# Check-in — Clean transcript > Source: `25 Aug 08_28.txt`. Speaker labels are machine-generated and may be wrong. Break chatter and obvious recording spill have been removed; sequence, wording, language switches and uncertainty are preserved. ### Speaker 2 · 00:17 Great to see you all again, and great to see you in, in, uh, in flesh or, uh, in person. Kevin. Uh, welcome back. Uh, it was, uh, s- almost six months or five months since we, uh, since we last met. And, uh, lots of things has happened. It was funny because when, uh, when me and Jocke started, uh, planning this session, we started reminiscing, okay, looking back at, hey, what has happened in AI since, uh, since end of March. And I'm guessing it's like, uh, in Internet years, six months is probably a couple of decades. But in AI years, it's probably moving between Bronze Age to Renaissance or indus-industrialiZations, I would say. So I'm guessing a ton of things have happened, uh, also for you, uh, not just as leaders, but also as a company. Uh, so it's gonna be really fun to dig in to see, okay, how can we equip you with things that can help you and support you, uh, in this time of change, uncertainty, and I'm guessing lots of complexity. And we landed on kind of framing the day around this subject. So today, we're gonna build something. Uh, we're gonna build up both energy. ### Speaker 2 · 01:42 We're gonna use a lot of tools similar to last time, lots of hands-on things. Uh, and we're gonna introduce you to a new colleague that you are gonna hire, an AI colleague. Um, and the thing is that everything we do today is deliberately designed to kinda help you out first thing of tomorrow. So you will be given the assignment to come and say, "Okay, how can we shape an AI companion, colleague, coach, something that can help you manage complexity, lead, but also mundane tasks that you're still doing, I'm guessing, copy-pasting and all of that, sorting out data?" So that's the kind of framing of today. Uh, and we will do that, uh, more and more. You will do the work, we will do less of the work, and in the afternoon, you will kind of take over the ship and steer it. Uh, we will start today by going through a bit, okay, what has happened the six months that has, uh, gone by since we last time met. Uh, and looking at, okay, where are you at? What has worked? What has not worked? Where are you stuck? ### Speaker 2 · 02:55 Uh, and then, then moving to, okay, how, how can this, these insights, these reflections shape how you want to show up tomorrow with our new companion? Does that sound good? ### Speaker 4 · 03:08 Yes. Any questions? Great. ### Speaker 2 · 03:15 I want to-- Do you remember this slide? Does anyone remember this? This was a slide I introduced on the first day when we met, talking about a kind of our approach, uh, at Novare to, to learning and to leadership and to new technologies. And I think I want to-- I mean, I kinda wanna remind us of this because, uh, as we see it, rega- irregardless of what kind of techna- technology that is introduced into society that is kind of disrupting things, we as organizations, we, we try to adapt, but the real kind of either bottleneck or enabler is for us as humans, right? So the behaviors, the habits, the culture that we either enable or stop is what kind of moves the needle or stops it. And I fell in love with this kind of thing as well, uh, this model. So just to show what I mean in, in another way. But we have, uh, AI, the technology layer that, uh, uh, is introduced either through data security platforms or integration. Uh, it, it is a promise around business impact, productivity, quality, innovation, speed, progress. But the hamburger meat is really in the middle. ### Speaker 2 · 04:33 So what is either, like I said, a bottleneck or enabler is the leadership layer, how we show up as humans and how we can lead not just people anymore, but AI agents and processes. And that's a big shift because leading others or leading leaders has until now been leading team members. But now it's all about, okay, what kind of processes and what kind of, uh, AI agents are we gonna introduce into this? So we're gonna do a lot of technical things with your computers today, lots of tools, but the real kind of reflection needs to be around this. Okay, how can we together shape a culture that enables, uh, habit and behaviors that supports this new shift? Does that make sense? Great. Jocke, do you wanna add anything? ### Speaker 5 · 05:29 No. ### Speaker 2 · 05:30 No? ### Speaker 5 · 05:31 I'll, I'll talk enough later, so. ### Speaker 2 · 05:34 Um, let's start by kind of moving down the funnel of where we've been. So, uh, we're gonna do a check-in, and, uh, obviously, we want to know, okay, where, where are you at with AI? Uh, so what has kind of exceeded your expectations since last time? And, and where are-- w-where, where are you still stuck? Where do you hit resistance? What is frustrating you? But also, what is one thing that others in the room might be missing? And this is crucial because what we see a lot around the companies right now is that people tend to hesitate to share what they're doing with AI, either because they're thinking, "Okay, they-- my colleagues are probably way far away or far gone, or what do you say? Far ahead than me. So this is probably not news for them." Or you feel like, "Okay, I don't know if this is really serving us," but I think the key insight here is that it's, it's good to learn fast, but it's even more important to share faster of your learnings. So let's take a few minutes, uh, write on Post-its if you want, but, uh, answer these questions one by one. ### Speaker 2 · 06:53 What is exceeding your expectations, like aha moments that helped you out? And where are you really frustrated now as well? Or hit a wall for resistance. And think of something that you think others in the room might be, uh, needing to hear to be beneficial. ### Speaker 6 · 07:18 One thing that they're missing. ### Speaker 6 · 08:43 We don't have a-- we don't have to rush this. So if, uh, like it's okay to, kind of, try to go back. ### Speaker 2 · 14:02 Right. So let's, let's have a discussion. Two and two and you have to end three. So you two, you