Top AI tools: Boosting productivity with GitHub Co-pilot and Azure AI

hello everybody and welcome to the Thursday afternoon webinar on top AI tools for devops and software development we're going to kick things off in episode one with boosting productivity with GitHub co-pilot and azour Ai and today the presenters will be me Maro and hanu with me I will start by going a bit through the the AI and the topic about Ai and and how organizations are changing and then we'll jump right into GitHub co-pilot features and how it changes kind of the developer flow and then after Mian we'll be talking more about azour AI once again welcome everybody as we're speaking in Zoom you will find a poll it's two questions so not very laborious you'll go ahead you walk to the poll and you'll answer the two questions and then in the end we'll see some of the results let's go so already in 60s 70s uh Roy Amara um created am Mars law which is essentially saying that we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run and as we've now been in the generative AI era for a bit over two years essentially like the the democratization of generative AI we've already seen that the tools have evolved hugely and already like this week we've had lots of new updates into some of the tools I'll be showing today and also some of the tools that that hun will be showing and we already know that the technology is running in front of us but at the same time how do we put them in place in organizations in an effective way how do you change your organization in an effective way we see that there are two change movements happening in the organizations the developer platforms and platform engineering is kind of continuing the devops movement in organizations making the organizational transformation possible consolidating the tooling consolidating the cognitive load within developers and providing the tooling as a service for the developer organization the organizational transformation where is then at the same time the whole AI driven development moving is coming on an individual level into organization so I as a developer and getting the tools that I can use so GitHub co-pilot has been around for now four and a half years I think and uh and we are able to start changing our way of working and of course that will then integrate into the organizational transformation and build something bigger in the future if we look at some of the statistics of AI co-piloting we already know that most of the organizations and their developers have already started using AI driven coding in one way or another in their work yet only less than half of the companies have actually implemented AI in their software development and of course this number is growing all the time but it's still the reality that organizations that have the front running uh uh point of view they are the ones that are moving in the front but then there are lots of organizations who don't either see Digital Services in their core strategy or ar just aren't yet ready for this movement aren't there yet and this also reflects the fact that only 11% of organizations are actively using AI in their workflows meaning that it's still very much a developer tool AI co-piloting promises are us big changes like big boosts in our development flow and this is of course from the GitHub research saying stating that AI co-piloting is making developers 55% faster however when we look at it from the organizational point of view this is not always the truth we know that if organizations only small fraction of organizational R&D budget is is spend on the actual development work doing it 55% faster doesn't change the organization so much on average organizations tend to use 5 to 10% of their R&D in the actual development so making that 55% faster would be 2 and a half to 5% faster organization so we're still far away from getting the actual benefits that are promised by by these researches however from our customer base we've seen two things happening clearly so the first one is from uh from a Fortune 500 company C CTO stating that we've seen benefits of three to four hours per week this is by the way one of the questions in poll so if you're answering the poll and you're developing your organization I would highly value your answer in this this is the finding we've had is that developers will be 10% faster in their daily work with the means of AI development and 10% increase is huge like it's one of the biggest changes within development and development future that we've seen so already it's not 55% but the change is huge and it's only going to accelerate in the future the other interesting finding from a very regulated Fortune 500 company is that developers and end up writing more compliant and consistent code with the means of AI development and this is kind of natural because AI is helping the developers from the knowledge base of code that's already seen as compliant and consistent it's come it's bringing the ideas that are already from its base compliant and consistent and this only supports the learning curve for developers and of course when the suggestion are coming in they have been more or less trained and pre-screened this doesn't mean that you wouldn't have to yourself understand what's happening but on the basic level this is what actually happens with these organizations in mind and the changes that are happening within organizations we see that developer efficiency is definitely a top level priority but the engineering organizations aren't yet able to have an effect so we know that 45% of the cxos of organizations are not measuring the productivity change within the organization they're not measuring how the transformation is affecting the actual business outcomes they might be looking at R&D hours or tickets or velocity but the business outcomes are still yet to be defined to be seen how what is the effect in in regular organizations the second thing we always see which is also one of the reasons for this webinar series is that onethird of developers time is still spent on maintenance and tooling so tinkering with the tools that the developers should be using fluently in their organization and even while we are talking about the a tooling most of the vulnerabilities are still found in the very lates of the software development life cycle usually after the code has been merged it's gone to either security team or some more extensive automation finding these vulnerabilities and taking it all the way back into the development organization to the beginning of the life cycle so before jumping into the co-pilot and its features I want to talk