0:00 Oh, I can code 55 times faster. How many hours a day do you code as a developer? The value was not in the coding. 0:06 The value was in finding that solution, in defining that solution. Junior engineers are more likely to adopt AI. They are more native to the technology. They're not too opinionated on that yet. They're the change agents. 0:19 If you cannot experiment on a day-to-day, which means at your job, it might be actually a good reason. 0:24 Just expand your mind, make sure that you you are aware of a little bit of everything. This is a craft that is changing super fast. Coding is cheap. 0:33 Software is expensive today. I'm joined by Susanna Daniels, Chief Developer Advisor over at Microsoft, and with close to 30 years of experience, she has more than yours and mine combined. Potentially. 0:47 We talk about the future of software engineering with tools and a gentic, the things we have nowadays, but more specifically how you as software engineer can navigate this and much more. 0:59 So enjoy, I think a big narrative nowadays and a trend that really focuses on specifically productivity. 10X engineer is something that was already a phrase in the past and now people are saying, oh, it might be a reality. 1:15 What is your view on this emphasis on productivity specifically? So yeah, that's a bit of, that's an interesting question. Early days when we, when we had AI, we were talking about mainly code completion and asking an agent to do some work for you, right? 1:36 Task agent, human. Back then the focus was on productivity because that was what you could achieve, right? 1:47 But that marketing was so good that we were talking about, oh, I can code 55 times faster. How many hours a day do you code as a developer? How how much is that? 1:58 Like 14%? You can say one hour, you can say 2 hours. So while that is an enormous benefit, don't get me wrong, the question is, is all there is productivity? We learn now that with these agents and these models becoming more powerful and us 2:18 understanding better how to operate them that we can also talk about improving the quality. 2:24 We can also improving, talking about improving the outcomes that we, that we would like to see. 2:32 We can talk about instead of productivity, we can talk about the pace of innovation. So there's a lot more to it. 2:41 I think that the productivity narrative has worked really well, but we, we are way beyond that. 2:48 I think productivity is maybe the first level of AI adoption, right? I'm freeing up some time. 2:54 You're going to use the time to learn more, to understand more about these agents and how they can help you. 3:01 And that way you're going to add more and more value. 3:04 But yeah, it's a it's a very, very stubborn, reminiscent of of previous marketing, I think heartache maybe. 3:13 But yeah, this isn't it's not all story. 3:15 Yeah, No, I like your emphasis on creating value. 3:18 That's why I always had my joy. It wasn't necessarily on the the act of creating software, although I did have a lot of fulfilment in there. But that part is now I think the most evolving. The outcomes I'm working towards, I think those are still what makes me take at least as 3:33 engineers specifically. But there are different flavours of engineers. Some people really like the the hands on part and they really like and they hold a lot of passion in the craft of creating code where they say this is elegant code and we've refactored it and it's not 3:46 really nice. That fulfilment part, I think that's impacted the most so far. Yeah, it's look, if you, if you look at software development, at least in my opinion, what we used to do is we would have a problem, a challenge and we 4:03 would, would start figuring out what is the algorithm to get to a solution, AKA the coding the you know, but if you look at overtime, software development changed a lot, right? 4:19 So for instance, gathering the right requirements, that's so much work. It became a full time job for a person, right? If you look at other parts of software development that you know, that became jobs, that became roles, that became things that other people would do. 4:40 So if you look at the essence of software development, a lot of it was stripped away basically. Why? 4:46 Because we're humans. So if we have big things, if we have big teams, we, we need to figure out the process, we need handoffs to ensure that we, that, that we, that we keep that flow going, that we satisfy the requirements of the next person or the next rule that picks up a task. 5:12 Is that really software engineering And not to me when I started software engineering, what I loved about, about it is really knowing the problem, knowing who the customers are, matching that with some ambitions coming down from 5:26 business and figuring out how I could make that all work. 5:30 And yes, I spent a lot of time coding to do that, but the failure was not in the coding. The failure was in, in, in finding that solution, in defining that solution in, in getting that feedback and, and translating that in some technical, technical solution for that. 5:49 And with, with now with AI, we, we see that we are more involved in the beginning. This will be increasingly more I think that the boundary between products and software engineering will will be non existent at some point. 