two. Oh, let's see. You two and. Oh no, you two and you three. Sorry. So let's, let's have a discussion. One second. Let's, let's listen to each other. And, and also have when you listen, what kind of patterns do you see? So what kind of patterns? Is it processes? Is it task? Is it onboarding? Is it sharing? What is it that you're hearing that you're either succeeding with or struggling with? Det här var små diskussioner om det. ### Speaker 8 · 14:54 Seven minutes. ### Speaker 1 · 15:14 Tina kan börja. ### Speaker 8 · 15:28 My expectations. Det är att gud vad du tittar på mig. Vad konstigt exempel. Eller så känner jag. ### Speaker 3 · 15:38 Exakt. Vi vill dig väl. ### Speaker 8 · 15:39 Tack. Att det är så lätt att använda och att man kan you get so much more out of everything. Jag som inte är en early adapter generellt. Man är ju tvingad till att vara det. Men just med det här tycker jag att jag var ändå ganska liksom tidig. Så det, det har liksom generellt living to my expectation, my expectation, exceeding my expectation. Och ja, i en part of everyday life. ### Speaker 1 · 16:11 Så att. Och ja, jag använder den till och med. Och det trodde jag inte att jag skulle göra. ### Speaker 8 · 16:23 Han kopierar. ### Speaker 1 · 16:24 Nej, det är bara Ja, lyssnade du aktivt? Är det aktivt lyssnande om du liksom Ja ganska, ganska mycket. Förstå att man kollar på kameran. Det är när man aktivt lyssnar. Försöker. Ja, absolut. Nej men jag tycker har lite mer strukturerat saker som man kan använda och det funkar ganska bra. Ja, det här jag Så är det texter som kan bli 10 gånger bättre för att du sätter in dem i contexten och samlar sådär ganska mycket och kan bland det här skriva något rätt så eller summera det på en plats där vi har det. Det funkar riktigt bra. Jätteskönt, mycket bättre. Sen har vi en funktion som är fin att faktiskt redigera. Om man har bra alfasenhet material så vi tittar på det liksom om man vill ha bakgrund på vi kallar det för repurposing. Man vill göra om det till en lång dos eller man vill göra något kortare eller något annat med materialet. Och vi har också fått en spår i testet hos Kompano. AI-agent som också ger bättre access bara till vissa marknadsmaterial som har med en artificiell intelligens där vi liksom visar Vad gör agenten? Agenten, den är marknadsexperten. ### Speaker 1 · 18:02 Det är ingen agent som ligger i hela systemet. Man kan använda den och göra styrkor med den. Man kan göra marknadskampanjer. Sen håller vi på att utvärdera den, att testa om den liksom levererar all den där bara visdom. Det är väldigt bra. Det passar till de saker som man har tillgång till. All information man har. Var det har varit. Det är bra. ### Speaker 4 · 18:35 Sen har jag lagt upp några ställen där man kan se AI-agenter som skapar rapporter. ### Speaker 1 · 18:43 Det är marknadsanalys. Det är att man analyserar sin egen business. Potentiella, både interna och externa. Och en daglig sorts analys som man tittar på alla rapporter. ### Speaker 12 · 19:00 Frågan var vad som har Ja, och den har en CDX-bit. ### Speaker 1 · 19:03 Jag tycker det börjar funka bra. Faktiskt. Det måste man säga. Det tycker jag faktiskt att den verkar bra. Och sen så det som också har blivit mycket bättre sen vi använde Workshops. Copywrite så är det man måste skriva vanliga titlar liksom på tänder. Vi har gjort en match också, typ som bloggar. ### Speaker 11 · 19:31 Det tror vi går över. ### Speaker 1 · 19:35 Det är ett bra jobb kan man säga. ### Speaker 11 · 19:37 Man har gjort den här men delar den med tusen. Nu gjorde vi ju det här i Claude. ### Speaker 1 · 19:42 Det börjar röra sig. Men PowerPoints. Vad sa Claude? Den sa att Jag hade klart på ett glas. ### Speaker 8 · 20:01 Ja just det. Åtta komma fem av tio. ### Speaker 12 · 20:03 Åtta komma fem av tio. ### Speaker 1 · 20:04 Men UX leads customer experience more and more. ### Speaker 12 · 20:08 Nej men, grejen är att jag har fuskat då. Dels har jag Claude, vilket är bra. Och sen så har Håkan gjort en speciell grej med att man kan göra som en wiki i ett projektform. Och liksom containar man. Fördelen med Copilot är att den tittar över allt. Ibland är det en nackdel, men då kan man liksom avgränsa. Man kan kolla mer. Då kan jag få in dokument, till exempel ett mail från juni eller listor på anställda. Ja, fan. Så ska jag mata in det i inboxen. Så kan jag säga ingest så kan den liksom göra det. Gör det till, vilka wikis är det? Så kan man be den göra sånt efteråt. Det är ju skitsmidigt. Och då efter det då kan jag säga, ja men nu ska vi ha ett möte med advokaten eller med Clemens eller någonting. Filtra grejerna som vi ska tänka på eller göra projektplan eller leta efter, räkna ut vad det kan kosta. Och då har den redan anställda och allt vi ska kolla in och sånt. Det är otroligt smidigt. Kan man få en, komma mycket, mycket längre också. Man får koll. Men så måste advokaten då siwidosist säga. ### Speaker 12 · 21:27 Då har man väldigt bra koll. Det har typ varit mina. ### Speaker 13 · 21:38 One more minute. Och sen Kittelfis. Då ser vi release, nyaste releasen av Kittelfis. Den innehåller fem gånger mer än alla andra releaser vi sett. Och det är på grund av AI. Så teamet kommer att växa jättemycket. Lång, långsiktigt. Det, där är det liksom lite grann Learning curve and what is different learning curve, which is fine. Men hon är ju superdyrtrog. Ja, absolut. ### Speaker 1 · 22:16 Det gör sig mycket, mycket mer tester som inte är maskiner. Det gör sig liksom redan när den skriver koden så gör den det bättre. Och sen så byggs det in tester, automatiska tester på en gång samtidigt. ### Speaker 13 · 22:32 Får jag längs det här inte bort kan det göras. Det är otroligt. Det är otroligt. Ja, det har bara varit Okay everyone. I ### Speaker 13 · 22:59 put, I put up two headlines here. ### Speaker 2 · 23:01 So I'm gonna try to capture some of the things that you talked about and that you reflected on under the headline of our AI wins and where we're kind of still stuck or hitting the system. Så Johan och Lena, what, what's your AI wins? What has worked? I think, yeah, the thing is really amazing. How do you use it in project? ### Speaker 14 · 23:34 Like basically in everything I do when I do things and I ask the AI to