1 00:00:00,080 --> 00:00:01,920 Hi everyone. My name is Patrick Akil and 2 00:00:01,920 --> 00:00:05,480 joining me today is Shaheen Shakarami, Director of Data and 3 00:00:05,480 --> 00:00:08,640 AI over IKEA Retail. And we discuss how AI has 4 00:00:08,640 --> 00:00:13,400 impacted the tech industry. Nowadays, 2025 companies are no 5 00:00:13,400 --> 00:00:17,480 longer hiring junior talent, but Shaheen argues that now is 6 00:00:17,480 --> 00:00:19,320 actually the best time to hire junior talent. 7 00:00:19,600 --> 00:00:22,000 And how can companies do that and distinguish themselves? 8 00:00:22,360 --> 00:00:29,960 Listening to find out, Enjoy. And the tech industry is very, 9 00:00:30,080 --> 00:00:33,200 very evolving. And I wanted to hear your 10 00:00:33,200 --> 00:00:36,240 opinion on kind of how newer people get into the industry 11 00:00:36,240 --> 00:00:38,600 because I feel like it's it's very much changed. 12 00:00:39,040 --> 00:00:40,920 Hiring is now kind of on the back burner. 13 00:00:40,920 --> 00:00:42,920 I see a lot of bigger organizations laying off people 14 00:00:42,920 --> 00:00:45,480 in the 1st place. And I also see this trend where 15 00:00:45,480 --> 00:00:48,360 a lot of people coming into the industry are having a harder 16 00:00:48,360 --> 00:00:50,160 time finding a job in the 1st place. 17 00:00:50,520 --> 00:00:52,760 What's your take on that? How can younger people kind of 18 00:00:52,760 --> 00:00:55,280 distinguish themselves? Or should companies really be 19 00:00:55,280 --> 00:00:58,120 focusing on seniority instead of also hiring new talent? 20 00:00:58,480 --> 00:01:00,480 Right. It's a really good question. 21 00:01:01,360 --> 00:01:05,160 My main take around this is, yeah, indeed, we are going 22 00:01:05,160 --> 00:01:07,120 through a lot of changes during tech industry. 23 00:01:07,120 --> 00:01:11,080 We still have not found our feet on the ground of where AI 24 00:01:11,080 --> 00:01:13,520 exactly is bringing the most amount of value. 25 00:01:13,560 --> 00:01:17,080 Yeah, now we can write more lines of code, but does that 26 00:01:17,080 --> 00:01:18,880 really bring the value that we want? 27 00:01:19,880 --> 00:01:23,160 Now you have more P Rs to review, you have more XYZ and 28 00:01:23,160 --> 00:01:26,320 the whole chain is getting faster and faster and maybe some 29 00:01:26,320 --> 00:01:29,520 parts, maybe you're spending a lot more time designing things. 30 00:01:29,520 --> 00:01:34,720 So I, I think a lot of companies are stop are stopping either to 31 00:01:34,720 --> 00:01:38,840 hire new people or just completely saying, OK, AI can do 32 00:01:38,840 --> 00:01:41,400 what a junior talent can do. Therefore we are not going to 33 00:01:41,400 --> 00:01:43,880 hire anymore. I actually hear this a lot in 34 00:01:43,880 --> 00:01:47,680 the circles that I am in. And I personally think that's a 35 00:01:47,880 --> 00:01:51,400 not a very good approach because a young person is beyond just 36 00:01:51,400 --> 00:01:54,240 someone doing some job. It's also bringing new 37 00:01:54,240 --> 00:02:00,160 perspective, new fresh talents and also a lot of new talents 38 00:02:00,280 --> 00:02:03,800 have already been AI native during their study time. 39 00:02:03,800 --> 00:02:06,280 They've been dealing with ChatGPT or whatever other tools. 40 00:02:06,280 --> 00:02:09,320 So they're not like me that after certain years, they now 41 00:02:09,320 --> 00:02:11,480 need to change their way of working, the change of their way 42 00:02:11,480 --> 00:02:13,360 of thinking. And they use it a lot more 43 00:02:13,360 --> 00:02:15,960 natively as well into their day-to-day activities. 44 00:02:16,320 --> 00:02:22,280 So I think just stopping how stopping saying, OK, we're not 45 00:02:22,280 --> 00:02:24,800 going to hire any junior talent and we're going to focus on 46 00:02:24,800 --> 00:02:26,800 seniors. It's actually going to be a 47 00:02:26,800 --> 00:02:30,760 massive mistake for a lot of companies down the line. 48 00:02:31,720 --> 00:02:35,960 My main, actually, my main advice would be to the 49 00:02:35,960 --> 00:02:39,320 companies, actually, if you have the means, hire now that 50 00:02:39,320 --> 00:02:42,560 everybody is not hiring because these are the talents that you 51 00:02:42,560 --> 00:02:44,680 need to train and still need to learn. 52 00:02:44,680 --> 00:02:48,040 And while your senior people are still there, they can teach 53 00:02:48,720 --> 00:02:52,200 these folks the ways of doing and the ways of the craft. 54 00:02:53,000 --> 00:02:56,560 But also my advice to the younger people is to really be 55 00:02:57,600 --> 00:03:01,000 because they come with less baggage around tech. 56 00:03:01,000 --> 00:03:04,360 So they can be a lot more agile in changing, a lot more agile in 57 00:03:04,720 --> 00:03:08,240 what tool they use, a lot more agile in how they think and all 58 00:03:08,240 --> 00:03:10,760 of that and always do something extra. 59 00:03:11,480 --> 00:03:14,880 So this is what I do when I hire very junior talents. 60 00:03:14,880 --> 00:03:18,160 I always look to see what else do they have beyond their 61 00:03:18,160 --> 00:03:22,080 studies and activities, whether it's a very cool beyond their 62 00:03:22,080 --> 00:03:24,520 internship. Some people are hustling on the 63 00:03:24,520 --> 00:03:26,800 side, they have some side jobs selling coffee and whatever. 64 00:03:26,800 --> 00:03:29,080 And to me all of these are signals that these people are 65 00:03:29,440 --> 00:03:33,840 very, very agile and very open and have a growth mindsets and 66 00:03:33,840 --> 00:03:37,160 that they have that level of curiosity that they will survive 67 00:03:37,200 --> 00:03:39,360 and they will learn the skills very quickly. 68 00:03:39,360 --> 00:03:42,400 So that would be also my advice to more younger folks. 69 00:03:42,520 --> 00:03:44,960 Yeah, I like that a lot. I feel like exactly as you 70 00:03:44,960 --> 00:03:47,680 mentioned, if a lot of companies are no longer looking at younger 71 00:03:47,680 --> 00:03:51,120 talent OR less and less than there is an opportunity to hire 72 00:03:51,160 --> 00:03:52,880 that talent. But then still indeed 73 00:03:53,160 --> 00:03:55,880 distinguishing yourself if you find yourself in that group kind 74 00:03:55,880 --> 00:03:58,880 of early in career is going to be more and more difficult. 75 00:03:59,280 --> 00:04:02,080 I like your advice in doing something extra, whether that's 76 00:04:02,080 --> 00:04:04,520 like something entrepreneurial, something that's like open 77 00:04:04,520 --> 00:04:07,320 source, a really cool project that you've done for me, it 78 00:04:07,320 --> 00:04:09,000 doesn't necessarily matter what it is, right? 79 00:04:09,000 --> 00:04:11,920 If I can see you're passionate about it, it aligns with 80 00:04:11,920 --> 00:04:14,280 something that you want to do professional, even better. 81 00:04:14,520 --> 00:04:17,760 Or it shows me some of the core values that you are as a person 82 00:04:18,120 --> 00:04:20,959 through kind of your upbringing or kind of your discipline. 83 00:04:20,959 --> 00:04:23,520 Then it adds a lot to that person hiring. 84 00:04:23,880 --> 00:04:27,440 In the end, we'll hire people we hire also shortly pay a little 85 00:04:27,440 --> 00:04:30,000 bit extra for skills and knowledge, but we want to work 86 00:04:30,000 --> 00:04:32,840 with these people and that's still its core to what we do. 87 00:04:33,120 --> 00:04:35,440 Yeah, absolutely. I mean, if everybody in the team 88 00:04:35,440 --> 00:04:38,160 is senior, I think actually the mindset becomes a little bit 89 00:04:38,160 --> 00:04:41,000 negative toward everything. So that's why I do really like a 90 00:04:41,000 --> 00:04:43,200 team that has a right balance between senior and junior 91 00:04:43,200 --> 00:04:45,120 people. Maybe that's why I'm because I 92 00:04:45,120 --> 00:04:47,840 do really like that positive atmosphere at work, that 93 00:04:47,840 --> 00:04:50,840 optimism that certain people have. 94 00:04:50,840 --> 00:04:53,960 And and I think, I think that optimism just a little bit 95 00:04:53,960 --> 00:04:58,400 becomes sometime less and less because you get just more world 96 00:04:58,400 --> 00:04:59,760 experience to become more realistic. 97 00:04:59,880 --> 00:05:00,800 A little bit. Jaded. 98 00:05:00,800 --> 00:05:04,360 A little bit jaded. So that's why I was think also 99 00:05:04,360 --> 00:05:07,560 in general for general culture of your working space is very 100 00:05:07,560 --> 00:05:09,680 important. Of course, this advice might not 101 00:05:09,680 --> 00:05:11,320 be true. If you are a very, very small 102 00:05:11,320 --> 00:05:13,160 organization, you start up with five people. 103 00:05:13,680 --> 00:05:18,160 I'm not sure if having a a, a very, very junior talent might 104 00:05:18,160 --> 00:05:20,320 be the right choice for you. I mean, there are still really 105 00:05:20,320 --> 00:05:24,360 good ones out there that can do beyond our imaginations. 106 00:05:24,360 --> 00:05:29,240 But I think just the advice becomes more and more true if 107 00:05:29,240 --> 00:05:32,480 you're a bigger and bigger organization with a lot more 108 00:05:32,480 --> 00:05:37,400 resources that you can absorb these junior talents in that 109 00:05:37,400 --> 00:05:39,320 way. To play devil's advocate, then 110 00:05:39,320 --> 00:05:42,160 if companies are looking from the perspective of OK, I can put 111 00:05:42,160 --> 00:05:44,960 in the time and effort to hire junior engineers and they have 112 00:05:44,960 --> 00:05:47,680 value once they're there. But junior engineers also want 113 00:05:47,680 --> 00:05:50,640 to learn and kind of have career growth and they also want to 114 00:05:50,800 --> 00:05:52,400 learn from different organizations. 115 00:05:52,760 --> 00:05:54,960 So they might move after one or two years, which is kind of a 116 00:05:54,960 --> 00:05:57,960 wasted effort. It is kind of this balancing act 117 00:05:57,960 --> 00:06:01,080 of return, like investing and then getting a return on 118 00:06:01,080 --> 00:06:03,560 investment with regards to the knowledge, skills and expertise. 