6:06 Maybe somebody has a role or or you know is is better at doing 1 task than another and and you know keeps the control and, and has some some responsibilities there. 6:18 But it should be seamless. And with that, a developer is involved, a developer is involved in the outcomes because that is what we're actually doing. 6:26 That algorithm, algorithm that will be solved by AIAI will will take care of that work for us. And when it doesn't work, we can still dive into the code and do the nitty gritty. 6:36 Some things are maybe way too bleeding edge for an agent to understand. Could be, but that is then the more fun work, right? If you ask most developers, we did internal internal survey and one of the things that came out out of it is that that's actually the work that 6:55 developers enjoy doing and figuring out the the solution, figuring out the outcomes. So yeah, it is shifting. 7:03 It is changing. I think that for the for the coders that like to code, that's great. 7:09 We still have people also writing the Linux kernel, right? 7:13 We still have people that worry about, you know, that's writing assembly. We have people that write COBOL, that's all fine, that will change. 7:21 But yeah, there's something to do for everyone. 7:24 I think we're in really exciting times, as you mentioned, where teams are kind of navigating this newer way of working. 7:31 For me, it's probably one of the first times where I've seen this much impact on what I did as a craft. 7:37 I don't have as much years of experience as some other people, even the people that are listening here, But for me, this is really cool to experience this part. 7:45 There's also fear because as individuals are kind of amplifying and creating value to a level that is different, that is evolving. I've also seen organizations that say, whoa, for the people that are really senior, they 7:57 know kind of the boundaries and they know they have quite some knowledge under their belt to accelerate with this. 8:03 I saw a chart where it said that senior positions went up and then the junior positions actually went down. 8:09 And I've also seen a narrative just now I was reading on Reddit where someone said, well, I have my degree and I have really good grades, four point OGPAI study from this school, but I've done X amounts of thousands of applications and I still don't 8:21 have a job. That is a little bit concerning when it comes to new and early in career people actually coming to the market. I'm curious because you engaged with a lot of engineering leadership, what your perspective is on that? Yeah, that's interesting, right. 8:34 The research contradicts what we're seeing that what we're seeing happening slowly. So for and this is maybe also the positioning about AI. So no judgement there. 8:49 I think that a lot of a lot of misunderstanding is that AI can do the job of a junior developer. 8:59 So we don't need the junior developer anymore. 9:05 There's there's a few things to that. 9:08 I think the most important one is that when you, when you look at the skill and doing software development at skill, it doesn't make sense. So you can't only hire senior engineers, There's not enough of them even like with with the 9:33 projections that we see with the number of engineers that that will be there in the following years, it will not match what is actually needed. So that's a bit counterintuitive. I think one other thought is that when you look at actual AI adoption, research tells us that 9:56 junior engineers are more likely to adopt AI. 10:02 They are more native to the technology. 10:04 They are more able to go beyond the like, what some few as classic developer work. They're not too opinionated on that yet. They're, you know, willing to use AI to do something beyond what is in their original job description. They're the change agents. 10:24 So they're that's also a way to look to look at it. 10:29 And I think that makes perfect sense. 10:31 And my expectation is that what we'll see happening is that at some point this trend will be broken and we will actually see that the junior engineers becoming high demands because these junior engineers are able to navigate this more perfectly. 10:48 They can handle very much the ambiguity that will certainly certainly exist in software engineering role. 10:59 And the senior engineers can be there to as SM ES to codify their tribal knowledge, to codify best practices so that agents and humans can leverage that. 11:13 And also to train these junior engineers because we, we, we need steady supply of, of people that will be able to, to become like, you know, senior software engineers at some point in their career. Then, then there's also the profile. So instead of like only looking 11:36 at the T shaped engineers, we will also see M shaped engineers that are, you know, better qualified at two different, different areas. So yeah, it's an interesting, interesting thing that is happening. 