review it or, or if I, if I want expertise in areas where I'm not so good, I, I ask it and I, I learn a lot as well when I do. ### Speaker 2 · 24:00 Do you have like a specific moment when that was like, or is it like it's been a- Yeah, yeah, we have a case. ### Speaker 14 · 24:07 You put the tap on and then it's just. I mean, everybody knows what that is, but we have a case in Europe here and I, I basically need to know every VAT rule in Europe because we negotiate it. And AI, you know, it's perfect for that. It, it like, it replaces a super expensive consultant basically. And it gives me like all that and I know that so. ### Speaker 5 · 24:30 Wasn't the EU supposed to be one single market? ### Speaker 14 · 24:34 Yeah, that's what they thought, but it's just even more complex. ### Speaker 5 · 24:37 AI to the rescue. Yeah Helena? ### Speaker 15 · 24:43 Yeah, so it's a-about this, but I, I expressed it like AI's capability to, uh, I mean, summarize and structure a topic. And you learn from it, you get all the info and, uh, and so. Both in private life and at work. ### Speaker 2 · 24:59 Mm. Have you used it, uh. Yeah, I guess you use it seamlessly between private and- Yeah. Yeah. ### Speaker 15 · 25:11 Well, kind of- You guys use the same, or is it- No, I mean, I use Chat- Yeah .GPT privately and then Copilot at work. And also language, and then- Yeah .you get help- Yeah .with expressing yourself. ### Speaker 2 · 25:27 Like translation or- Yeah .yeah, or language? ### Speaker 15 · 25:29 Good English. ### Speaker 10 · 25:30 Or just like it softening your reception. ### Speaker 15 · 25:33 Yeah. ### Speaker 10 · 25:34 Oh, yeah. ### Speaker 15 · 25:35 Or like, uh- Yeah .being more. Yeah. Yeah, you rewrite it. ### Speaker 10 · 25:37 You can ask- Yeah, you can get, like, get a tone. ### Speaker 2 · 25:39 Be more direct. ### Speaker 15 · 25:40 Yeah, exactly. Yeah. ### Speaker 10 · 25:41 Okay, nice. ### Speaker 15 · 25:42 Or do you want to present this to Clement? ### Speaker 16 · 25:45 What is, what is the thing what you do differently? ### Speaker 10 · 25:50 But it starts asking you now. ### Speaker 15 · 25:51 Do you want me to write this, uh, for- It's for Clement. Exactly. What is different? Ask, ask Copilot. Um, yeah, so it's the broad reach, uh. ### Speaker 2 · 26:05 How do you mean? How do you mean? ### Speaker 15 · 26:06 Yeah, but, uh, the, the, the, the capability- Yeah .I mean, to go out and, and search everywhere for everything- Yeah .more or less, via the things or whatever. ### Speaker 2 · 26:16 Um- Where are you stuck? Or where do you kind of, uh- Yeah .feel like a resistance or frustration? ### Speaker 15 · 26:24 Yeah, I think, uh, in the organization, maybe it's still, uh, I think it's a few using it. So if we have like, um, um, one person, for example- Adoption? Yeah .in, yeah, in the second line team, then the rest is not that interested or, uh, feels, uh, hesitation to start using it, for example, for troubleshooting. And then when it-- I think also it repeats itself. I mean, if you discuss for a long time, you get- Mm. And maybe it's me prompting the wrong way, but I think I get-- I try to, to get it to- It's like a loop? .shall-- Yeah, exactly. And then also, uh, yeah, summarize it into a PowerPoint. I think I prefer doing that maybe myself, and that could also be me. ### Speaker 4 · 27:08 Mm-hmm. Mm-hmm. ### Speaker 15 · 27:09 Mm. Um, pictures. Uh, I tried to do home styling with, with it. Yeah. And that, I mean, they are selling that, uh, through AI now, but, but, uh, yeah, it didn't work perfectly. So when pictures is involved, I, I at least have a problem on creating good enough material. ### Speaker 14 · 27:31 Yeah. How about you, Ivan? Uh, I think we're still stuck in, in, in, in, in the, um, in the local IT organization or within the company, in like how we use it and access and all that. Mm-hmm. It's still, uh, it's, it's a still a lot of frustration and still takes a lot of time. And then we have also a different landscape of applications now, and it's very difficult to work as a team when everybody have different accesses. They use different AI apps. Mm. Some use Claude, some use, uh, ChatGPT, some use Copilot. And, and, and if you build then something, an app- Mm .or, or an agent, um, it's difficult to get its broad usage or so, um- Mm. That, that hinders a little bit, I think. Yeah. At the same time, it's good to explore, but it's, um, yeah. Exactly, yeah. Great. Anything else? No. ### Speaker 5 · 28:31 Can I just ask a follow-up to- Yes, yes .to Helena? When you work with ChatGPT, do you have a paid subscription or do you just use the free one? ### Speaker 15 · 28:38 No. Yeah, I use the free one. ### Speaker 5 · 28:39 Okay. Yeah, because when you said that it was bad at pictures- Mm .there is a massive difference- Yeah .between the paid account and the- Yeah. ### Speaker 15 · 28:46 I actually tried my partner's, and he, he is paying for it. It was much better on his phone. ### Speaker 5 · 28:53 Can't he pay yours as well? No, I think- What a cheap partner you have. ### Speaker 2 · 28:56 That is, that is the one, one thing that others in the room are missing. Someone else is paying. Someone else, yeah. Burn someone else tokens. Yeah, that's the. The insight. Yeah. How about you? You are nodding, Katrin. ### Speaker 10 · 29:10 Want to start? Uh, um, okay. What succeeded? What's the, the AI wins? I think, uh, a lot of the same, but, like, this summarizing and structuring, like going through- Mm .large amounts of, of data. But I think that is already up there. Um, and then another thing, I think, uh, like building models or framework that I could use with my team. ### Speaker 7 · 29:34 Mm-hmm. ### Speaker 10 · 29:35 Uh, like I- In what way? Like, for example, I, uh, we needed to do a gap analysis in competence from, like, where we are now- Mm .to what our strategies as where we're gonna be in, like, three to five years. And, like, we, we, uh, we're gonna look at what competencies are we missing or, like- Mm. And I didn't really know how to ad-- like approach it with my team. So then I built kind of a framework that we could use- Mm .when doing that analysis. ### Speaker 3 · 30:05 Nice. ### Speaker 10 · 30:06 So that