119 00:06:04,160 --> 00:06:07,120 So from that perspective, I can see people being more afraid of 120 00:06:07,120 --> 00:06:11,480 like investing in junior talent. I mean that that is true. 121 00:06:11,480 --> 00:06:15,440 If you don't have a very good way of retaining your general 122 00:06:15,440 --> 00:06:18,280 talent, that's can be true also for your senior folks. 123 00:06:18,280 --> 00:06:21,480 They can just leave after a year or two and that actually is even 124 00:06:21,480 --> 00:06:25,480 a bigger loss because they've they've been building a large 125 00:06:25,480 --> 00:06:28,800 portion of your code base or, or, or your tool or your 126 00:06:29,240 --> 00:06:31,560 software and now they're leaving. 127 00:06:31,720 --> 00:06:34,200 So I think you need to really balance it with the overall 128 00:06:34,200 --> 00:06:38,720 cultural change within your company to, to really bring them 129 00:06:38,720 --> 00:06:41,280 forward. And at the end of the day, in 130 00:06:41,280 --> 00:06:47,120 the grand scheme of life, we're thinking about the society or 131 00:06:47,120 --> 00:06:50,640 the city as a whole. And I think if a junior talent 132 00:06:50,640 --> 00:06:54,360 moves from company X to company Y and if all companies are 133 00:06:54,640 --> 00:06:58,120 hiring junior talents at the end of the day, and we have the 134 00:06:58,120 --> 00:07:01,120 right amount of flow of talent into the market, I think it 135 00:07:01,120 --> 00:07:05,160 benefits everyone. So it's, it's a larger benefit 136 00:07:05,160 --> 00:07:09,120 for the for the whole ecosystem, which eventually makes the 137 00:07:09,200 --> 00:07:13,200 ecosystem of the area better. And I think that's the same in 138 00:07:13,200 --> 00:07:15,080 Bay Area. I think you, you will see also 139 00:07:15,080 --> 00:07:18,880 for companies like Open AI, they hire a Group A lot of junior 140 00:07:18,880 --> 00:07:22,000 folks as well as well as very, very, very senior and very the 141 00:07:22,000 --> 00:07:25,240 best in their class work. So, so I think that's just the 142 00:07:25,240 --> 00:07:29,000 balance balancing act for that whole ecosystem. 143 00:07:29,760 --> 00:07:31,840 Yeah, that's a beautiful mindset I think to have. 144 00:07:32,360 --> 00:07:36,160 How can companies distinguish themselves and make sure at 145 00:07:36,160 --> 00:07:39,320 least have a higher chance of people staying with regards to 146 00:07:39,360 --> 00:07:42,000 what they do on a day-to-day? Or what can companies change or 147 00:07:42,000 --> 00:07:47,880 offer in the first place? I think we cannot as a as in a 148 00:07:47,880 --> 00:07:52,520 company you cannot really be good for everyone and 149 00:07:52,520 --> 00:07:55,440 everything. So how I think about it 150 00:07:55,560 --> 00:07:59,160 personally, my own wouldn't talk about on behalf of any company 151 00:07:59,160 --> 00:08:02,280 or myself. I try to look at what is the 152 00:08:02,440 --> 00:08:06,120 motivating factor of people in the team. 153 00:08:06,800 --> 00:08:10,520 Some people are just really motivated by doing the best and 154 00:08:10,520 --> 00:08:14,480 the coolest and the fanciest or the newest thing, which is 155 00:08:14,480 --> 00:08:16,640 actually myself. I'm part of those people. 156 00:08:16,640 --> 00:08:19,800 I just really like this, this idea of being on the forefront 157 00:08:19,800 --> 00:08:22,880 of things, even though it can burn me at the end of the day, 158 00:08:23,560 --> 00:08:25,520 but I still really enjoy it. And some people are really 159 00:08:25,520 --> 00:08:28,240 motivated by money. Some people are motivated by 160 00:08:28,240 --> 00:08:32,760 promotion, some people motivated by stability, work life balance, 161 00:08:32,760 --> 00:08:35,760 a wide variety of range of different levels of motivation. 162 00:08:36,600 --> 00:08:39,760 And I think sometimes as a company you cannot offer all of 163 00:08:39,760 --> 00:08:41,520 it. So you need to really think 164 00:08:41,520 --> 00:08:44,760 about what is, what are you good and what exactly can you offer 165 00:08:44,760 --> 00:08:48,720 to the people within your means. And that is for me always. 166 00:08:48,720 --> 00:08:51,960 The way of retaining a talent in the team is to, at the end of 167 00:08:51,960 --> 00:08:55,760 the day meetings, meeting their expectation and their 168 00:08:55,760 --> 00:08:59,040 motivational factors with what the company can offer. 169 00:08:59,040 --> 00:09:02,160 And if we cannot offer that, then that is really beyond like 170 00:09:02,160 --> 00:09:04,760 if, if you want a certain level of let's say salary, which is 171 00:09:04,760 --> 00:09:07,400 very numerical thing and you can, it's less Gray. 172 00:09:07,680 --> 00:09:10,200 It's like I need X, we cannot offer you X. 173 00:09:10,240 --> 00:09:12,480 That is totally fine. And maybe that's that's a 174 00:09:12,480 --> 00:09:14,320 motivating factor and we cannot retain you. 175 00:09:14,800 --> 00:09:17,680 But from my experience, I think a lot of people really need 176 00:09:18,000 --> 00:09:20,560 motivations from the perspective of they think they are doing 177 00:09:20,560 --> 00:09:25,600 something good for and they are having an impact and, and 178 00:09:25,600 --> 00:09:30,040 creating that area that's within your team that people can have 179 00:09:30,040 --> 00:09:33,440 impacts on the goals, on the environment and the culture. 180 00:09:33,760 --> 00:09:39,000 I think that is where I would suggest to focus on for creating 181 00:09:39,040 --> 00:09:44,040 a good culture to retain talent. I spoke with Dario, I think it 182 00:09:44,040 --> 00:09:46,600 was a couple of weeks ago, he's from Elastic, and he says at 183 00:09:46,600 --> 00:09:49,440 some point I have enough money. Like I earn enough money. 184 00:09:49,440 --> 00:09:52,720 It's no longer about the money, it's about me feeling meaningful 185 00:09:52,720 --> 00:09:54,680 in the work that I do. I was like, that's an 186 00:09:54,680 --> 00:09:57,160 interesting perspective because especially in bigger tech 187 00:09:57,160 --> 00:10:01,280 organizations, everyone earns like down the line, above 188 00:10:01,280 --> 00:10:03,440 average, well above average, very, very well. 189 00:10:03,680 --> 00:10:06,520 So then what distinguishes it? It's like the interesting impact 190 00:10:06,520 --> 00:10:09,080 that you make on customers, or if you're actually contributing 191 00:10:09,080 --> 00:10:12,040 towards business outcomes, or if your project didn't get killed, 192 00:10:12,040 --> 00:10:14,800 or if you actually got very interesting discussions with 193 00:10:14,800 --> 00:10:16,720 your colleagues and there's enough for you to learn and 194 00:10:16,720 --> 00:10:18,600 grow. There's so many aspects. 195 00:10:18,920 --> 00:10:22,520 I really like the emphasis of learning and growing and finding 196 00:10:22,520 --> 00:10:25,440 the balance that works for you. If all of a sudden you get 197 00:10:25,440 --> 00:10:27,960 twins, you don't just have one kid, you have two kids. 198 00:10:27,960 --> 00:10:29,880 So your life might completely be different. 199 00:10:30,280 --> 00:10:33,520 And then this kind of grind that you're on of always kind of 200 00:10:33,520 --> 00:10:36,600 working on career progression that likely gets a little bit on 201 00:10:36,600 --> 00:10:39,520 the back burner and you need a little bit more time for family 202 00:10:39,520 --> 00:10:40,840 and friends. And maybe that's no longer the 203 00:10:40,840 --> 00:10:43,200 right organization or the organization's flexible enough 204 00:10:43,200 --> 00:10:46,560 and says doesn't matter who do you that's the best? 205 00:10:46,720 --> 00:10:48,000 Absolutely. Yeah, absolutely right. 206 00:10:48,240 --> 00:10:51,680 You mentioned yourself, you love newer technologies kind of being 207 00:10:51,680 --> 00:10:54,600 on the forefront of things, which means my assumption is 208 00:10:54,600 --> 00:10:56,720 that especially in the last few years, things have been 209 00:10:56,720 --> 00:11:00,480 interesting for data and AI. How things been for you and 210 00:11:00,480 --> 00:11:02,640 let's start from the personal productivity sense. 211 00:11:02,800 --> 00:11:05,600 Have things changed in how you work or how you approach things? 212 00:11:06,320 --> 00:11:10,000 Definitely. I think I mean, first the whole 213 00:11:10,000 --> 00:11:13,080 world of AI is very, very interesting because constantly 214 00:11:13,440 --> 00:11:15,880 the definition of what is AI is changing. 215 00:11:15,880 --> 00:11:18,720 So I think some of those deep learning models that we were 216 00:11:18,720 --> 00:11:22,800 building even two years ago and now suddenly I can go to into my 217 00:11:22,800 --> 00:11:26,440 own organization and then my own team members and they might not 218 00:11:26,440 --> 00:11:29,800 count that as AI anymore because it's not the generative AI, it's 219 00:11:29,800 --> 00:11:31,840 not the latest and the greatest and you're not doing that. 220 00:11:32,440 --> 00:11:35,280 So I think the definition of what is AI and what is what is 221 00:11:35,280 --> 00:11:38,360 the whole ecosystem is constantly evolving and and I 222 00:11:38,360 --> 00:11:40,760 think it's a very loose term, so. 223 00:11:40,960 --> 00:11:43,560 It's become a loose term. It's become a somehow somehow 224 00:11:43,560 --> 00:11:47,680 very loose term, very overused in certain contexts, underused 225 00:11:47,680 --> 00:11:52,400 in certain contexts. But personally, yes, I, I've 226 00:11:52,400 --> 00:11:57,720 been, I think it has impacted a lot the way that I work, the way 227 00:11:57,720 --> 00:12:00,120 that I do things. And I think there's a lot more 228 00:12:00,120 --> 00:12:05,480 room to actually change because for example, just to give you 229 00:12:05,480 --> 00:12:09,800 some example, I think with this whole ecosystem of lab coding, 230 00:12:09,800 --> 00:12:12,160 that's that's actually really beautiful if you think about it, 231 00:12:12,160 --> 00:12:17,440 because I'm a very visual person and I always need to jump on a 232 00:12:17,440 --> 00:12:21,120 whiteboard and do things. And now in a digital world that 233 00:12:21,120 --> 00:12:24,680 my teammates or team members or stakeholders or whoever that are 234 00:12:24,680 --> 00:12:27,240 in different parts of the world, I always had this difficulty of 235 00:12:27,480 --> 00:12:29,080 showing them exactly what I mean. 236 00:12:29,240 --> 00:12:33,440 So you use mirror, you use whatever and and and but still 237 00:12:33,440 --> 00:12:37,000 kind of was missing. And I think now we have these 238 00:12:37,000 --> 00:12:40,760 wipe coding sessions with, let's say a product manager or a UX 239 00:12:40,760 --> 00:12:42,520 designer. And then that is for me is very 240 00:12:42,520 --> 00:12:44,440 beautiful. That's we can come up with a 241 00:12:44,440 --> 00:12:47,800 proof of concept, build something discarded almost 242 00:12:47,800 --> 00:12:50,320 immediately. But we understood each other 243 00:12:50,320 --> 00:12:53,000 much better in a much deeper level than what it was. 244 00:12:53,000 --> 00:12:56,280 So I think that is one of the thing and I think also the the 245 00:12:56,280 --> 00:13:01,200 whole world of writing codes. I mean, I'm not a software 246 00:13:01,200 --> 00:13:05,080 engineer, so I don't write as many lines of code as maybe a 247 00:13:05,080 --> 00:13:07,720 software engineer does. So maybe they see a lot more 248 00:13:07,840 --> 00:13:11,120 improvement there. But just really recently looking 249 00:13:11,120 --> 00:13:16,640 at clouds code, it's just, it's just insane to think about. 