11:51 But if you look at everything combined to skill, culture, the ambiguity and the role that that will be probably what we'll see happening over the following year years, yeah. 12:08 For a lot of my career, I was the person that was early in career. I haven't been in a position where I actually engaged with A-Team that is quite younger than me. I'm now 30. 12:19 And I recently had an experience where there was a whole team of people younger than me. I was like, they actually have a really good energy and they're experimenting on their own and they're really wanting with a lot of drive to push this 12:31 project forward. And I was like, yeah, like I used to be like that. I can still see people like that. I think when you have experience under your belt, you become a little bit more sceptical, realistic. Maybe you burned your fingers a 12:42 few times like it's normal, but I do like that energy. 12:46 I miss that energy. And I feel like if an organization wouldn't have that, it also doesn't inspire the people that are there anymore to a level that they should be inspired, I feel like as well. It's that exactly. 12:57 And remember, coding is not only about the tools, it's not only about the people, it's also about the culture. 13:06 And that is also something that's, that's those juniors bring in these can be your change agents to, to, to get to that culture. Technology helps us there as well, right? So if we, if we would see that, let's imagine you have an A Gentile team. 13:24 In my mind, that's three people, lot of customers experimenting with exactly that number. And that could be like 1:00 PM and 2:00 juniors. And you know, anecdotally, people say, oh, that's a nightmare scenario because you will never know what they will create. 13:41 And those agents will will run riot. 13:44 And in all their enthusiasm and with their lack of knowledge, yeah, they might create like something we would think is suboptimal, subpar or, or whatever negative connotation. 13:57 And that doesn't need to be true. 13:59 Because if you look at like that model, and again, this is this is a bigger change, right? This is culture. 14:07 Imagine that these SM ES codify their best practices. 14:11 We implement platform engineering in the right way. 14:16 We have that culture where where collaboration, where creativity, embracing uncertainty, those principles where, where that where that is where that is in in in the DNA. 14:34 Why would it get off script? Because if they're able to define the right outcomes and have to write guard feels and constraints in place, then neither did the human nor the agent can can, you know, can go off script too much. 14:52 You're setting them up for success. 14:54 And I think that is a bit of the like the mind shift. 14:56 How can it make others successful? 14:59 We have a thing in Microsoft and it's it's high culture, high performance culture. High performance doesn't mean that one person is super cute and like, you know, wears a cake all day. That's also the person that enables others to be successful, that enables others to build the 15:19 next great thing. And I think that that is something that's it just makes sense to to look at it in that way instead of like, oh, I'm going to replace juniors with AII think you shouldn't replace necessarily anyone without AI 15:37 unless that there's like literally no activities anymore for a person. But yeah, the strong conventions and the guard rails and a platform where people can accelerate, I think those are a lot of the ingredients that platform engineering brings to the table or also a lot of the 15:54 problems that it solves when that is not there, when there's friction or whether there's different flavours of technology and kind of wild spread, which really goes hand in hand. 16:03 And I feel like platform engineering now should also have an emphasis on making this work in energetic way of working, which is really cool, to be honest. 16:12 That might be then in this vision where seniors really make a difference, where they say, OK, these are from what we know, best practices, strong conventions, differentiating factors for effectiveness. When we talk about small 16:25 incremental teams that deliver value, which can be very cool. 16:29 Are there organizations experimenting with this way of working as you mentioned already, also with regards to the seniority within teams? Yes, yeah, yes. 16:37 So then for people that are early in career and that are listening, they just have to find those organizations. 16:42 Yeah. Look, hang in there. 16:45 I think it's, and this is basically advice that's that's I'm giving to anybody that is, that is new in career. 16:55 Just expand your mind. Make sure that you, you are aware of a little bit of everything. 