was nice. ### Speaker 3 · 30:07 Yeah. ### Speaker 10 · 30:09 Uh, but still stuck, I think information overflow. Mm. You know, because you prompt, and then you get, like, so much words added. And I think you can probably prompt more efficiently and, and maybe get around that. Mm. But, uh, in general, easy to just get so much information that you then have to work through and see- Yeah .what's relevant and. So it's probably because of bad prompting, but, um- But how do you handle that when, uh, that happens? ### Speaker 2 · 30:38 Like, what, what do you do in the room? ### Speaker 10 · 30:40 Uh, either I move on to something else. Or I ask it to, like, I try to, uh- Sort it, yeah .you know, uh, reframe my prompting or work around it. Yeah. But then I also think it, sometimes it feels like you get contradictory, uh, answers and then- Mm .um. So but, but I think me and my team, we have, uh, in our weekly meetings, we have one checking point, which is like sharing, um, AI wins or whatever- Mm. Yeah . like things that we did. And but one thing that keeps coming up is this information overflow. Like, if you have, for example, one team member that, uh, they're typically recording or doing the transcript- Yeah . on meetings. ### Speaker 6 · 31:24 Mm-hmm. ### Speaker 10 · 31:24 And then AI or Copilot creates all these actions that they feel compelled that they have to do because they're there, like- Yeah . in the min-minutes of meeting, that these are the actions that came out. And then there's actions all over the place and, you know- There's too many actions . Uh, and if you're a person that, uh, you know, want to follow the rules and- Yeah . do what you're told- Mm-hmm . then that becomes a problem. ### Speaker 6 · 31:50 Mm-hmm. ### Speaker 2 · 31:51 So Mm-hmm. ### Speaker 6 · 31:52 Yeah. Yeah. ### Speaker 10 · 31:53 Luckily, I, I don't follow the rules, so I don't have to worry. But I have team members that do, so yeah. ### Unknown · 31:59 Mm-hmm. ### Speaker 5 · 32:00 Can I just ask- Yeah . a follow-up? Yeah. Because I've heard that a couple of times now on the, the long outputs and the word salad. I love that word. Um, are you using the custom instructions to, to tailor and, and tell Copilot what you're expecting, uh, in, in personalization? No, I haven't really. Okay. ### Speaker 10 · 32:18 So that's, uh, probably a good point. ### Speaker 5 · 32:20 Yeah. Yeah. So we're gonna look into that then- Yeah . because that's, that's a- I know it's there, so yeah. Yeah. Exactly. Yeah. Yeah. Thanks. Great. ### Speaker 10 · 32:27 Good. ### Speaker 5 · 32:28 Yeah. ### Speaker 11 · 32:28 Yeah. Um, I'm heavily user of this chat or speaking function. Mm-hmm. So I- Oh, yes. Yeah . I, I like we did, did, you know, in the g- training. So- Yeah . w- when I'm in the car in the morning, I have like half an hour and at least one hour back home. So almost every day, I'm consuming max. I, I'm getting kicked out of- . our Copilot. So I am consuming more than one hour. ### Speaker 2 · 32:53 Mm-hmm. ### Speaker 11 · 32:54 Uh- Yeah . yeah, so I, I use it for a lot of things. ### Speaker 2 · 32:57 Yeah. And what, what, what do you get- So- . from it? Like when the w- so you talk about like- No, I, I, I- . ### Speaker 11 · 33:02 work- . practice speak German, so I- Yeah . become, uh, you know, so much I. And now I started to, to take, uh- To Tesco . Yeah. We could do that in your lunch. I could take the subjects but in German. Mm-hmm. And it's a really big challenge because then I learn the words. Uh, so, so I think it's. And now it's improving. Earlier, it was only, uh, it was listening, and then it was speaking, right? Mm-hmm. It couldn't do it at the same time. But now if I do the like the open, uh, uh, not the within our Copilot, but, but then it could do. It's more interactive. Mm-hmm. It has really improved- Mm-hmm . the last, yeah, weeks or months. ### Speaker 5 · 33:37 Yeah. Yeah. ### Speaker 11 · 33:38 So that's, uh, one thing. And then the other win is that I think the, the, uh, the, the, in the organization and in my team, they're really. Everybody is now sort of- Everybody on board. Yeah . chatting and doing. Yeah. ### Speaker 1 · 33:49 Mm-hmm. ### Speaker 11 · 33:49 We are doing a lot more now- Yeah . versus six months ago. Yeah. I see a big, big difference. ### Speaker 1 · 33:54 Mm-hmm. ### Speaker 11 · 33:55 So a-a-and then the, the fact, I think we are, uh, both, uh, the, the. I, you know, I developed, uh, an okay agent, but it's, uh. I-I'm stuck on, you know. I'm playing with it. I develop it. But it's a different thing when we should roll it out in the organization- Mm-hmm . because then it's like, yeah, yeah, it's a software development project- Mm-hmm . basically. You know, uh, I need, I'm. I, I, I would need to turn my h- text prompt, the skill I built it in, in Claude into, um- Copilot . uh, yeah, yeah. I would. Copilot to roll it out- Mm-hmm . but, you know, it needs to be testable- Yeah. Mm-hmm . and probably some parts need to be exchanged with code to make it repeatable. I mean, I, I'm getting stuck to- You get to a certain point- Yeah . ### Speaker 2 · 34:40 and then it's like, "Okay, how do we-" Yeah. ### Speaker 11 · 34:42 And then I don't think we have any good sort of plan of what to- Mm. Okay . how should we proceed as- Yeah . an organization? What should we prioritize? ### Speaker 5 · 34:50 Scale and deploy and have it- Yeah. ### Speaker 11 · 34:52 Yeah. And what, what. Yeah, exactly. And increase the, like, the, the, the maturity of what if we, uh, deploy an agent. ### Speaker 5 · 35:01 Mm-hmm. ### Speaker 11 · 35:02 Like we were saying, it feels like we're playing with it right now. Yeah. Now it's more playing. It's like- It's okay for playing. Yeah. But, I mean, if everybody should use my agent- Mm-hmm . maybe we should- Have some plan for that. Yeah, mm-hmm . yeah, and maybe we should test it before. ### Speaker 5 · 35:12 Mm-hmm. Mm-hmm. ### Speaker 11 · 35:13 Somebody needs to prove it. ### Speaker 2 · 35:14 Exactly. ### Unknown · 35:15 Yeah. ### Speaker 2 · 35:16 Great. ### Speaker 5 · 35:17 I, I, I think you- Yeah . you forgot one thing. I know something that you're stuck on. You mentioned that you hit the, the, the barrier every