250 00:13:16,640 --> 00:13:20,280 That's that actually the way that you do things has got a lot 251 00:13:20,280 --> 00:13:25,600 faster. Also, I, I do chat with ChatGPT 252 00:13:25,600 --> 00:13:29,640 a lot about work events. So it's a little bit like a 253 00:13:29,640 --> 00:13:34,080 psychological things like, Hey, prep me for this conversation 254 00:13:34,080 --> 00:13:36,600 that I'm having, or I'm having this and this and that. 255 00:13:36,960 --> 00:13:40,040 Can you help me do that? I think I do quite a lot of 256 00:13:40,040 --> 00:13:44,680 these things throughout the day, but I do see every time I go to 257 00:13:44,680 --> 00:13:47,240 a conference or to talk to a meet up or something, I see 258 00:13:47,240 --> 00:13:51,680 people actually have all sort of cool things in their ecosystem 259 00:13:51,680 --> 00:13:53,480 and in their things. So I think there's a lot more 260 00:13:53,480 --> 00:13:57,400 room to figure out exactly what is the productivity tool for me 261 00:13:57,400 --> 00:14:02,280 individually that can make me even more and more productive as 262 00:14:02,280 --> 00:14:05,560 a person. Yeah, I, I haven't vibe coded as 263 00:14:05,560 --> 00:14:06,800 much. Like I feel like I'm late to the 264 00:14:06,800 --> 00:14:10,840 game because I did a year and a half of product and now I got 265 00:14:11,040 --> 00:14:13,200 Claude Coats specifically. And I was like, I have this 266 00:14:13,200 --> 00:14:14,360 idea. I always wanted to make this 267 00:14:14,360 --> 00:14:17,040 game that I can play with friends and I've seen stuff on 268 00:14:17,040 --> 00:14:20,200 YouTube and I, I wonder if I how easy it is to make a game like 269 00:14:20,200 --> 00:14:21,720 that. And I just did it. 270 00:14:21,720 --> 00:14:23,880 It was like 1 evening. It's not done yet. 271 00:14:23,880 --> 00:14:26,280 I still need to continue. But and then I looked at how 272 00:14:26,280 --> 00:14:27,800 much I spend. It was like $3. 273 00:14:27,800 --> 00:14:30,280 And sure, I have this super expensive monthly subscription 274 00:14:30,280 --> 00:14:34,080 plan and I still pay, which is like an insane model, but I am 275 00:14:34,080 --> 00:14:37,520 having a lot of fun with it. I feel like this kind of playing 276 00:14:37,520 --> 00:14:40,200 around and iterating and indeed, like you said, kind of 277 00:14:40,200 --> 00:14:42,680 validating assumptions. I'm also a visual person, so 278 00:14:43,040 --> 00:14:46,160 getting something up and running visually, We have so many tools 279 00:14:46,160 --> 00:14:48,160 now. I'm even to the point where I'm 280 00:14:48,160 --> 00:14:51,360 like overwhelmed with regards to OK, do I try this thing and 281 00:14:51,360 --> 00:14:53,720 which one do I master? And like where do I place my 282 00:14:53,720 --> 00:14:55,600 bets in? What do I think is going to be 283 00:14:55,600 --> 00:14:58,440 there for the long term or what is just a fad and what is more 284 00:14:58,440 --> 00:15:01,520 hype? I don't know yet, but I do like 285 00:15:01,520 --> 00:15:04,320 this experimentation mindset. I feel like it's been a while 286 00:15:04,320 --> 00:15:07,720 since we've had an innovation like this, which has been this 287 00:15:07,720 --> 00:15:10,200 big and kind of, yeah, changes a lot of things. 288 00:15:10,200 --> 00:15:14,000 I feel like, yeah, like cloud was one thing and then a lot of 289 00:15:14,000 --> 00:15:16,080 people were hesitant. And now it's, it's pretty 290 00:15:16,080 --> 00:15:18,160 established. I would say companies are there 291 00:15:18,400 --> 00:15:21,840 or companies are migrating and startups due to the fact of 292 00:15:21,840 --> 00:15:25,560 total cost of ownership, basically they start there, 293 00:15:25,560 --> 00:15:28,280 which is way easier. But with AI, we haven't really 294 00:15:28,280 --> 00:15:31,640 figured out yet or established yet what is there to stay with 295 00:15:31,640 --> 00:15:33,760 regards to productivity and what is there to go in the first 296 00:15:33,760 --> 00:15:34,800 place. Right, right. 297 00:15:34,880 --> 00:15:38,040 Absolutely. I mean it's also true in how you 298 00:15:38,040 --> 00:15:42,040 build models or how you do data science or AI as a whole. 299 00:15:42,160 --> 00:15:45,880 That is also massively changing day-to-day. 300 00:15:45,920 --> 00:15:51,160 I think looking at maybe 3-4 years ago, fibers is still a lot 301 00:15:51,160 --> 00:15:52,880 of a lot happens in predictive AI. 302 00:15:53,320 --> 00:15:58,320 You spend quite a lot of time tuning models, fine tuning 303 00:15:58,320 --> 00:16:04,360 things, building algorithms and doing feature engineering and 304 00:16:04,360 --> 00:16:07,440 all of that sort of things. Now it's kind of constantly 305 00:16:07,440 --> 00:16:08,520 shifted to a different direction. 306 00:16:08,520 --> 00:16:12,120 But still a lot of those abilities that you had, the 307 00:16:12,120 --> 00:16:14,400 learnings you had are actually still true. 308 00:16:14,800 --> 00:16:19,560 And I think, I think that is also also the whole ecosystem 309 00:16:19,560 --> 00:16:21,160 around it. What is data? 310 00:16:21,160 --> 00:16:25,560 What type of data do you need for this type of AI versus how 311 00:16:25,560 --> 00:16:27,360 do you put these models in production? 312 00:16:28,120 --> 00:16:30,240 Is it the same? Is the model artefact the same? 313 00:16:30,240 --> 00:16:32,760 And all of those sort of things are also constantly evolving, 314 00:16:32,760 --> 00:16:35,640 which is very exciting and interesting at the same time to 315 00:16:35,640 --> 00:16:38,920 figure it out. What is the right tool stack to 316 00:16:38,920 --> 00:16:43,240 do AI and data science compared to two years ago or two years 317 00:16:43,240 --> 00:16:44,840 ago? What has been the the biggest 318 00:16:44,840 --> 00:16:47,080 change that you've seen? Because I, I haven't worked with 319 00:16:47,080 --> 00:16:49,640 data science in, let's say, a product development team for a 320 00:16:49,640 --> 00:16:52,600 long time. When I did, I, I was there from 321 00:16:52,600 --> 00:16:54,720 the product perspective. And 1st I saw them kind of 322 00:16:55,400 --> 00:16:58,880 isolated working on models and then kind of testing and proof 323 00:16:58,880 --> 00:17:01,040 of concepts and actually seeing if there's value. 324 00:17:01,520 --> 00:17:04,040 And then once we got to a point where we're super confident, I 325 00:17:04,040 --> 00:17:06,920 saw them going very close to the user side and actually testing 326 00:17:06,920 --> 00:17:09,599 and validating assumptions and seeing if they could deliver the 327 00:17:09,599 --> 00:17:13,560 value as they had theoretically. But then with actual users, I 328 00:17:13,560 --> 00:17:16,800 don't know if that's normal. I think that's very healthy also 329 00:17:16,800 --> 00:17:19,520 from data science to go more towards a user perspective to 330 00:17:19,520 --> 00:17:23,240 see how we make this means to an end actually achieve business 331 00:17:23,240 --> 00:17:26,160 outcomes in the end of the day. I saw them also be involved more 332 00:17:26,160 --> 00:17:28,160 from an infrastructure standpoint on the engineering 333 00:17:28,160 --> 00:17:29,520 side. So I don't know if that's 334 00:17:29,520 --> 00:17:32,000 normal, that's a newer trend, but what have you seen? 335 00:17:32,000 --> 00:17:33,640 How is the data science kind of evolved? 336 00:17:34,240 --> 00:17:38,680 I think the whole ecosystem is constantly shifting between 337 00:17:38,680 --> 00:17:44,000 predictive AI and generative AI. And the main thing is that it 338 00:17:44,000 --> 00:17:49,120 has been in a way democratized in how people can interact with 339 00:17:49,120 --> 00:17:53,840 models. And AII think after ChatGPT, the 340 00:17:53,880 --> 00:17:58,960 main change in the society or in the tech world has been now you 341 00:17:58,960 --> 00:18:03,880 suddenly could interact with an AI that sounded OK, goods are 342 00:18:03,880 --> 00:18:07,960 very well or convincing. And I think that opened a lot of 343 00:18:07,960 --> 00:18:10,040 new doors. But at the same time, it created 344 00:18:10,040 --> 00:18:12,240 also a new shift in how you do things. 345 00:18:12,240 --> 00:18:18,600 For example, building a proof of concept going from zero to 80% 346 00:18:18,880 --> 00:18:23,160 has gotten really easy. It's, it's like almost like you 347 00:18:23,160 --> 00:18:27,560 could even simplify it as much of we're going to call an API, 348 00:18:27,760 --> 00:18:31,840 write the prompt and it's put something wrapping around it and 349 00:18:31,840 --> 00:18:34,920 calling it an application. Or I think now we're talking all 350 00:18:34,920 --> 00:18:37,760 about agents calling an agent, like an agent calling a few 351 00:18:37,760 --> 00:18:41,240 different AI models in a loop and then we call it an agent. 