17:03 This is a craft that is changing super fast. 17:07 It doesn't hurt to know about proper UX, you know, and, and being able to understand that doesn't, you know, with, with all of these topics, security, the product management, it, it helps if you know a bit of that because that is the value that 17:25 you will bring extra with it and just don't give up. 17:31 I I think and I see that this will change at some point because economics of skill tell us that this is not a sustainable model, right? Yeah, a lot of my emphasis I think also in the impact that I had early in career was the fact 17:48 that I'm a good communicator, collaborator and a kind of bridge gaps. My fear is that, and this was also just me coming into a project and being the sole person that needed to build an integration, just get it done. 17:59 I was like I I could do it. It was a lot of fun. 18:02 I did not interact with as many people within my team because it was kind of a one man army task. And I feel like that will or there is the option to do that more and more. 18:11 So then there would be less inner operatings with a bigger team. You already mentioned that 3 might be the the solid number and I see there's a good thing and a bad thing. I still think that the soft skills aspect of things collaborating, we will still be 18:25 a lot of interaction with people in and of it's own. 18:29 But some of the complexity also goes into and genetic work which is amplified on a different level. 18:34 Yeah, if you were to advise people what to focus on, it could be people that are early in career, it could be existing software engineers. They're experimenting with genetic work flows. What should they now looking at first principles from software engineering focus on with 18:48 regards to knowledge from your perspective? 18:52 That is a great question in terms of knowledge, of course, beyond coding itself and understanding these agents. 19:00 It's a good name now. Yeah. 19:08 What it what it is that I I literally see happening is that a lot of a lot of principles, a lot of frameworks, best practices that are older than I am, maybe even they they, they get a Polish up. There is a lot of principles that that have been around forever. 19:39 There's a lot of best practices and we see that with 19:49 specifically the the specification parts, specifically the beginning of of that software, software development life cycle. I, I it is those principles I would. Advice to take another look at right a, you know, if, if you really understand everything 20:10 there is about behavioural driven development, that's a great one, right. That's valuable understanding a bit more about systems, understanding more about, I don't know, the SOLID principles and there's so many, like there's an enormous list of things that you can just now 20:35 acquire in, in terms of knowledge. 20:36 It's not difficult to do. But but if, if you understand these principles, then that will help you massively in, in defining these specifications and defining these constraints in making an making an actual software products instead of codes. My manager has a nice one and I 21:03 just have to put it out there. She always says, says coding is cheap, software is expensive. And that's just the difference between between just like putting out some codes or being really intentional about what you're doing, knowing your, your 21:22 heritage, knowing, knowing whatever that these researchers have researched for years, all those best practices, leverage them. Learn that be, be, be good at that. Be be the one that says, but what if we would like, instead of using this method, use 21:42 another methodology? You know, I think that's that's, that's that would be beautiful. Yeah, I'm happy that a lot of strong principles and like foundational knowledge, knowing that will still help in the end. We're not really vibe coding on 21:59 enterprise skill. We're still creating software that for the most part is going to be there hopefully for years on end. Creating is a lot easier. 22:08 I feel like creating in the right way we're getting better and better at, but we will go into a mode where creation is just there in abundance. So then we have to make sure that we create the right thing, created the wrong the right way, 22:20 and then also has the longevity component to it. 22:23 One of the reasons why I really like Go as a programming language was that the creation process in the end, what you see in code was quite simple. It's still on version 1 dot something something. It's probably in the 20s now, 22:35 but it's still a one major version, which means I never had to migrate my code. If there was a new update, it was kind of seamless. That process was always pretty simple. There are not many things that I could do in different ways in Go as I could compare to other 22:50 languages, which is why I really enjoyed it now. 22:53 It has a strong readability component, which is fantastic when we're talking about software that is generated because if you have a language that's easy to read, you will be reading a lot more I feel like nowadays, which is also funny 23:04 when we're talking about programming languages specifically. I wonder if agentic software engineering is also going to influence in the end the language that you choose right now. 23:14 I think it will because we are doing a lot of the reviewing work as humans still with an agent still probably like pre reviewed, but in the end still as a final pass. 23:23 There's humans in the loop. I don't know how that's going to evolve, which I think is still quite exciting to see how it plays out. Is this a long intro to asking the question? Will everything be JavaScript? 