day, that you, you hit- Yeah. The tokens . your usage limit, right? Yeah, but then I change. I think that's definitely something that you guys need to, to talk about, right? Because that is definitely a feeling of getting stuck when you run out of tokens, and you're not allowed to use the tool anymore, right? ### Speaker 11 · 35:39 So- Yeah, but then I sh- I switch. ### Speaker 5 · 35:40 Uh, y- but you shouldn't have to work around it, right? Yeah. Exactly. It should be kind of- Yeah . institutionalized. ### Speaker 10 · 35:46 I didn't even know that was possible, so maybe we could have a shared pool- Uh, but I, I'm like a corner case. Nobody So you don't, don't find me. Then, then you shouldn't run out because- Your second partner . uh- Yeah. Exactly . you should be able to use my tokens because I- Mm-hmm. ### Speaker 5 · 35:58 Yeah, exactly. ### Speaker 2 · 35:58 Yeah. ### Speaker 5 · 35:59 Something to figure out. ### Speaker 10 · 36:00 I think one hour probably. ### Speaker 2 · 36:01 Kevin and, and Clemens, what's your, uh, AI wins since last time? ### Speaker 13 · 36:06 Yeah, the, the, the wins are definitely. A big win for me was Copilot Cowork- Mm-hmm . that I have the APIs to- Mm-hmm . my entire infrastructure. I think that really made a leap, leapfrog improvement. ### Speaker 2 · 36:16 In what way? ### Speaker 13 · 36:16 Um, yeah, that is suddenly really, um, I had access to my open, um, my OneDrive, my email- Yeah . everything, I think, in a, in a much more professional way than what we focused then in Copilot. That wa- that was a big win. Yeah. Mm-hmm. And, uh, especially used it then for, uh. So Robert and I, we started working then on, uh, on a project in Japan, and Robert used Claude, yeah. And, um, I used, uh, then Copi- Cowork, and I was impressed how it really. how quickly it created file structures, quick documents, started creating drafts on emails, yeah, based on my input and also based on history. Mm-hmm. I think that was, to me, phenomenal. Mm-hmm. And I, I gained so much time back, which I could use for other things. Uh, I, I have to hon- be honest So, and then I prepared a trip to Japan, meeting with law firms, et cetera. But when I would have done that all on my own, the preparation, I would have never done that. ### Speaker 6 · 37:20 Mm. ### Speaker 13 · 37:21 I would have gone much more intuitive in the meetings with gut feeling. Yeah. You know? Whether or not I was really professionally prepared. I think that, that was, that was a significant gain for me. Yeah. Um, I s-- I think it's, it's getting more and more tailor-made to me when I interact with, with AI. ### Speaker 6 · 37:39 Mm. ### Speaker 13 · 37:40 I have the impression in the beginning it was more like I talk to the AI, and the AI talks to me in general, but now it's really getting tailor-made. Mm. It, it's, it's much more professional towards me, yeah. Whatever. When I, for example, I use speech to text, and then by some reason it interrupts me in the middle of a sentence and then asks me, "But you didn't finish your last sentence. Was there something relevant?" Yeah. And then. ### Speaker 2 · 38:08 Yeah. ### Speaker 13 · 38:08 What's going on? Uh, the, the. I think it's, it's another- Yeah . it's another quality. Uh, so th-this is to me a real, a real win. Uh, I'm still getting stuck on. And I still haven't, haven't yet found out why. So speech to text is extremely good on ChatGPT, in my, in my private, um, use, use. It's extremely good on phone and iPad. Uh, I struggle with speech to text on Copilot, and I struggle on speech to text with a laptop. ### Speaker 4 · 38:39 Mm-hmm. ### Speaker 13 · 38:40 You know? I don't know if. Now I bought a professional microphone because I think maybe hardware, which also can improve it a bit. But th-this is something where at least I have the impression it's better on ChatGPT or on the iPad. Mm. Um, and the third question or way where I'm still stuck, I think it's important, and also I've improved myself, is creating agents and training the agent on finding the right skills. You spend a lot of time in improving the skills, and as more time I spend in improving the skills, as better the outcome is. Mm. I think this is, this is something you need to force yourself not to be- Mm . lazy on preparing the AI or the agent, yeah. ### Speaker 1 · 39:24 Mm. ### Speaker 13 · 39:25 And that, that's something where I think we, uh, we, we are still stuck, but this is, I think, the normal learning curve, I would guess. So there's no blame with myself or with maybe anybody. It's about trying to have the desire to be my better self together with AI. Mm. Because the AI. You always, you know, or you. Even I, a human being, I think there's always. I'm tempted to dump something. ### Speaker 16 · 39:51 Mm. ### Speaker 13 · 39:52 And then it's done, yeah. But I force myself. I think it's not about dumping it to get more time for the couch and Netflix. Mm. It's about dumping something to gain time for being together- Mm . my better self to improve- Yeah . quality together with AI. Mm. But th-th-this is I think a little bit against human, human nature, and you have to force it. If otherwise, you risk that your quality goes slightly downward over time. Mm. Mm. And you need to make sure, I think, that I help you to be better, yeah. ### Speaker 5 · 40:23 Mm. You have to invest in the relationship. ### Speaker 13 · 40:26 You have to invest. Yeah. And, and when you don't invest, and use it just as a, whatever, lazy assistant, do it for me, I think then, then I think we won't gain something. And, and this is. Th-this needs discipline, and I'm not yet fully there. Mm. But I force myself to, to. ### Speaker 6 · 40:45 Mm. ### Speaker 13 · 40:46 I think here we are somewhat stuck to, to do this. Yeah. And I also think, um. We, we talked about this. I, I, I don't think it's, it's about finding the. Of course, it's all good that you have a trainer or a mentor who helps you. But I think you need to spend much more time on pro-improving yourself rather than- Mm . trying to find somebody who can help you. ### Speaker 7 · 41:10 Mm. ### Speaker 13 · 41:11 You know? ### Speaker 2 · 41:11 Mm. ### Speaker 13 · 41:11 I, I think it, it's a real personal