352 00:18:41,320 --> 00:18:46,040 I think that level of creating proof of concepts has increased 353 00:18:46,080 --> 00:18:48,800 a lot, which basically makes the noise a lot. 354 00:18:49,120 --> 00:18:53,240 But still doing that last 20% is still equally as hard in my 355 00:18:53,240 --> 00:18:57,440 opinion. For example, now we are 356 00:18:57,440 --> 00:19:02,680 investing a lot more time and efforts in in how our data are. 357 00:19:03,160 --> 00:19:06,360 Of course, data has been always the cornerstone, corner 358 00:19:06,360 --> 00:19:10,760 cornerstone of AI and organizing it has been always a thing. 359 00:19:11,400 --> 00:19:14,040 But now you need to look into, do you have the right 360 00:19:14,040 --> 00:19:16,560 ontologies? Do you have the right knowledge 361 00:19:16,560 --> 00:19:19,040 graph in your company? Do you have the right semantic 362 00:19:19,040 --> 00:19:22,240 data layers? And all of these, I think do you 363 00:19:22,240 --> 00:19:24,360 have the right embeddings? Do you have all that sort of 364 00:19:24,360 --> 00:19:28,120 things to do things? And, and that has been 1 main 365 00:19:28,120 --> 00:19:32,040 change in my opinion. That's how we look at data will 366 00:19:32,040 --> 00:19:35,760 evolve a lot more and companies need to figure out what is their 367 00:19:35,760 --> 00:19:38,320 data and is their data foundations ready for the 368 00:19:38,320 --> 00:19:41,640 generative AI world. And then the the second thing is 369 00:19:41,640 --> 00:19:45,800 how you put models in production has changed significantly, of 370 00:19:45,800 --> 00:19:48,280 course, in the basis of things still the same. 371 00:19:48,280 --> 00:19:50,840 But one question that we constantly ask ourselves, like 372 00:19:51,120 --> 00:19:56,600 on top of the cloud provider AI tools like Vertex AI in GCP or 373 00:19:56,720 --> 00:19:59,720 AI Foundry in Azure and all of that sort of things, what else 374 00:19:59,720 --> 00:20:01,640 do you need? For example, we used to have or 375 00:20:01,640 --> 00:20:04,640 we still have wait time biases as a tool that we track 376 00:20:04,640 --> 00:20:06,920 experiments and we track different model runs and 377 00:20:06,920 --> 00:20:09,160 etcetera. Is rates and biases still the 378 00:20:09,160 --> 00:20:12,720 right tool for doing that or do we need a new tool on top of it 379 00:20:12,720 --> 00:20:16,800 to do LLM OPS, for example? And also what are the model 380 00:20:16,800 --> 00:20:19,840 artefacts? How do you evaluate the models 381 00:20:19,840 --> 00:20:24,840 as significantly changed in now you need to structure your 382 00:20:24,840 --> 00:20:26,520 ground truths in a completely different way. 383 00:20:26,520 --> 00:20:32,120 So I think and also people wise around the company or around 384 00:20:32,120 --> 00:20:35,480 just think one thing is you building the models, but also 385 00:20:35,480 --> 00:20:37,400 the experiences for the customers, exactly what you were 386 00:20:37,400 --> 00:20:40,280 mentioning. Everyone else, like if an 387 00:20:40,280 --> 00:20:43,360 experienced designer, UX designer, service designer, a 388 00:20:43,360 --> 00:20:47,440 product manager needs to also understand what the generative 389 00:20:47,520 --> 00:20:52,480 AI is and how is it different than predictive AI And, and how 390 00:20:52,480 --> 00:20:55,320 do you build an experience for this new world that things can 391 00:20:55,320 --> 00:20:58,800 be generated on demand? Do we do you would you look at 392 00:20:59,320 --> 00:21:02,480 an e-commerce website the same way if you knew everything can 393 00:21:02,480 --> 00:21:05,280 be generated in a in a perfect world. 394 00:21:06,000 --> 00:21:08,480 So how would you look that? I think these are all different 395 00:21:08,480 --> 00:21:13,040 changes that are coming and we have we don't have very good 396 00:21:13,040 --> 00:21:16,080 answers for all of them yet. And I think it takes some time 397 00:21:16,080 --> 00:21:21,240 to to just really figure out what is the right tool to do it 398 00:21:21,240 --> 00:21:24,760 and how to do it and and where is the place of predictive AI 399 00:21:26,040 --> 00:21:30,000 has versus generative. AI for the specifically, let's 400 00:21:30,000 --> 00:21:32,760 talk about generative AI. Things like how your ground 401 00:21:32,760 --> 00:21:35,560 truth has changed and model optimizations or even the 402 00:21:35,560 --> 00:21:38,720 observability in production. Where do you see those 403 00:21:38,720 --> 00:21:41,200 responsibilities lie then? Is that data science? 404 00:21:41,200 --> 00:21:43,160 Is it more data engineering? Is it still software 405 00:21:43,160 --> 00:21:44,960 engineering? I feel like it's a good 406 00:21:44,960 --> 00:21:47,960 combination, but I don't know in the end who's going to be core 407 00:21:47,960 --> 00:21:51,160 responsible for what yet. I think that's a really good 408 00:21:51,160 --> 00:21:55,320 question, and I think we are trying to put all of these in 409 00:21:55,320 --> 00:22:00,960 the boxes that's already exists. But maybe the answer is that 410 00:22:00,960 --> 00:22:04,400 none of these are the right people, the right people to do 411 00:22:04,400 --> 00:22:06,400 that. And maybe indeed it's a 412 00:22:06,400 --> 00:22:10,320 combination of all of that. There will be always a place for 413 00:22:11,200 --> 00:22:14,840 doing actual data science. For example, building an actual 414 00:22:14,840 --> 00:22:22,160 rack is not as easy as putting some documents in a in a vector 415 00:22:22,160 --> 00:22:26,520 database, do a a search in it and say, hey, tada, we have a 416 00:22:26,520 --> 00:22:28,120 rag. It's really not that easy. 417 00:22:28,120 --> 00:22:30,960 You really need to spend quite a lot of time. 418 00:22:31,080 --> 00:22:33,520 It's a search problem at the end of the day, and you really need 419 00:22:33,520 --> 00:22:35,360 to optimize that as a search problem. 420 00:22:35,840 --> 00:22:37,400 Understand the whole space very well. 421 00:22:37,400 --> 00:22:40,360 So there is a space still for optimizing that type of thing 422 00:22:40,480 --> 00:22:43,960 within any organization or any company, small or big. 423 00:22:44,480 --> 00:22:48,840 And then there is and then there is also all these newer things 424 00:22:48,840 --> 00:22:51,720 that are coming up. The way that you keep the 425 00:22:51,720 --> 00:22:54,880 infrastructure together. It doesn't require only software 426 00:22:54,880 --> 00:22:56,640 engineering. It also requires you to learn a 427 00:22:56,640 --> 00:22:59,880 little bit about what is this generative AI model does and 428 00:22:59,880 --> 00:23:01,480 etcetera. It requires you to do a little 429 00:23:01,480 --> 00:23:03,640 bit data engineering and all of that. 430 00:23:04,280 --> 00:23:09,320 So I think in the future there will be new, we can call them AI 431 00:23:09,320 --> 00:23:12,120 engineers. That's how we can call them very 432 00:23:12,120 --> 00:23:15,560 easy solving the problem to to call them AI engineer that needs 433 00:23:15,560 --> 00:23:19,320 to to have different skill sets from different crafts. 434 00:23:19,880 --> 00:23:24,280 If you are very, very specialist in this, I think majority of the 435 00:23:24,280 --> 00:23:27,120 skills are not going to be super duper specialist and they're 436 00:23:27,120 --> 00:23:31,240 going to be more and more skill sets that needs to come together 437 00:23:32,160 --> 00:23:36,160 to to actually deliver products in the new in the new age of 438 00:23:36,160 --> 00:23:38,040 tech. I like that perspective a lot 439 00:23:38,040 --> 00:23:40,400 and I'm wondering how you then support people that are 440 00:23:40,800 --> 00:23:43,080 interested in this from their own domain. 441 00:23:43,080 --> 00:23:45,000 Right? Because if I, if I talk to RAG 442 00:23:45,000 --> 00:23:47,600 to my colleagues, software engineers, they go search 443 00:23:47,600 --> 00:23:51,600 problem software engineering. But then in organizations, data 444 00:23:51,600 --> 00:23:54,400 science, people are working with generative AI and trying to get 445 00:23:54,400 --> 00:23:57,800 that up and to production. And then they see their model 446 00:23:57,800 --> 00:23:59,320 performance will increase if they do RAG. 447 00:23:59,320 --> 00:24:01,760 So then they try and figure out how to implement RAG, even 448 00:24:01,760 --> 00:24:03,000 though it might be a search problem. 449 00:24:03,360 --> 00:24:06,520 It's like everyone that is interested will contribute and 450 00:24:06,520 --> 00:24:09,040 will help. And is that also what you foster 451 00:24:09,040 --> 00:24:10,480 with regards to people's personal growth? 452 00:24:10,480 --> 00:24:12,760 Like if they're interested, can they do kind of the whole chain 453 00:24:12,760 --> 00:24:16,520 of things or do you still like see what their expertise is and 454 00:24:16,520 --> 00:24:21,120 what they can add most value? I always suggest to and I don't 455 00:24:21,120 --> 00:24:23,560 know if this is the exact right suggestion, though, I think I 456 00:24:23,600 --> 00:24:29,200 always suggest to explore a sister capability to your own 457 00:24:29,200 --> 00:24:32,320 main craft. So if your craft is a software 458 00:24:32,320 --> 00:24:36,560 engineer and there's a sister capability of understanding and 459 00:24:36,560 --> 00:24:39,280 learning certain type of generative AI models, 460 00:24:39,360 --> 00:24:42,600 understanding them a little bit deeper than just calling ChatGPT 461 00:24:42,600 --> 00:24:46,000 and calling an API and etcetera, to really understand how to 462 00:24:46,000 --> 00:24:47,640 write a really, really good prompts. 463 00:24:48,040 --> 00:24:50,000 Maybe in the world we don't need to write prompts anymore. 464 00:24:50,000 --> 00:24:53,960 But understanding all these sister capabilities is what I 465 00:24:53,960 --> 00:24:56,720 suggest to to people. What if it is getting like so 466 00:24:56,720 --> 00:24:58,760 far away? And I think and, and it really 467 00:24:58,760 --> 00:25:01,400 distracts you from doing your main tasks. 