23:36 I hope not. But yeah, what? 23:37 How do you see that? Luke, I think that yes, of course there there there will be some influence here and there because the agent, if you don't stare at the agent, picks what the agent will pick. Yeah, right. 23:52 On the other hand, who is defining what the agents should build? It is us humans normally. 24:00 Yeah, still so, so I'm, I'm not too, too grim about that. 24:07 But one one thing that's this is maybe a bit too hypothetical, but you could wonder like not if in 10 years or so the the default programming language will be JavaScript, which we can argue about, we can argue about this for an hour maybe. 24:30 But it is like, what is the future of software even? 24:35 And what is the future of software languages at all Like? 24:40 Is it too far fetched to think that in, I don't know, a few years we will create applications using natural language and that will be compiled. 24:52 Who who cares? Like what, how it is compiled, what language it's using? And so that that's one thing. 24:59 And the other thing is what even is an application? 25:03 I work with customers that are rethinking their their apps and they're saying, well, but my customer, yes, sure, they go to a website right now, but if they are in an in an, in an agent, they're using, I don't know, something like chit GPT or 25:22 clouds or whatever. And they they ask a question like, Hey, how can I get how, how do I choose? 25:29 I don't know life insurance. So why wouldn't it just suggest me the good life insurance and I just close it right from there? 25:40 Or even would it be too far fetched to think that when I would a life would want a new life insurance that there would be agents being kicked off to negotiate on my behalf And just look which one comes back with the right offering or when I 25:57 want to do my groceries. So this is super exciting. 26:03 But yeah, there there's, there's of course a lot of beliefs about the way that that we are looking at software development right now. I think that will change massively. But but yeah, I on on the programming language. Yeah, it is of course there was 26:20 a preference to the colour purple according to some LLM right. Oh really, if the people would build a web app, then it would use the colour purple because The Color Purple and then some reason, some research, I don't know. And thus like everything becomes 26:36 purple. So if then in the next training would take the place, then everything would be purple. 26:42 Or you know, there's there's a lot of things to to say about that. I was reading an interview with it was Reddit and like someone these stated, yeah everything we say on Reddit becomes becomes the future training data. 26:59 Yeah. And I don't know if that's a good thought or or a horrible thought, depends on, you know, which subreddit you're on. But yeah, there, there. 27:09 There's a lot of implications with that. 27:14 Every every so now often I'm on Reddit or I'm on YouTube and I scroll through the comments and someone makes a comment that says all of these other comments. 27:22 They sound like bots and I'm like I have no clue who I'm actually talking to nowadays if that is the truth or not. 27:28 Is it like with the M dash? So I. 27:32 When I see that, I'm like, that's the smell. 27:35 But so, yeah. But we think then that that's that's a thing that an agent does. 27:44 But on the other hand, like, I always learned that I used too many commas. Oh, yeah, yeah. 27:51 So what did I do? You removed the commas, yeah. 27:54 I moved to M Dash because I that's the way that you know, that's that's my way of communication. 28:03 And then I had a whole period where I was like, yeah, I'm just going to going to replace all the M dashes with a minus sign or something maybe. But it's just, yeah, there's no at some point, no way of telling, I think. 28:16 No, I mean, I'm fully there with you actually. 28:19 That's so funny. I saw on LinkedIn everyone people were like, I was doing M dashes before it was cool and now all my articles are getting rushed. 28:25 Like, this is also not a good movement, but yeah, yeah, that's funny. As that's humans and you know, the, the, the point of this is as humans, we, we value that somebody communicates as a, as a, as a natural person, as a human. Yeah, authentically. 