thing. Yes. You know? Mm. That's, that's what I think. ### Speaker 2 · 41:18 I mean, there. Like we said, I mean, it can. There's YouTube videos, and there's just. You can just start prompting. Yeah, yeah. That's a good coach. ### Speaker 13 · 41:29 Yeah. And for example, then when, when you. When, when I missed, uh, whatever, when my quality creating an agent was bad, yeah- Yeah . then I said, "Why should I find somebody to teach, to teach me make an agent? Why don't I ask Copilot- Yeah . to teach me make an agent?" Yeah. And when I then, for example, then, whatever, use Copilot to write a better email, yeah, then I save time. I think then I said, "Now this is the time I gained to have another training session for me," yeah. ### Speaker 6 · 42:00 Mm. ### Speaker 13 · 42:01 And not who is, who is my trainer for the future. Mm. Mm. It's. I have my trainer with me. Yes. And this is a, a learning, you know, which it's a completely different to whatever we have done in history. Yeah. I think it's a really strong insight there. ### Speaker 5 · 42:15 Yeah. ### Speaker 13 · 42:15 Uh, yeah. ### Speaker 5 · 42:16 Yeah, and it goes back. I mean, Aristotle, he said that what we have to learn to do, we learn by doing. So he was ahead of the curve there, right? And, and then we, then we had the thousands of years of thinking that learning is something we go somewhere to do, and someone spoon feeds us those insights. But now we're back to learning by doing. Mm. It has never been as easy before as it is today- Yes . to actually engage in that, right? It, it takes investment and time, but the, the results you get are amazing if you do it. ### Speaker 2 · 42:44 Kevin, what's your AI win? ### Speaker 3 · 42:48 What are we missing here? I would say AI wins for me are mundane tasks. ### Speaker 2 · 42:53 Yeah. ### Speaker 3 · 42:54 Um, reviewing contracts, NDAs, stuff that I'm not good at and don't enjoy doing. I think that's an AI win for me. It's super helpful. Um, com-comparative, compar-comparing documents He does a super good job. Um, so I would say that's my leading win. I would also say that, uh, got some team members who are really embracing it. ### Speaker 4 · 43:24 Mm-hmm. ### Speaker 3 · 43:25 I think that's an AI win, um, encouraging them to share their wealth and share their knowledge and share their, their experience. But I would say conversely, my biggest point of being stuck is that some folks on my team, many folks on my team, are not embracing it- Mm. Nor have they any interest in embracing it. And, and I think that is, as a leader of a, a good-sized team, that is, I think, my biggest- Sure .point of being stuck. ### Unknown · 43:59 Yeah. ### Speaker 3 · 44:00 Um, and I think different to your experience, Clemens, I think, you know, some people want to use AI- Mm .and want to do better and want to be more efficient, and some of us-- some, some folks just don't, don't. ### Speaker 4 · 44:16 Yeah. ### Speaker 3 · 44:17 And so, um, how do you get those folks- Yeah .to embrace it and to embrace change? I think as an organization, we don't embrace change very well. ### Speaker 2 · 44:28 Mm-hmm. ### Speaker 3 · 44:29 And I think this is another example of that. ### Speaker 2 · 44:31 I think it's really strange- Some folks don't want- Yeah, sorry. No, sorry. ### Speaker 3 · 44:34 Yeah, yeah. ### Speaker 2 · 44:35 No, I think it, I think it's really strange that when it comes to AI, it's like, uh, uh, for a lot of people, it's like, "Oh, it's-- maybe it's a choice to use it." But I mean, uh, if you would have, uh, introduced a CRM tool or, uh, or a salary system, and three team members are like, "Yeah, well, I'm not gonna report salary in this, the new system," that would have been like, "Yeah, that, that doesn't work, right?" Yeah, but that's bad too. So I think it's like, uh, it's not a choice anymore. ### Speaker 4 · 45:02 Yeah. ### Speaker 2 · 45:03 It is, uh, and, um, I think there are serious conversations around this, like. And it also, it's. Like Yuk-Yuk-Jukka said, it's never been easier to onboard yourself to your first AI moment. Either starting in your private life or at work. ### Speaker 5 · 45:22 Yeah. Are, are, are they passive or are they detractors? Are they aggressively against? ### Speaker 3 · 45:29 No, I would say passive. ### Speaker 5 · 45:30 Passive. That's the worst kind, right? Yeah. Because if they're actively against, at least you know where to push, right? Yeah. You, you- Yeah, yeah .you get something to hold on to. You can have an argument about it. You can have a discussion. Yeah. But that gray mass that is just passive, it's really- Yeah .deteriorating an organization. ### Speaker 2 · 45:45 And there's a lot of that. ### Speaker 5 · 45:47 Yeah. Yeah. ### Speaker 2 · 45:48 Thank you. ### Speaker 5 · 45:49 Let's fix that. ### Speaker 2 · 45:49 Tina. ### Speaker 8 · 45:50 Okay. Just to comment on that, there's probably some kind of a fear there- Yeah. Uh, which is hindering them from, from using it as the development goes so fast. And maybe on the bottom part there, we are all on different levels as well. Yeah. Uh, working against, uh, your AI with my AI, we have a-- we had a situation last week when I did a template with my AI and you did with yours, and we were prompting it in a, in a similar way, but still yours became better. Uh, and how that then affects the, the broader organization, uh, when, when it is AIs talking. Mm. Uh, so maybe there's some kind of a fear, not only with these individuals, but in the organization in, in, in a broader aspect as well. So that was one of my- Yeah. My hinders here. Sure. ### Speaker 7 · 46:43 Yeah. ### Speaker 8 · 46:44 Generally, the wins, using it i-in everyday life nowadays. Um, coaching, learning, amazing way of learning about new things and, and it can be tax reasons, it can be labor law, it can be whatever. Uh, so, so definitely part of my everyday life. And at the same time, I always want to go back to the, okay, the life is happening here- Yeah .between individuals, and I do understand that the coaching with an AI can be good and, and is good. And personally, I, I do want this interaction- Yeah .to be there as well. Um, so yeah. ### Speaker 4 · 47:26 Nice. Yeah. Oh, yeah. Yeah. Great. ### Speaker 