use cases and the business question that you have so many Many complexities or challenges that yes, sir in that one because I think that's that's pretty common as well And I give you the last key if your data scientist is doing a lot of clicks to go into production That's a bad sign there you he has to be coding So that's another indicator of how complicated is going to be to put models into production if you are using a mouse It's going to be complex if you are using some code is going to be replicable And it's going to be automatable with some DevOps or whatever. So think that take that in mind So if someone is making too many clicks, maybe for going to production, we should be writing code One half time from our one more question. I guess if any minute of potato Okay, thank you very much