28:44 Authentic, human, real. On the other hand, we all love it when AI does that for us. With with coding, you know, it's the same story. We still own the code, we are still responsible. The human is still responsible for what we're writing. If AI have written 80 or some 29:07 customers, even now, 90% of the code, 90%, their engineers are still responsible. They are still the humans that are in the loop. We're trying to remove it where it doesn't make sense if we have enough confidence, but but still, it's a human that's that's responsible for for what 29:27 gets out there. And if I use AI to translate my ideas and people can actually read it, I think that's way better than when I spend too much time on writing it myself. 29:42 And people read it and think, wow, she's unhinged. 29:46 But this doesn't make sense. I don't know. 29:48 I stopped halfway, right. So that's, that's always a bit of the, there's always a train off there. 29:57 Yeah, yeah. I had a conversation with Gregor Hope on the podcast as well and his stance was really OK. 30:03 The if the baseline has moved because the baseline is now you have output with regards to AI, you still have to add value on top because otherwise anyone can do the same. 30:12 Maybe the reasoning or the prompting is slightly different, the flavour is a little bit different, but if you don't add your value on top then basically anyone can do it. 30:21 Your value is really the value you add, so keep doing that and otherwise it's useless what you're doing. 30:27 And that's an excellent remark, right? 30:28 And This is why it doesn't make sense if all people use AI to, you know, I think the most visible thing is LinkedIn. 30:39 If it's like the 19th article that is written by AI and it's all looks the same and it's like nothing is being added, yes. 30:47 What is the purpose? The same you can say for software, right? And on one end, we, we with AI, with this technology and with the right culture, we enable everybody to, to get the most out of it. 31:05 Because the the remark of a customer, the idea of, of somebody in the business, the the, the experience by the PM and by the engineers that are involved that can, that can create something great. On the other hand, we also see a 31:25 lot of software companies now struggling because they created the product, but basically it didn't add much. 31:32 Think of the example where you have, I'm not going to name a shame, but you have an open source product. 31:38 Monetization with open source has always been challenging. 31:41 So one of the scenarios is of course you release the premium feature and you maintain that and the customer buys that and with that secure that there flows money into the open source project. I would wish that everybody 31:56 would think that way, but also somebody could just say, well, I just used the product sheet and the recorded demo and I told AI built this for me and I got it 90% right. 32:09 And I think that's good enough. And of course, I think that everybody should pay for open source, but but yes, support, support, support the makers, support the the contributors and the companies that make the project a life, that invest time and and resources in it and money. 32:32 But yes, you need to add something really special. 32:34 Yeah. Otherwise it's not good enough anymore. I feel like open source is the value of open source and kind of the monetization model. 32:43 It was always challenging. So open source projects really nailed it and they are thriving. And the other ones that added it like an add on that one is the, I think impacted the most in terms of now people being able to kind of copy and and match 32:57 things. Which means if that incentive is gone, I hope we don't get into a cycle where we have less and less things that are open out for people to use. 33:06 Because I do love a lot of my open source software. 33:09 And I do think it is, quite frankly, one of the most awesome things we have about this industry is the fact that people create, they share with almost nothing in return. 33:19 Even the open source projects that have monetised, it's typically just a fragment of it. It's not the main thing. 33:25 And the people that are maintaining it have a second like a full time day job. This is a side thing for them. 33:31 So yeah, I I hope we don't get into a a position where we have less and less open source, but this is a risk for sure. 33:39 Yeah, I, I, I hope the opposite. I hope that if we will see more open source and, but yeah, it's, it's something to, to that we need to actively work on. Yeah, the world depends on open source. It's like that for the foreseeable future. How many AI she'll build on top 34:04 of something and how many platforms she'll build. 34:07 Like there's still always that XKCD with that one package maintainer in Nebraska. But that's, that's the reality, right? And I, I see a lot of creativity happening and I think my expectation is that open source will still be there. It will be, will still be 34:29 booming. I don't know about the larger projects though, because now it's easier to say, oh, well, I wanted to in another way or, oh, I'm really specific about something. I don't have the answer to that. 34:42 But I think that's a, that's a thing we should solve as an industry as well. And tell people what open source is and and why it matters to contribute to like these projects and and work together. But. 34:56 I have one final thought and it goes into the line of I think tooling that we now have available as software engineers specifically. I know a lot of people in start-ups or smaller organizations where tooling is kind of experimented with. They might fall in love with 35:11 Cursor, for example, just to throw one out there. 