468 00:25:01,400 --> 00:25:04,840 For example, if you have to maintain certain infrastructure, 469 00:25:04,840 --> 00:25:07,680 for example, now you are you to put certain models in 470 00:25:07,680 --> 00:25:11,480 production, you don't want to use certain cloud capabilities. 471 00:25:11,480 --> 00:25:14,840 You want to use, I don't know, run a Kubernetes cluster to do 472 00:25:14,840 --> 00:25:16,960 certain things. And then it requires you to 473 00:25:16,960 --> 00:25:19,360 learn a completely different sets of skills and it's going to 474 00:25:19,360 --> 00:25:22,280 take some time and etcetera. Then I'm like, OK, maybe ease 475 00:25:22,280 --> 00:25:25,240 into it. So you don't need to go all the 476 00:25:25,240 --> 00:25:28,160 way till the end. So that is how my recommendation 477 00:25:28,160 --> 00:25:33,080 is into learning these different capabilities together. 478 00:25:33,320 --> 00:25:34,960 Gotcha. Yeah, I like that a lot. 479 00:25:34,960 --> 00:25:38,360 Kind of adjacent technologies that are easier to hop in and 480 00:25:38,360 --> 00:25:40,920 out of and you can kind of leverage the craft that you have 481 00:25:40,920 --> 00:25:42,320 already. Yeah. 482 00:25:42,680 --> 00:25:47,360 You mentioned earlier that AI is kind of gotten fluffy because 483 00:25:47,360 --> 00:25:50,360 everything is now AI, everything and nothing at the same time are 484 00:25:50,360 --> 00:25:53,200 only specific things are now all of a sudden AI and some people 485 00:25:53,200 --> 00:25:55,400 don't understand the difference between generative and 486 00:25:55,400 --> 00:25:57,200 predictive AI. I'm assuming you have 487 00:25:57,200 --> 00:25:59,160 conversations with that are technical but also 488 00:25:59,160 --> 00:26:02,160 non-technical. How important is it to 489 00:26:02,160 --> 00:26:06,720 understand and be very specific in a language perspective with 490 00:26:06,720 --> 00:26:10,840 what we talk about nowadays? I've been on two different 491 00:26:10,840 --> 00:26:13,440 spectrums of this over the courses of my career. 492 00:26:13,440 --> 00:26:18,040 I think earlier on in my career, I used to spend a lot of time 493 00:26:18,160 --> 00:26:21,440 correcting people and saying, hey, this is not like that. 494 00:26:21,440 --> 00:26:25,120 This is like that. And and I think it is not always 495 00:26:25,120 --> 00:26:30,760 very helpful to to do that. And on the other hand, now I'm a 496 00:26:30,760 --> 00:26:34,120 lot more focused on bringing business values. 497 00:26:34,120 --> 00:26:38,840 So I don't really care as much if person X uses a certain 498 00:26:38,840 --> 00:26:43,680 language to begin with. But I do think the it's, it's a 499 00:26:43,720 --> 00:26:46,640 it's not my responsibility always to correct people, but I 500 00:26:46,640 --> 00:26:50,640 do think it's a responsibility of the organizations to teach 501 00:26:50,840 --> 00:26:56,160 people the languages, the risks, the ethical side of things and 502 00:26:56,160 --> 00:26:58,240 all of that. Things are sometime you can get 503 00:26:58,240 --> 00:27:01,200 this through a conversation and some once the person trusts you, 504 00:27:01,200 --> 00:27:03,840 they of course listen to that in the same way that I learned a 505 00:27:03,840 --> 00:27:08,080 lot about probably I've already in this podcast, I used the word 506 00:27:08,200 --> 00:27:10,440 experience design, UX UI already wrong. 507 00:27:10,680 --> 00:27:12,680 And I'm sure if a colleague is listening is like, no, this is 508 00:27:12,680 --> 00:27:15,440 not that and that, and I think it's all in learning. 509 00:27:15,440 --> 00:27:19,040 So I think it's just going with an open mind on both side and, 510 00:27:19,040 --> 00:27:22,600 and to, to really learn each other's languages is always 511 00:27:22,600 --> 00:27:26,480 really helpful because a lot of our work is just communicating 512 00:27:26,480 --> 00:27:28,160 with each other. And the better you can 513 00:27:28,160 --> 00:27:31,920 communicate with each other, then the better you can push the 514 00:27:31,920 --> 00:27:35,840 products forward. So, but I do think personally, 515 00:27:36,200 --> 00:27:39,720 understanding the differences in the tech is very, very important 516 00:27:39,720 --> 00:27:43,600 because you need to know what's the tech can offer you. 517 00:27:44,240 --> 00:27:47,960 And then different people need to think about how it can create 518 00:27:47,960 --> 00:27:50,760 a business value. Because if you do it fully tech 519 00:27:50,760 --> 00:27:54,120 driven business value, you might end up on one spectrum of 520 00:27:54,120 --> 00:27:55,840 things. If you do completely, let's say 521 00:27:55,840 --> 00:28:00,880 based on on only business and neglecting what tech can do for 522 00:28:00,880 --> 00:28:03,760 you, then you end up in a different zone. 523 00:28:03,760 --> 00:28:07,480 So I think a balance is, is very important unless you're Google 524 00:28:07,480 --> 00:28:10,280 and you're really good at the tech that you do and tech 525 00:28:10,280 --> 00:28:13,120 defines everything and then you be Google. 526 00:28:14,600 --> 00:28:18,040 I mean, even Google, I feel like they're so tech driven that 527 00:28:18,040 --> 00:28:21,520 sometimes the marketing of whatever cool they have is 528 00:28:21,520 --> 00:28:23,600 sometimes a little bit lacking where you don't really hear 529 00:28:23,600 --> 00:28:26,760 about this or it's like a ninja release and then no one actually 530 00:28:26,760 --> 00:28:28,240 sees it. It doesn't have the same hype 531 00:28:28,640 --> 00:28:31,800 like an open AI is now trying to do, which is kind of everywhere. 532 00:28:32,080 --> 00:28:33,320 They're also at the forefront of it. 533 00:28:33,840 --> 00:28:36,720 But you mentioned business outcomes with generative AI, if 534 00:28:36,720 --> 00:28:40,400 you like, personal productivity is changing, but so is kind of 535 00:28:40,400 --> 00:28:43,440 what we can offer our customers nowadays, right? 536 00:28:43,440 --> 00:28:46,920 If you've seen kind of this predictive AI, or say I should 537 00:28:46,920 --> 00:28:50,440 say generative AI, really leverage the technology to then 538 00:28:50,440 --> 00:28:51,760 have really cool business outcomes. 539 00:28:51,760 --> 00:28:52,960 Do you have a few examples there? 540 00:28:53,560 --> 00:28:56,440 Yeah, we've got, I mean we've been on this journey of finding 541 00:28:56,440 --> 00:29:01,200 out where generative AI brings outcomes and where it brings 542 00:29:01,200 --> 00:29:05,320 business values. I think you in general you 543 00:29:05,320 --> 00:29:09,880 should look at your the value chain of whatever team or idea 544 00:29:09,880 --> 00:29:14,240 that you're looking at and focus on a few areas that brings 545 00:29:14,240 --> 00:29:16,400 business outcome. I think from my experience so 546 00:29:16,400 --> 00:29:19,840 far, the areas that I can see there has been tangible business 547 00:29:19,840 --> 00:29:24,480 outcome 1 is actually the whole idea of search because because 548 00:29:24,480 --> 00:29:29,120 now people can go beyond keyword search and have a lot more 549 00:29:29,120 --> 00:29:32,840 semantic way of explaining what they want and how they want. 550 00:29:32,840 --> 00:29:36,600 And, and even go to the other end of it, which is 551 00:29:36,600 --> 00:29:40,200 conversational search, which you can actually have a conversation 552 00:29:40,200 --> 00:29:42,320 till you figure out what you're exactly looking for. 553 00:29:43,040 --> 00:29:45,480 Some people are really good at describing what they they have. 554 00:29:45,480 --> 00:29:50,200 And also when you think about search, it's a multi modal thing 555 00:29:50,200 --> 00:29:51,800 as well. Sometimes you search with an 556 00:29:51,800 --> 00:29:55,240 image because you don't know actually how to describe what 557 00:29:55,240 --> 00:29:57,040 you want. You're looking for a wife you're 558 00:29:57,040 --> 00:30:00,400 looking for. And I think generative AI can 559 00:30:00,400 --> 00:30:03,440 really help with product discovery as a whole. 560 00:30:04,080 --> 00:30:08,520 So I think that's one area that within IKEA where I work right 561 00:30:08,520 --> 00:30:11,720 now, I see quite a lot of value being generated and we are 562 00:30:11,720 --> 00:30:13,520 experimenting with it. But still there is quite a lot 563 00:30:13,520 --> 00:30:19,600 of like pros and cons of the cost versus the infrastructure 564 00:30:19,600 --> 00:30:22,800 versus everything. How do you maintain it in a good 565 00:30:22,800 --> 00:30:26,640 way and in a in a very stable way? 566 00:30:27,080 --> 00:30:29,400 Are the costs high usually when you're talking about kind of the 567 00:30:29,400 --> 00:30:32,640 search advancements of conversational search or multi 568 00:30:32,640 --> 00:30:35,200 model give me a vibe similar to this. 569 00:30:35,520 --> 00:30:41,240 Are the costs high in general? It could be high, yes, but you, 570 00:30:41,480 --> 00:30:43,640 I think for me, I personally do not. 571 00:30:43,680 --> 00:30:46,720 Maybe I do not optimize always at this stage. 572 00:30:46,720 --> 00:30:50,520 The first step on a cost for me is figuring out what the 573 00:30:50,520 --> 00:30:54,840 customer wants goes beyond, of course, within reasons. 574 00:30:54,840 --> 00:30:57,720 If something costs like completely absurd and it's like 575 00:30:57,720 --> 00:31:01,960 way out of line and then the whole ROI needs 20 years to to 576 00:31:01,960 --> 00:31:05,320 break even, Yes, of course not. But I think you could do so much 577 00:31:05,320 --> 00:31:08,800 with training, more specialised model training with more open 578 00:31:08,800 --> 00:31:12,880 source models, quantised models, reduce the size of these models 579 00:31:12,880 --> 00:31:14,160 and etcetera. Because at the end of the day, 580 00:31:14,160 --> 00:31:17,680 you're working a very specialised area in a company. 581 00:31:17,680 --> 00:31:23,240 You're not a world knowledge on. You're not you're not offering 582 00:31:23,240 --> 00:31:25,440 world knowledge to people in a lot of companies. 