5 · 47:30 I-I think you mentioned you on the, the IT, uh, kind of management of things, and we have a situation with you, right? With, with Cowork- Yeah .where it looks like since you are part of the HR organization, they're extra restrictive in giving you access to Copilot Cowork. Mm. I just think that's also something that there is still some kind of institutionalized- Yeah. Yeah .stuckness in things, right? Yeah. Where maybe the decision isn't always made where the decision should be made because it's IT who has a blanket no, right? And we need to challenge that, I think. ### Speaker 11 · 48:04 Yeah. You're not getting the same access or? ### Speaker 8 · 48:06 No, there's only three to four individuals per division who should get Cowork, was the answer I got when, when applying for Cowork. And also referring to my-- the sens-sensitivity of my role. ### Speaker 12 · 48:21 Oh. Um- But maybe they, maybe they thought it was Core Cowork. ### Speaker 8 · 48:25 No, no. Co-Copilot Cowork. ### Speaker 13 · 48:27 I, I, I-- Tina, you should have told me. I didn't know that. I thought we already had that on. Um, uh, that, that could be, of course, for me, it was a pretty quick and easy to get it, but- Mm .Cowork is still in a, in a test mode in the company. Yes. So we have a restricted, uh, small group now who's working on Cowork, already monitored, and I think there's a whatever, a community doing this. Yeah. And the reason is, of course, I think because it could be that we decide whatever to raise our content. I think then all is gone what is in history. It could be that we roll it out, I think. I don't know even the status yet. Yeah. There's- But I, I didn't know, I think, that it's three or four per division. ### Speaker 5 · 49:08 No, that's, that's what I got as- But it's actually also a cost matter, right? Yeah, yeah. It's also cost. Because Cowork is, is pay-as-you-go, right? Yeah, yeah. So, so we need to monitor that initially. What I'm saying here is that just because you're HR, someone at IT shouldn't be allowed to just say no to you, right? ### Speaker 8 · 49:25 That was one of the arguments. Yeah. It was more about the- Yeah .the size, the number of, of users. ### Speaker 13 · 49:30 So it's not- Let's- Not everybody has yet Cowork. ### Speaker 8 · 49:33 Everyone has except for your team. Everyone else has except, except me. ### Speaker 2 · 49:36 Everybody has Cowork now except of you. ### Unknown · 49:38 Okay. ### Speaker 2 · 49:41 Right. I'm curious to hear from Simon and Robert. Simon, what do you, what's your, what's your AI wins? ### Speaker 9 · 49:47 The, um, a lot have been said. Yeah. I think one AI win, uh, that will be really great going forward now is to use the AI for translation processing- Mm. -external communication. Yeah. Because that had historically been like a big, big, big hurdle- Mm. -you know, translate the web, getting, uh, uh, team members internally to proofread, et cetera. And now it's started to become so good, so I don't think we have to do that. It's- Yeah . been much, much easier. But then I think other AI wins in the, let's say, a little in the marketing team is, uh, at least when you have already good material to work from, it's really good in, we call it, like, repurposing to make LinkedIn posts or write- Mm . an article on this. Uh, we've tried to create like, um, iconic marketeer that only looks into approved material, like in our sales portal and on the Mycronic web, because there we have all the release material. ### Speaker 2 · 50:49 Yeah. ### Speaker 9 · 50:50 And, uh, that I think will become a really, really, uh, good. And, and being able to, uh, let's say, push out more stuff and communicate more. So that is really good. ### Speaker 2 · 51:02 Communicate more, yeah. ### Speaker 9 · 51:04 And then also I think it has, uh, become even better. I think it's really good also in becoming a good writer, especially if you have some material to start from or- Yeah . a PowerPoint, et cetera. So I mean, like the lobby news or other communication, it's, uh, yeah, really good. ### Speaker 2 · 51:22 I think it's like. I mean, if you compare it to beginning of the year and now and how well it writes and finds the tone and how well you can kind of put structure around it, so it doesn't feel like it's an AI repetitive force, but it's, it's, it's unbelievable. ### Speaker 9 · 51:40 Uh, so what I, I think we still struggle with a little bit, but it's, uh, like, uh, you said, Göran, like, uh, we're playing around with AI internally in Mycronic, but the IT infrastructure is not there yet everywhere. ### Speaker 2 · 51:54 Mm. ### Speaker 9 · 51:54 We have access to, you know, the data in Azure or in M365, but, uh, we haven't really opened up the AI to use our systems and the data- Mm . in our systems on, in the. What we have now, the data platform that we built, uh, they talk about doing this MCP so that- Yeah, yeah . uh, but that will happen next year. It's not there yet- Mm . et cetera. So with, um. ### Speaker 6 · 52:18 What is MCP? ### Speaker 9 · 52:19 It's like, uh, like an architecture to give, uh, the right, uh, people, uh, through agents access to the correct, uh, data that you should have- Connection between, uh, Copilot and M3, for instance. ### Speaker 12 · 52:32 Yes. Mm. ### Speaker 9 · 52:33 I mean, and, uh- And depending on- Connect like- Yeah. And depending who you are, you have access to certain part of the data that the agent have access to. ### Speaker 2 · 52:42 Thank you. Robert, final words. What do you wanna add? ### Speaker 12 · 52:47 Final words. Uh, I work, uh, mostly with Claude- Yeah . but then I work not every day maybe, but, uh, have Claude in the Cowork with a project. ### Speaker 2 · 52:55 Uh-huh. ### Speaker 12 · 52:56 Uh, and then I have helped to, to do like a wiki structure- Mm . in there. Mm. So, so that, uh, I can put mails or, or lists of, like we have in this project, list of, uh, employees or whatever in the inbox, and then I ingest it, and then it makes, um. I can keep the data. So Claude is good because it's more contained into a specific folder. So it doesn't go everywhere. That's both a pro and a con. That's a difference to Copilot in a way. And then