35:13 And if they then move to a bigger organization, they will no longer have access to the tool that they love. 35:17 So now they use whatever is in our organization, whether that be something open, AI related, clawed, or get a copilot doesn't really matter. In the end. 35:26 They have to be more flexible when they switch jobs to also switch tools, which is pretty funny because I love my tooling. 35:32 I love mastering tools because then I feel like I can leverage that to achieve my outcomes faster. 35:39 It was also in the IDEI love my hotkeys, my bindings to make sure that when I actually execute code, I do that efficiently because like to be effective in what I do. 35:48 So letting go of a tool and then mastering another one, it's not really something I look forward to. 35:53 There are a lot of, I feel like, underlying principles when we're talking about agentic development, the way you prompt, the way you context manage, how you kind of parallelize agents or how you navigate certain structures. 36:08 We're still learning and we're still evolving also with how we structure context nowadays. There was this discussion on skills and agents, but I'm curious from your perspective, when you advise engineering leadership on the tooling they 36:21 make available for their department, for their teams, for their organization, what do you make sure that is there? 36:27 Because I'm quite curious from the people that are listening, they might not have the tooling they want, but it might be effective for what they need to do. 36:35 So I'm curious to hear how you think. 36:37 Yeah, that's, that's a great one. 36:39 I'm a lot of the, a lot of the tools, to be very blunt here, they're not so much different. Can you achieve 80% the same thing with all of these tools? Probably, most likely. 37:01 I didn't do the research on it, so don't pin me down. 37:03 But look in, in general, these tools are, are different, but but not that different. One thing that that's, and this is regardless of the technology, we cannot just ship a tool and give it to a developer, especially not when it's highly 37:22 sophisticated like an AI coding assistant or an an agent for doing anything basically. So what I advise these leaders is to at least enable the developers. 37:38 And with enablement, I do not mean like, oh, this is how you get the license and this is where you find it. 37:43 Although that's a really good start, don't get me wrong. 37:47 And already better than like maybe like half of the companies, you know, it's like usually it's just like, here's a license. Good luck. 37:55 You need to train on that. You need to understand like where do I put constraints? What are these constraints? 38:01 Like this is a lot of concepts to grasp agents. 38:05 Oh, I can write my own agent. You have to know how. 38:08 You have to know where. You need to understand the specifics of these tools and how they process this information. 38:15 Luckily, we have more and more standards, right? 38:18 This is why we as an industry trying to figure out like, OK, where do we put these instructions? 38:25 Let's adopt these skills. Let's look at certain protocols that help us. MCP is a great example of a highly fast moving, fast based industry that competes with each other to their life. OK, we need to settle on this. 38:39 We need to find a, find a standard for that. 38:41 Whatever you think of MCP, it's a standard, right? 38:44 Yeah, but people need to learn what is MCP, how does that work? 38:48 How do we build that securely? How do we skill that? 38:50 What can't we do with MCP? Looking at the different rules? 38:56 If you are a software developer, that's really great. 39:00 There's tons of information out there. 39:02 Even then, we need to teach you to understand how you can leverage the tool. But we also have data scientists, we have test engineers, we have PMS, we have, there's so many people that are involved in that whole coding thing and they all need to be enabled and understand like how 39:21 they can get the best out of the tools that they use. 39:26 Also about the models, what happens when I pick one LLM over another LLM? Sharing best practices internally, making sure that you set up an inner source culture. 39:39 There's a whole, you did not need to do that all at one time, 39:46 but there, there's like this whole trajectory where you can just mature and, and for each phase, you can add a little bit to it to ensure that you're successful with the tools. 39:57 And then when you need to switch the tool and you're like, oh, this is like, it feels uncomfortable because it's different or it's in another IDE even, or it doesn't even have it's own IDE or whatever. It's ACLI suddenly. 