583 00:31:26,760 --> 00:31:29,600 So I think there's a lot of ways of optimizing for the cost, but 584 00:31:29,600 --> 00:31:31,960 I wouldn't start with the cost. I would just figure out what the 585 00:31:31,960 --> 00:31:33,960 user wants. Is this a experience that they 586 00:31:33,960 --> 00:31:37,320 want? But yeah, the cost can be high 587 00:31:37,360 --> 00:31:40,160 at the start. And then you start looking for 588 00:31:40,160 --> 00:31:42,640 ways of optimizing that further and further. 589 00:31:42,760 --> 00:31:46,520 You don't have to rule it out on everyone until you optimize the 590 00:31:46,520 --> 00:31:49,080 cost. For example, what if you know 591 00:31:49,080 --> 00:31:51,640 this is what your user wants? Then it's different. 592 00:31:51,840 --> 00:31:55,440 I feel like knowing what the user wants has always been 593 00:31:55,440 --> 00:31:58,640 challenging but also users have now been so educated with 594 00:31:58,640 --> 00:32:02,440 regards to OK I go to an e-commerce and I know what I 595 00:32:02,440 --> 00:32:05,080 need to search on but I know also how to use the search bar 596 00:32:05,080 --> 00:32:06,720 for example. So they automatically will do 597 00:32:06,720 --> 00:32:08,680 keywords. They will not describe in 598 00:32:08,680 --> 00:32:11,000 linguistics what they're searching for even though it's 599 00:32:11,000 --> 00:32:14,920 now possible or even the ability to be like I want a vibe similar 600 00:32:14,920 --> 00:32:17,680 to this living room. If your IKEA sounds like a super 601 00:32:17,680 --> 00:32:20,480 cool use case to then offer something that is similar. 602 00:32:20,920 --> 00:32:22,160 But how do you educate a customer? 603 00:32:22,160 --> 00:32:23,880 Like take them with you? Or how do you know if this is 604 00:32:23,880 --> 00:32:25,120 actually a problem you're solving? 605 00:32:25,360 --> 00:32:27,160 I feel it has become even more challenging. 606 00:32:27,680 --> 00:32:32,720 Definitely it's challenging, but it is changing, I think how 607 00:32:32,720 --> 00:32:37,760 people even right now use Google versus ChatGPT and everything. 608 00:32:37,760 --> 00:32:40,200 I often hear, I was like, hey, I've searched this on ChatGPT 609 00:32:40,200 --> 00:32:43,880 and, and, and then you don't want to argue with a person. 610 00:32:43,880 --> 00:32:45,200 Is this the right way of searching? 611 00:32:45,240 --> 00:32:46,560 For. For a thing because. 612 00:32:46,760 --> 00:32:49,120 If it works, it works is. This the most sustainable way of 613 00:32:50,000 --> 00:32:54,560 finding the answer to your questions, but at But at the end 614 00:32:54,560 --> 00:32:57,560 of the day, if you're building something that users truly love 615 00:32:57,800 --> 00:33:02,360 and it's exciting, I think user behaviour slowly and organically 616 00:33:02,640 --> 00:33:03,880 changes. So there there is this whole 617 00:33:03,880 --> 00:33:06,760 trend of change, the user behaviour that changes 618 00:33:07,040 --> 00:33:10,560 automatically by itself. Like right now you cannot really 619 00:33:10,560 --> 00:33:12,920 think of a brand not having a website. 620 00:33:12,920 --> 00:33:15,320 So that's not a competitive advantage anymore. 621 00:33:15,320 --> 00:33:17,760 It's just a given. And I think it's the same way 622 00:33:17,760 --> 00:33:19,680 with this. It's like if you are giving a 623 00:33:19,680 --> 00:33:21,760 really, really good experience and you're figuring it out 624 00:33:21,760 --> 00:33:26,840 exactly what it is you're getting into, into that space. 625 00:33:26,840 --> 00:33:29,360 But at the end of the day, we don't know the answers to these 626 00:33:29,360 --> 00:33:32,920 questions and it's all bets, but you need to put bets on things 627 00:33:32,920 --> 00:33:37,240 that are that you are secure about, that you're like, OK, I 628 00:33:37,240 --> 00:33:40,800 have a hunch as a business, this is where we want to go and this 629 00:33:40,800 --> 00:33:43,840 is what we want to be as our competitive advantage. 630 00:33:44,520 --> 00:33:47,800 Otherwise you are building the same thing as everyone else and 631 00:33:47,800 --> 00:33:51,120 you're just following and you're not never getting the best out 632 00:33:51,120 --> 00:33:56,080 of the tech, you're just following and and you're not 633 00:33:56,080 --> 00:33:59,000 differentiating yourself. And I think that's my core 634 00:33:59,000 --> 00:34:02,160 philosophy. That's everyone in this current 635 00:34:02,960 --> 00:34:05,720 space. While it's possible you should 636 00:34:05,720 --> 00:34:09,360 start differentiating yourself maybe in five years it's not 637 00:34:09,360 --> 00:34:12,120 possible anymore because everything's taken our new 638 00:34:12,120 --> 00:34:15,920 experiences on maybe the whole world of e-commerce website does 639 00:34:15,920 --> 00:34:19,080 not even exist and now we are only shopping through a 640 00:34:19,080 --> 00:34:21,000 completely different ecosystem and etcetera. 641 00:34:21,000 --> 00:34:23,719 I think now that it's possible, you need to, as a business, 642 00:34:24,000 --> 00:34:27,719 figure out what makes you special and and bets on those 643 00:34:27,760 --> 00:34:30,320 very specific capabilities that makes you special. 644 00:34:30,960 --> 00:34:32,400 Yeah, I think now is the perfect time. 645 00:34:32,480 --> 00:34:35,199 Like there's whenever there's a new technology, there's always 646 00:34:35,199 --> 00:34:37,400 something, but it might not be relevant for a lot of people. 647 00:34:37,840 --> 00:34:41,520 And now because it changes like one of the core components like 648 00:34:41,520 --> 00:34:45,440 you mentioned search, search all of a sudden becomes more than 649 00:34:45,440 --> 00:34:48,320 what we used to have before. You can leverage that in 650 00:34:48,320 --> 00:34:50,960 whatever you're trying to do and help the customer experience. 651 00:34:51,520 --> 00:34:53,960 Search was a a big component where you said, OK, that's where 652 00:34:53,960 --> 00:34:55,239 it aligns with business outcomes. 653 00:34:55,239 --> 00:34:56,880 Have you seen any other? Yeah, absolutely. 654 00:34:56,920 --> 00:35:01,400 I think there are a couple more. In general, the search was a 655 00:35:01,400 --> 00:35:05,960 classic one that you already exists, so it's relatively easy 656 00:35:06,480 --> 00:35:08,720 to improve. But then there are also new 657 00:35:08,720 --> 00:35:10,680 experiences that you can build for customers. 658 00:35:10,680 --> 00:35:15,440 And for example, one in the whole world of generating 659 00:35:15,440 --> 00:35:19,920 content, it's a, it's a very specific and interesting area 660 00:35:20,240 --> 00:35:21,840 and you can look at it from different angles. 661 00:35:22,160 --> 00:35:25,000 Maybe you're just generating everything from scratch using 662 00:35:25,000 --> 00:35:27,120 generative AI, which I don't think is the right approach 663 00:35:27,120 --> 00:35:32,400 because then your whole website is filled with AI images and 664 00:35:32,400 --> 00:35:33,720 there's nothing unique around it. 665 00:35:33,720 --> 00:35:36,280 But then there's also something about looking at content as a 666 00:35:36,280 --> 00:35:39,080 whole and gap inspiration for the customer. 667 00:35:39,080 --> 00:35:44,720 For example, a lot of customers have trouble getting inspired. 668 00:35:44,720 --> 00:35:46,360 What do they need? How do they need it? 669 00:35:46,360 --> 00:35:48,360 Where do they need it? When do they need it? 670 00:35:48,360 --> 00:35:50,960 So now we can get a lot more deeper into the context of the 671 00:35:50,960 --> 00:35:55,960 customer and understanding from that angle and gap, this 672 00:35:55,960 --> 00:35:59,920 imagination, imagination gap for them and feel it's in a way that 673 00:35:59,920 --> 00:36:05,080 they can get really, really better recommendation for their 674 00:36:05,080 --> 00:36:08,800 home in in this context of IKEA or, or if you're in different 675 00:36:08,800 --> 00:36:14,280 contexts of what is exactly a deep way of personalizing your 676 00:36:14,280 --> 00:36:18,640 offering to the customer beyond how we do it today, which is 677 00:36:18,640 --> 00:36:21,560 like a carousels of things, sending emails and etcetera. 678 00:36:21,560 --> 00:36:24,120 Beyond that, what can we do to really, really personalise and 679 00:36:24,120 --> 00:36:26,000 give a new experience to the customer? 680 00:36:26,000 --> 00:36:29,840 So I think in this area, we we are experimenting quite a lot 681 00:36:29,840 --> 00:36:33,240 with the 3D World and etcetera and see what can be done in that 682 00:36:33,720 --> 00:36:37,560 area. Then then other areas that I've 683 00:36:37,560 --> 00:36:41,840 seen value is actually a lot of a lot of Co workers because 684 00:36:41,840 --> 00:36:44,080 those are also customers in a big organization. 685 00:36:44,440 --> 00:36:47,480 You're dealing with a lot of a ton of internal toolings, people 686 00:36:47,600 --> 00:36:51,360 working in different places. And we still see value in, for 687 00:36:51,360 --> 00:36:54,520 example, how you do supply chain and how do you optimize that in 688 00:36:54,520 --> 00:36:57,840 a good way. But that's less of a generative 689 00:36:57,920 --> 00:37:01,440 AI, more of a predictive AI, but it could move into a generative 690 00:37:02,120 --> 00:37:07,120 AI world or, or for example, now we have a higher tolerance of 691 00:37:07,480 --> 00:37:14,360 unstructured data, meaning now you can understand images very 692 00:37:14,360 --> 00:37:16,400 well. So how do you manage your 693 00:37:16,400 --> 00:37:18,280 contents? How do you create metadata 694 00:37:18,640 --> 00:37:21,240 around them with a good human feedback loop? 695 00:37:22,000 --> 00:37:25,160 I think all of those are also another area that I can see 696 00:37:25,160 --> 00:37:28,800 already creating value, but there's a ton to be seen to be 697 00:37:28,800 --> 00:37:35,000 done, to be proven value. So, so either you could optimize 698 00:37:35,000 --> 00:37:38,080 existing features with generative AR to give a much 699 00:37:38,080 --> 00:37:43,560 deeper personalization or completely create a competitive 700 00:37:43,560 --> 00:37:47,040 advantage, which I think is is much harder and but it's also a 701 00:37:47,040 --> 00:37:48,720 lot more exciting. Yeah. 702 00:37:49,120 --> 00:37:52,880 How do you then get your priorities in order in terms of 703 00:37:53,000 --> 00:37:57,400 what do we bet on 1st and how do we validate assumptions and do 704 00:37:57,400 --> 00:38:01,720 we go for the more longer term big impact ones or do we do that 705 00:38:01,720 --> 00:38:03,520 in parallel with a few smaller ones? 706 00:38:03,960 --> 00:38:05,800 What do you choose to experiment with first? 