very briefly about the adoption curve within organizations so at these two years that kind of the underestimating the change that's happening in organization we've been focusing on developer happiness we've been serve we've been served by GitHub copilot as a developer tool easy to implement easy to start using and within organizations this is also important for you to focus on serving the innovators and early adopters in your organization to use these tools in an efficient way being the innovators within your organization but the real challenge comes then when you have to focus on the late majority the lards the people who don't feel comfortable using these kinds of tools or don't naturally start using them in your organization and there you have to start building kind of the automation the platforms the way how these people get automatically these kind of tools in use and get the benefits with less and less uh effort done by themselves so putting Focus then on the productivity and control within the organization as some of the statistics on organizations most of the code natural is still written without the help of AI tools most of the organizational team still working on addressing technical debts and 75% of the biggest organization still use main frames so we already know that we're very much in the innovators and early adopter face on all of this but the change will happen super fast which you will also see from my demo how well the AI is already integrated into the development flow a few months ago with uh with my colleague henko we made kind of a prediction of how AI driven development will look like in the future and we had this slide as uh 2030 plus as in defining Target State and starting an AI engine working with the developer on stating what we need the AI creating software and then with the help of users and the data collected AI will be proactively able to suggest changes in the actual application and funny enough just a few months ago the co-pilot workspace was released I'll be showing it to you if it works I think it works um I said it's changing all the time but it's the first concept of this kind of a development flow for the future where I only make the requirements that are needed from the software and then I start a discussion discussion together with AI on where should I take the development that of that actual feature and it's not too far away from actually collecting the data and understanding the customers and users and bringing that insight into the automation Lo now before jumping in I'll I'll remind you all when listening to me that when you go back to your organization there are three key things you have to keep in mind to make the change happen so the first one is the benefits might lie somewhere unexpected this means that when you start using these kinds of tools you might might end up seeing that oh this works actually much better in quality assurance or our security is so much better when we start using these tools so you have to be open-minded for Innovation within the organization you have to make this kind of a tooling easier to adopt than any of the Alternatives this also helps the organization to avoid Shadow it and centralize consolidate the tool chains and then being able to balance between the autonomy autonomy and standardization so making the rule sets clear but still leaving space for Innovation let's jump right into it so introducing a driven development today I'll be talking about GitHub co-pilots and I think there are four key areas that you should keep in mind when working with Aon tooling on a developer key organizational level the first one is generating code so how do you generate code with the means of aif and one of the ways of course is for example now the C pilot edits at the agent OD which is able to generate all your bootstrap code all of all of the uh boilerplate code already from for you from the very beginning so kind of that's the extreme of generating code understanding AI helps you understand your code base the libraries you're using the interfaces and the new code that you might have to read and understand automate so I said co-pilot is not just for creating U features it's also for testing security building the automation pipeline continuous integration continuous deliver and then also automating the infrastructure ultimately and then fourth is learning so from learning to new languages to mastering new skills like testing and security for yourself with the means of AI let's get into it so I'll start with autocomplete inline and shell for this purpose I have and I'm doing this on a very simple application I'm doing it on Purp purpose this all works on a more complicated projects as well but it's easy for me to show here's a week. hopefully. works it's an application that shows the week for today and I have my terminal here mpm start I've started my application it's running on this smaller browser you so you'll see it and then we'll go to editor week number we open the week number JS so we haven't implemented anything it shows a line for us I'll create function get week number date is today's date and then copilot kicks in so this is kind of the essentially the simplest way to use GitHub copile you start working on your current functionalities AI picks the domain it understands I want to do get week number the code looks like JavaScript it prop proposes it to me and with AB I'll just auto complete it proposes me export default but since I'm using vanilla JavaScript I need to use module exports get week number save looking at the browser and there we have it the week number 10 amazing so this is the easiest way of using the T cop now let's go to slash commands so I have this existing code I'll actually take another project so you see that can be used for a bit more complicated programs slash command here I have initialized socket here keyboard shortcuts uh I'm using OSX but it's very similar also in the code in in Windows forment I directly in my editor I have asked co-pilot and I'm using slash command explain enter what happens is GitHub co-pilot kicks in it reads the code it explains to me what it does and then I add if I add the context here I will also be able to attach it to all of the other functionalities within the application and by the way you'll see here that there are the models that you can select to use so it's not bound to only the gbd4 you can also use a bunch of other different models for your AI engine as well now let's go back to my week number I want to write a test it should return week end for 6th of March 2025 there we go I written a test let's go back to the