40:12 Look, it's the. Basically these LL Ms. and then this is of course talking, you know, working at Microsoft, we we can just switch the LL Ms. to whatever LLM that you use of course. And the most popular ones are in there and will be probably more. If the one is not in there, your 40:32 organization could edit even, right, to make sure that you can access that. And this will be further commoditised. We already see that, like what we now do with GitHub. We're our platform, we're welcoming. So if you want to use Clot as an agent, sure. You use Clot as an agent. 40:55 If your org enables that, that's really great. 40:58 Then you only have another bit of a different interface, sure, but your agent is still there, uses the LLM that you're used to. It's the agent that you're used to. And, and I think that that's, yeah, that commoditisation will will go on and I think that 41:15 everybody will benefit of that. But yeah, you can't just give it all to a developer. That's that's no, that's, that's not a driving adoption. Yeah, I like, I like that you say that an arbitrary percentage, but most tools right now, they kind of do the same thing. 41:31 And The funny thing is, I thought of this and yeah, if they don't do the same thing, then within a few weeks, within a few months, they will still do the same thing because people look at each other. If my tool has something that is 41:41 very special and other tools are out there, they are definitely going to also adopt something very similar to achieve the same outcome. It just makes sense. 41:49 It's good that we have competition because as consumers we get the best products in the end that really I enjoy. 41:55 Yeah, and, and me too and I love what's happening. 41:58 I, I love looking at what's what's happening with, with competition and of course I'm ready to, you know, and I follow these discussions and people showing off and on YouTube and on TikTok. It's just by looking at how they are using a product and how they're talking about these 42:16 products and helps me understand what is what is important. 42:21 And there's some nice innovations that are being put out there. Not everything I would say, well, that's that's a nice implementation or that doesn't really sound enterprise to me. But yeah, in the end, competition is what makes software good and great and 42:44 makes it inspirational to be in this field at this at this time. 42:47 For the people that are listening and they are not enabled to a degree where you say this is kind of the tooling, that you should at least have at least one of the tools, do you feel like that warrants them changing the organization that 43:01 they are in? Because I think I do almost if it was easy to just switch organisations. 43:07 Disregarding that, I feel like now things are evolving and accelerating so much in our craft that if you cannot experiment on a day-to-day, which means at your job, it might be actually a good reason. Yeah, totally. 43:19 If you if you look at at organizations and looking at their, their strategies on on retraining and retaining talent, there's a lot to do. Having a good culture, having a mature tools, which goes from your day-to-day software development to how easily can I obtain a platform? 43:58 How can I skill? How can I make these changes during all your platform engineering implementation? 44:04 Yeah, there will be, there will be choice, There will be choices. And you know, the more reason for organisations to think this true really, I'm certain that we will see more and more platform engineering conversations popping up. I don't know if we did a great 44:23 job at defining platform engineering. 44:25 I think we crammed way too much in there with a focus on engineering. So maybe it's time for another iteration on on on that approach. 44:35 But there's so much in there and we need that. 44:37 We need, we need that for the agents that help your customer. 44:40 We need that for the agents that help us write codes. 44:43 We need that for the agents that deploy our stuff into production and manage that and observe that for us. 44:50 We, we need all of that. People that are listening or watching might think, hey, yeah, but shouldn't it always have been that way? Yes. 45:03 But when you slap a high onto some project like that, chances that you get the budget for it and you get the resources for it are a bit higher as well. Yeah. 45:15 So do yourself a favour, implement platform engineering, make sure that you have the best culture look, be creative with with your collaboration model, your operating model experiments a bit with that. Make sure that you celebrate 45:30 your success, that people are enabled and encouraged to use AI and to fill. And I think if you do that right, you'll be solid and you're you're doing probably better than some other company and you will attract the talent. 45:48 Yeah, beautiful. Thank you so much for coming on, Susanna. This was a blast. 45:52 Yeah, thank you. It was so fun. 45:55 Cool. If you're still here with us listening, let me know in the comments section what you thought of this episode and we'll see you in the next one.
Community discussion