707 00:38:06,520 --> 00:38:11,800 In general, we or my team, my strategy always is A2 fold 708 00:38:11,800 --> 00:38:16,160 strategy as I call it. And I, it doesn't always work as 709 00:38:16,680 --> 00:38:21,320 clean as I'm explaining it, but the one fold is you do want 710 00:38:21,320 --> 00:38:24,440 short term value because businesses are optimised on 711 00:38:24,440 --> 00:38:27,960 short term value. If I, if I explain for someone, 712 00:38:27,960 --> 00:38:31,280 hey, within three years I'm going to bring you XYZ. 713 00:38:32,480 --> 00:38:37,240 I think nobody's going to give me money to bring this vision to 714 00:38:37,240 --> 00:38:40,200 life because this is just, it's just too absurd. 715 00:38:40,200 --> 00:38:42,160 They've never seen it and they have to imagine it. 716 00:38:42,160 --> 00:38:45,680 Now I have to explain and or they have a high trust on me and 717 00:38:45,680 --> 00:38:46,880 they're like, OK, yeah, go and do it. 718 00:38:46,880 --> 00:38:51,480 But often time businesses are optimized on short term values 719 00:38:51,480 --> 00:38:54,120 because they're you need to solve the problems here and now. 720 00:38:55,360 --> 00:38:59,360 So I think and in short term value, you also need to look at 721 00:38:59,680 --> 00:39:03,960 are you filling a gap that should not exist in the 1st 722 00:39:03,960 --> 00:39:06,120 place because something was broken somewhere else? 723 00:39:06,120 --> 00:39:08,960 Now you're feeling it. So that that is something I 724 00:39:08,960 --> 00:39:12,600 would not call a short term gain, but shorter gain means 725 00:39:12,600 --> 00:39:16,000 like for example, in the search, it's like it's, it's given, it's 726 00:39:16,000 --> 00:39:18,720 like it's there. People are searching, they are 727 00:39:18,720 --> 00:39:20,920 searching longer, they're already going to ChatGPT 728 00:39:20,920 --> 00:39:23,480 searching for your products. So you better offer a better 729 00:39:23,480 --> 00:39:26,000 experience. So then there are some some 730 00:39:26,000 --> 00:39:29,040 other things which are mid term and long term which you need to 731 00:39:29,040 --> 00:39:32,760 really understand what is your core business And there is like 732 00:39:32,760 --> 00:39:37,240 OK in that area of my core business, what is it that brings 733 00:39:37,240 --> 00:39:40,080 me value. And I think there is a bit of a 734 00:39:40,440 --> 00:39:45,000 luck plus a business hunch, plus the tech all needs to meet each 735 00:39:45,000 --> 00:39:50,680 other to actually bring value. And, and those are the areas 736 00:39:50,680 --> 00:39:54,760 that you want to spend a little bit time longer term 737 00:39:54,960 --> 00:39:57,920 understanding the take around it and and etcetera. 738 00:39:57,920 --> 00:40:01,840 So I think that is how I usually prioritize and say, OK, does 739 00:40:01,840 --> 00:40:05,520 this and I put 60% of the effort usually on short term, 40% on a 740 00:40:05,520 --> 00:40:08,960 little bit medium, long term and say, OK, is this bringing me 741 00:40:09,000 --> 00:40:12,600 value in next 2-3 weeks, a month, a year? 742 00:40:13,280 --> 00:40:17,360 What is the the thing? And then I do from there. 743 00:40:17,360 --> 00:40:22,520 But then also this culture of. Experimentation is also really 744 00:40:22,520 --> 00:40:27,520 important within IKEA. We spend 20% of our time on just 745 00:40:27,520 --> 00:40:31,360 experimenting, forsake of experimentation and and. 746 00:40:31,360 --> 00:40:34,520 Everyone. Everyone can do it, but not 747 00:40:34,520 --> 00:40:37,200 everyone does it because it's also hard to next to your job. 748 00:40:37,200 --> 00:40:40,400 Now think of, OK, now I'm going to do some side project. 749 00:40:41,120 --> 00:40:44,640 But that also comes up with like you read a lot of paper, you 750 00:40:44,640 --> 00:40:47,200 read, you understand the market, you understand things. 751 00:40:47,640 --> 00:40:52,400 And then I think that medium long term work is now much more 752 00:40:52,400 --> 00:40:56,360 interesting because now you have you are creating that business 753 00:40:56,360 --> 00:40:59,400 hunch within yourself of like, oh, where is everything is 754 00:40:59,400 --> 00:41:02,280 moving to work to based on your experimentation. 755 00:41:02,280 --> 00:41:06,840 So it's not so much of betting on like it's closed eyes. 756 00:41:07,960 --> 00:41:11,080 So you're actually betting on things that you still can't see. 757 00:41:11,200 --> 00:41:15,000 Going to take a bit of time to bring value out of it, but 758 00:41:15,000 --> 00:41:17,560 eventually it will come. I think it's very smart, like 759 00:41:17,560 --> 00:41:20,120 20% that would be one day a week. 760 00:41:20,560 --> 00:41:23,680 If I were to get one day a week to like read through articles to 761 00:41:23,680 --> 00:41:26,920 experiment with programming languages that I like and trying 762 00:41:26,920 --> 00:41:30,480 to kind of fix a problem domain wise that I know with new 763 00:41:30,480 --> 00:41:32,520 tooling that is out there, I think a lot of people would get 764 00:41:32,520 --> 00:41:34,840 inspired to start doing something. 765 00:41:34,840 --> 00:41:37,960 And also there's good energy of sharing then from a perspective 766 00:41:37,960 --> 00:41:40,120 of learning. And I think it's really was that 767 00:41:40,120 --> 00:41:41,960 already there in IKEA before you joined? 768 00:41:42,160 --> 00:41:45,400 It's, it's just beyond technology part of the IKEA. 769 00:41:45,480 --> 00:41:49,000 It's it's like an IKEA thing to to do that. 770 00:41:49,000 --> 00:41:50,440 I think it's even written in my contract. 771 00:41:50,480 --> 00:41:53,720 Oh, that's really good. But I I will add this, not 772 00:41:53,720 --> 00:41:56,200 everyone is motivated based on this. 773 00:41:56,200 --> 00:41:59,240 A lot of people just want to come and do their job and leave 774 00:41:59,400 --> 00:42:00,680 and that is totally fine as well. 775 00:42:00,680 --> 00:42:05,120 So not everyone needs to experiment, explore, do things. 776 00:42:05,560 --> 00:42:10,320 Some people are just not meant for that and I think it's much 777 00:42:10,320 --> 00:42:13,600 harder then a lot because a lot of people think, Oh yeah, now 778 00:42:13,600 --> 00:42:15,480 I'm going to put every Friday on doing this Sundays. 779 00:42:15,560 --> 00:42:19,720 But it's really not that simple when you have a ton of business 780 00:42:19,720 --> 00:42:21,880 push and we need to get this done and that done. 781 00:42:21,880 --> 00:42:27,720 So it it takes discipline to actually explore and think and 782 00:42:28,400 --> 00:42:30,480 and be creative. It's not as easy. 783 00:42:30,680 --> 00:42:33,920 Yeah, you get that. It's like your regular day job 784 00:42:34,160 --> 00:42:35,920 is still there. So if the deadlines are still 785 00:42:36,000 --> 00:42:38,400 there, then it's like, yeah, it just files up. 786 00:42:38,480 --> 00:42:41,080 It's like when you have a four hour meeting and you're like, 787 00:42:41,080 --> 00:42:44,000 man, why was I in this meeting? Because all my work is still 788 00:42:44,000 --> 00:42:45,200 there and I didn't get anything done. 789 00:42:45,760 --> 00:42:48,480 I get that from a discipline. Hopefully your AI agent can go 790 00:42:48,480 --> 00:42:50,720 to the meeting. And give me a summary, what was 791 00:42:50,720 --> 00:42:52,800 said. Yeah, yeah, yeah. 792 00:42:53,000 --> 00:42:55,480 When I asked you where business outcomes and and newer 793 00:42:55,480 --> 00:42:59,760 technologies align and what you've seen, you didn't tell me 794 00:42:59,760 --> 00:43:01,520 about the traditional one, which I thought you were going to 795 00:43:01,520 --> 00:43:04,200 mention, which is chat bots. I see a lot of companies, the 796 00:43:04,200 --> 00:43:07,600 first thing to do is they make this custom chat bot, like man, 797 00:43:08,040 --> 00:43:10,720 how much value is in this custom interaction? 798 00:43:10,720 --> 00:43:13,520 Because it's sure it's a tangible use case, might be a 799 00:43:13,520 --> 00:43:16,280 good one, I'm not sure yet. There's a lot of companies that 800 00:43:16,280 --> 00:43:19,600 are doing chat bots and trying to solve this problem of 801 00:43:19,600 --> 00:43:22,560 customer service and human in the loop with AI and that's 802 00:43:22,560 --> 00:43:24,560 their core. And then I also see customers 803 00:43:24,560 --> 00:43:26,600 instead of getting that off the shelf, trying to do that 804 00:43:26,680 --> 00:43:29,600 themselves, which there's a positive part because they have 805 00:43:29,600 --> 00:43:31,680 their own knowledge bases, but then there's also the downside 806 00:43:31,680 --> 00:43:33,040 because they have no clue what they're doing. 807 00:43:33,480 --> 00:43:36,000 Chat bots are kind of everywhere I've seen as first example, 808 00:43:36,200 --> 00:43:37,560 right? What is your what's your take on 809 00:43:37,560 --> 00:43:39,720 that? I think this, this is really 810 00:43:39,720 --> 00:43:42,560 part of that is you going from zero to 80%, which is really 811 00:43:42,560 --> 00:43:44,520 easy. And chat bot is one of those 812 00:43:44,520 --> 00:43:47,600 it's like getting frequently asked questions, chat bots. 813 00:43:47,920 --> 00:43:50,680 I'm going to put some documents in it and now the answers from 814 00:43:50,680 --> 00:43:55,720 my knowledge base and etcetera. And, and that is not an area 815 00:43:55,720 --> 00:43:58,800 that excites me. There is there might be value in 816 00:43:58,800 --> 00:44:00,480 it. I have my own doubts. 