terminal stop it for a sec and run the test cases oh the tests pass so I've been able to also write the testing with the help of AI here and then I could select all of these and ask command ID IE fix I could say please refactor this code and see if there are duplicate tests then it would start it would run it reads my code it sees if there are any changes and if there are it will generate the edits for me so already kind of working towards the fact that I understand the coding but it will also propose me changes and now here it says fix the code REM any dolic tests refactor the selected tests so I'll keep it as this this is just an example an explaining I already showed then one super interesting extra feature which I see in the kind of my normal development flow so I can say GitHub copilot suggests how can I see the lines of code I haven't tried this out by the way I'm just asking it uh how can I see the lots of code so it asks what kind of command can I help you with I'll say generic shell command it says clock uh and the current folder I can either copy the commment to clipboard explain the comment what it does or execute the command or if I'm not happy I can revise it I'll copy it to clipboard and then if I paste it I can run it like so of course now it goes through also the node modules so it's going to be it's going to be a crazy number of of uh lines of code for one a weak number but you'll see soon there we go so 1.75 million lines of code for one week number that's dependencies for you you can also make an alas so how can I find all files with cons in them exclude node modules oh let's try again no there we go so I made a question mark question mark Alias which already wants me to do a command line command line request so here you can see gra con stex node modules I need to explain I'll just execute it and it would be executing the command for me so these are kind of the immediate benefits you get from just the co-pilot vanilla co-pilot so speak the one that I have integrated in my into my editor and the shell good let's take a look at copal chats and edits so I'm you'll see I'm working on insiders but these functionalities which I show you will be also working super well on on the VSS code and of course number of other editors we'll start with set so comment control I workpace what does this code do it takes a look at the workspace it gives me an explanation of this application and it tells me what the code does okay I was slightly confusing so I was asking asking of the code here present and not just the workspace but you'll start the chat and you can have a conversation with a essentially so this is basically the first part discussion on topic I could ask whatever and as chat participants I can ask like I can add workspace I can add project I can here I would I would be able to add more um more features on top of the chat so essentially that's how it works and then I would be able to have like and I go to co-pilot edit it will which is kind of chat with an editor I now have agent mode it's actually comment comment dots which I can say to change between edit and agent mode so I'll be in edit mode this is the one that you'll have in your vs code as of today I can say please add a test for next week's Tuesday and it reads my file and it will hopefully add me a test case for next week's Tuesday probably people write it as next week's Tuesday because I asked it to but this is once again also for me learning to write it I want so you'll see should return correct week for next week's th Tuesday there we go then I can either keep or undo it good and then the last but not least which I will show to you is the agent mode and the agent mode is one of the most exciting features that is around it was released I think last week's Monday and it's still very heavily in experimentation I'll make a directory C pilot agent I just call it copilot agent copilot agent I'll open uh Visual Studio code insiders I'll open the edits command dots I'll change it agent I want to create a v react project in the current folder so instead of just proposing me changes in the code you'll see that it gives me a run command in terminal and command enter it will run in terminal uh create V by projects it says I should also run npm install so let's run npm install this takes a while so now it's install the package is amazing then let me see please start the project okay npm run Dev continue it started the project let's open a browser oh there it is white plus react so running not not a single line Co of code return so now we moove the basic component and replace it with the weak number there we go so now what it does of course it kicks in the co-pilot edits it removes the component it generates the edits for me hopefully removes the kind of all of the basic um bootstrap stuff it has there seems promising Cent week number applying edits and I said once again I'm using weak number because even while it looks slow you'll already see from the code that coding this by myself would have taken way more time or than just some time in in stack Overflow there we go let's keep it save it and go back to the oh there it is week number nine and here as you see the week number is difficult to calculate it's week number nine so I've had lots of magic there's still what's wrong with the actual algorithm but we will not go there today I'll just push on and we'll talk about coiled workspace so all of this has been in the editor now uh presenting copal workspace it's something that was introduced uh not too long ago and how it works is I have this home weather monitoring system written in react and VES as you already saw so I can see it's 10° here where I'm currently and you'll see also some temperatures here and then the code is in GitHub I've created an issue add battery display to temperature boxes so I have the temperature boxes here and collecting the information from a Ru pag uh through Bluetooth and I'm not sure how long these tags will be running so I would need to add a battery display to temperature boxes this is my issue and what workspace does is it leaves the whole editor behind so it will actually start a session with AI which understands my code and it goes to the project says no the current C does not deplay dis display battery information in temperature boxes it gives some it gives a proposed solution uh temp box sh sockets then I can give more instruction like sockets already sends the battery level in curent dot battery and the threshold for red should be 2,800 reads the files understand what I said as adds it as an instruction I'm happy with the instructions let's generate a p