817 00:44:00,480 --> 00:44:06,720 I could not put my bets on it if it was a horse race. 818 00:44:06,720 --> 00:44:13,240 I would not put my bets on this. But I do think if you want to 819 00:44:13,240 --> 00:44:17,720 build a chat bot though, how can this chat bot is going to be the 820 00:44:17,720 --> 00:44:20,720 world's best chat bots? That is exactly what your 821 00:44:20,720 --> 00:44:23,560 company meant to do. If it is the same exact same 822 00:44:23,560 --> 00:44:28,240 chat bot as across the board. Now it's your logo is on it and 823 00:44:28,240 --> 00:44:30,640 you are doing exact same question and answers and you're 824 00:44:30,640 --> 00:44:33,760 doing exact same thing. I just don't think that is going 825 00:44:33,760 --> 00:44:38,480 to cut it for any customer and and I think the number of users 826 00:44:38,480 --> 00:44:41,560 is also going to drop. I think Klarna showed us what 827 00:44:41,560 --> 00:44:45,000 not to do in this space, firing everyone and then bringing each 828 00:44:45,000 --> 00:44:47,040 other. Like I would not want to call my 829 00:44:47,040 --> 00:44:51,480 bank and not be able to talk to someone because now they decided 830 00:44:51,480 --> 00:44:54,640 to cut cost and they thought now I'm going to be a lot more 831 00:44:54,640 --> 00:44:56,480 efficient if I talk to a chat bots. 832 00:44:56,840 --> 00:44:58,520 Because these are the moments that you're interacting with 833 00:44:58,520 --> 00:45:00,760 customers and it needs to be memorable. 834 00:45:00,760 --> 00:45:05,280 It needs to be amazing. And yeah, there is a space for 835 00:45:05,280 --> 00:45:07,320 chat Buster and I think it's also business decisions. 836 00:45:07,320 --> 00:45:11,920 Do you want to go to a world of no human interaction maybe makes 837 00:45:11,920 --> 00:45:16,720 sense for some businesses, but I wouldn't prefer to work in in 838 00:45:16,720 --> 00:45:19,600 that type of space because that's that's not what I think 839 00:45:19,920 --> 00:45:24,800 is right application. Yeah, for for within my 840 00:45:24,800 --> 00:45:27,720 proximity. And working, I think it's very 841 00:45:27,720 --> 00:45:30,880 business specific. It's like, would I rather, and I 842 00:45:30,880 --> 00:45:33,440 had this conversation because there's a person in the know is 843 00:45:33,440 --> 00:45:36,280 working on the startup, would I rather wait 20 minutes on the 844 00:45:36,280 --> 00:45:38,440 line to get a human in the loop like on the phone? 845 00:45:38,840 --> 00:45:41,400 Or would I rather an AI picks up the phone, which is like a 846 00:45:41,400 --> 00:45:43,120 human. I barely can like have the 847 00:45:43,120 --> 00:45:46,280 difference and they pick up and they do like the first analysis. 848 00:45:46,760 --> 00:45:49,000 It's like, yeah, I know. I don't like waiting. 849 00:45:49,120 --> 00:45:51,760 Like I don't think humans are patient innate. 850 00:45:52,160 --> 00:45:53,560 So then it makes sense to do that. 851 00:45:54,200 --> 00:45:56,800 It does need to be very well done. 852 00:45:56,800 --> 00:45:59,480 I feel like like 80% is not good enough, especially when you're 853 00:45:59,480 --> 00:46:02,680 interacting with humans. That is your customer base. 854 00:46:02,680 --> 00:46:04,560 Your reputation is also based off that. 855 00:46:04,760 --> 00:46:06,400 There's a lot of mouth to mouth marketing. 856 00:46:06,520 --> 00:46:09,200 If it sucks, my whole friends and family will probably know 857 00:46:09,200 --> 00:46:10,920 that it sucks. So like that's this little 858 00:46:10,920 --> 00:46:13,200 bubble. And if you underestimate that 859 00:46:13,200 --> 00:46:15,360 and you get something up and running and 80% and you think 860 00:46:15,360 --> 00:46:17,800 that's good enough, I think you might be underestimating or 861 00:46:17,800 --> 00:46:21,160 undervaluing your customers as well, which is a shame. 862 00:46:21,160 --> 00:46:23,120 At the end of the day it's like hard earned people that are 863 00:46:23,120 --> 00:46:25,400 already there, they already bought something, they need help 864 00:46:25,880 --> 00:46:28,160 or there's something wrong and they definitely need help still 865 00:46:28,560 --> 00:46:31,600 and then if they don't get help they won't come back likely. 866 00:46:31,800 --> 00:46:35,640 Right, right, No, I mean you're you're definitely onto 867 00:46:35,640 --> 00:46:38,160 something, you know, right. There is a space for it. 868 00:46:38,160 --> 00:46:40,760 I'm not saying burn all the chat box to. 869 00:46:41,240 --> 00:46:43,280 The ground. I think there's a space for it, 870 00:46:43,280 --> 00:46:49,760 but I just think there is a lot of like I can go to a place with 871 00:46:49,760 --> 00:46:53,680 even within a company within within where I work or within 872 00:46:53,680 --> 00:46:56,760 anyone. I think whoever that starts to 873 00:46:56,760 --> 00:47:01,200 try generative AI, their first use cases is like really 2 three 874 00:47:01,200 --> 00:47:02,920 things and it just constantly repeats. 875 00:47:03,520 --> 00:47:06,640 And we need to ask ourselves, is that easy to do that? 876 00:47:06,640 --> 00:47:09,720 If it's that easy to think about it, either it needs to be really 877 00:47:09,720 --> 00:47:14,360 good, well done, or it doesn't really bring any value because 878 00:47:14,440 --> 00:47:18,920 everybody is doing it. So then it's not valuable as 879 00:47:18,920 --> 00:47:21,240 much as spending your time on something else. 880 00:47:21,560 --> 00:47:24,680 But if you think about what is a chat boss, how the chatbot can 881 00:47:24,680 --> 00:47:27,760 be unique for the customer, I think then there's the the 882 00:47:27,760 --> 00:47:31,680 conversation kind of changes from the existing one that we 883 00:47:31,680 --> 00:47:32,400 have. Yeah. 884 00:47:32,960 --> 00:47:35,960 I mean, whatever we do, I think there's a lot of stuff that we 885 00:47:35,960 --> 00:47:38,000 can do. And I really appreciate people 886 00:47:38,000 --> 00:47:41,240 like you that are like, OK, I have this strong vision, both 887 00:47:41,240 --> 00:47:43,480 long term and short term. And they're very much aligned 888 00:47:43,480 --> 00:47:46,520 with regards to business outcomes and where technology 889 00:47:46,520 --> 00:47:50,320 meets business at the end of the day, where it solves a problem 890 00:47:50,320 --> 00:47:52,600 as a means to an end. Those are the people that I want 891 00:47:52,600 --> 00:47:55,200 to be working with. And I feel like especially now, 892 00:47:55,440 --> 00:47:57,640 if you're early in career or you're more senior in career, 893 00:47:58,000 --> 00:48:00,200 you also want to be working, working with people that have a 894 00:48:00,200 --> 00:48:03,360 strong vision where there's an experimentation culture where 895 00:48:03,360 --> 00:48:05,200 you can feel fast and learn quickly. 896 00:48:05,760 --> 00:48:07,360 I mean, I'm saying this from personal experience. 897 00:48:07,360 --> 00:48:09,520 That's the environment that I would want to be in, especially 898 00:48:09,520 --> 00:48:10,880 because things are kind of accelerating. 899 00:48:11,360 --> 00:48:13,400 It's now cloud code. And then next week, I don't 900 00:48:13,400 --> 00:48:15,840 know, AWS might release something and then all of a 901 00:48:15,840 --> 00:48:19,120 sudden there's something else is like, there's so much to learn 902 00:48:19,120 --> 00:48:22,040 and so much to experiment with. If your environment doesn't do 903 00:48:22,040 --> 00:48:23,600 that, then you might get outpaced. 904 00:48:23,600 --> 00:48:26,520 And then a few years you might find yourself in this weird spot 905 00:48:26,520 --> 00:48:29,240 where all of a sudden there's this new role, AI engineer, and 906 00:48:29,240 --> 00:48:31,440 you only see components of like what you used to do. 907 00:48:31,760 --> 00:48:33,640 And it's, yeah, you're like, what do I do now? 908 00:48:33,680 --> 00:48:36,840 Yeah, I feel like it's good to keep learning and keep growing 909 00:48:36,840 --> 00:48:37,520 of. Course, yeah. 910 00:48:37,720 --> 00:48:41,040 The fun part is that especially in this industry, there's a lot 911 00:48:41,040 --> 00:48:43,640 of stuff online which I've always appreciated. 912 00:48:43,640 --> 00:48:46,720 I think that makes this industry very unique and it's very kind 913 00:48:46,720 --> 00:48:48,920 of fluffy for me to say because it's the only industry I've 914 00:48:48,920 --> 00:48:52,200 worked in. But still, the only downside now 915 00:48:52,200 --> 00:48:55,360 is that there's so much content. It's like, where is the essence 916 00:48:55,360 --> 00:48:57,880 and where's the gold? But that's why I really 917 00:48:57,880 --> 00:49:00,240 appreciate you coming on this podcast and sharing your 918 00:49:00,240 --> 00:49:01,880 perspective and kind of take on things. 919 00:49:01,880 --> 00:49:03,080 I've really enjoyed this conversation. 920 00:49:03,120 --> 00:49:06,920 Of course, thank you for having me, I enjoyed the conversation 921 00:49:06,920 --> 00:49:08,000 as well. Cool. 922 00:49:08,360 --> 00:49:10,920 Yeah, thank you for having me. It's indeed, I mean, there's so 923 00:49:10,920 --> 00:49:13,800 many podcasts and contents and everything out there, there. 924 00:49:13,800 --> 00:49:17,800 I I have always like so many podcasts always saved and and 925 00:49:18,000 --> 00:49:19,760 never get to go through all of them. 926 00:49:19,760 --> 00:49:23,120 And I feel bad. So, so indeed, it's great that 927 00:49:23,120 --> 00:49:25,560 we have so much knowledge and so much different perspectives out 928 00:49:25,560 --> 00:49:27,520 there which are really exciting. Cool, man. 929 00:49:27,520 --> 00:49:29,680 Thanks again for sharing. We're going to round it off 930 00:49:29,680 --> 00:49:31,160 here. If you're still with us, let us 931 00:49:31,160 --> 00:49:32,880 know in the comments section what you've thought of this 932 00:49:32,880 --> 00:49:34,720 episode. Leave a like if you liked it and 933 00:49:34,720 --> 00:49:36,080 otherwise we'll see you on the next one.