that's really. I mean, that's super useful. So I can. Like Clemens for instance today, he's going to meet, uh, with a lawyer. What are the different things he need to know? For instance. Uh, it's a bit talkative- Mm . but I told it that. ### Speaker 2 · 53:42 On purpose. ### Speaker 12 · 53:43 So, so I, I can get like a gross list. So we can see, okay, what are the risks here? And we can do a lot of things. ### Speaker 14 · 53:50 Mm. ### Speaker 12 · 53:51 Um, and then, uh, so that really works super well. Um, I also learned how to do, uh, token management on that part. So not working, updating a PowerPoint all the time, for instance, but working in, in, uh, text form- Mm-hmm . going until I'm satisfied to do the revisions. Uh, could be five, six revisions before I'm. Because you always forget something. Yeah. I didn't want that. Yeah. So that really helped a lot on the token, uh, um, usage. I think it's. The difference is huge. ### Speaker 2 · 54:26 Nice. Mm. ### Speaker 12 · 54:27 Um, and on the team, uh, and on the team level, we have a new release of the software now that is, uh, uh, coming out. Uh, and I never seen so much content because the team uses- In a good way? Yeah. So we have GitHub, and then, of course, I'm in R&D. So, so, uh, we have GitHub, and then they have all the different models that they can choose from. Mm. Everything. Uh, a- and, um, it's really helping their throughput. ### Speaker 7 · 54:55 Mm-hmm. ### Speaker 12 · 54:55 So the next release, for instance, you will be amazed how much stuff is in there. So the productivity is really increased as a result of AI. ### Speaker 17 · 55:03 And the testing, the test first, the- Sorry, what- . automation. ### Speaker 12 · 55:07 The, the testing is, uh, uh- Of the software . uh, the upstream testing is done with AI as well. Mm. They write the, uh- Oh . test automation and everything like that. So, so it's, you know, the quality should be better. Yeah. That's it. Yeah. Um, should be better. Interesting. Yeah. And, uh, we also released, uh, AI team that I have released a, a software that was done also extremely quickly because they had AI to work with a, a open tool, uh, for inspection product. ### Speaker 5 · 55:40 Uh- So you're building AI with AI now? ### Speaker 12 · 55:44 Yeah, yeah. ### Speaker 5 · 55:44 Yeah. Yeah. ### Speaker 12 · 55:46 And, but also in general, token management is really important. Mm-hmm. We're watching this every day to see, you know, some, some people use it a lot. ### Speaker 6 · 55:55 Mm-hmm. ### Speaker 12 · 55:56 And then, then we'll say, "Can we do something?" I mean, we don't want to hinder that, but also it's not going to be too much- No . because it costs, right? So we need to understand also in budget time how much budget do we need for tokens- Mm-hmm . ### Speaker 1 · 56:11 uh, next year. ### Speaker 12 · 56:12 And that's not gonna be nothing. It's gonna be a couple of million for a big production. ### Speaker 6 · 56:17 Mm-hmm. Yeah. ### Speaker 12 · 56:17 So we need to factor that in, but how to use it. So that's some stuff, um, uh- Some struggles . and then some struggles. We don't really have, uh, a good governance of all the possibilities with AI, uh, both internally a-a-and, and, uh, for the product. Uh, there's a lot of options. We don't really have. We have a little bit of a backlog for the AI team, but we need to work on that a little bit more, I think. So Göran and I will sit down with Romane next week and discuss it. ### Speaker 7 · 56:50 Mm-hmm. ### Speaker 12 · 56:50 Uh, we have a little bit of re-resistance of recording meetings. So that, that is also something that, um, we talked about that a couple of times, and people say that it could be that you don't want. If you're recording, people are more hesitant to express ideas and concerns and, and, uh, you know, just speaking. So, so we're a little bit cautious with that, um, right now to do. We-- If somebody's presenting something to an audience, then it's okay to record because that's anyway more of a one-way communication. But if it's, uh, a meeting, then, then we're a little bit hesitant. And then, um, uh, when I worked with Claude, I, I changed. I wrote, you know, "The, the date is," you know, fixed date and so on. Then a couple of weeks later, I said, "We need to change this date." And then it got really annoyed a-and irritated and a little bit arrogant, actually- . I think to question it. "Okay, so this is, uh, something you can change obviously, and, and why did you change it?" And I go, "Because of the flight schedule." And then after that, when I did the plans- Mm-hmm . it said, "Oh, well, we're gonna allow plus minus a couple of days here and here." How very gracious of you. ### Speaker 5 · 58:05 Oh, thank you so much. Sweet overlords. ### Speaker 12 · 58:07 I got the stamp like, uh, I'm a sloppy person- . and, and cannot be trusted and everything like that. So that's quite in-interesting. You know, it's like one strike and you're out. ### Speaker 9 · 58:19 Maybe it noticed your arrogance. ### Speaker 12 · 58:21 Mm-hmm. Yeah, I think so. Let's meet in Oslo. And then when, when I tried to build apps, um, yeah, with a pre-work on Claude, uh, i- like recording, uh, uh, what you do at the gym prep, that it, it's native in the phone- Mm-hmm . but then you don't get the data in-integrity and so on there. Yeah. I didn't manage to, to how to fix these sort of everyday things, uh, yet, but, uh, it's amazing what you can do. So I, I want an app that does this. ### Speaker 2 · 58:50 Mm-hmm. ### Speaker 12 · 58:50 Sixty seconds. ### Speaker 2 · 58:51 Mm-hmm. ### Speaker 12 · 58:52 And then it's there on the phone and I can- Mm-hmm . I can do things. It's, you know, quite interesting. ### Speaker 2 · 58:59 Thank you for that. Thank you all for checking in, and it's been a kind of a, uh, twofold intention with this check-in. I mean, part one is kind of landing in the room and what has happened obviously. But, uh, and part two is getting some clues for this afternoon when we're looking at, okay, what, what, uh, what would help you? What's the pain points, uh, for introducing an AI as a, as a kind of a collective member of this? ### Speaker 9 · 59:23 But you are still stuck with PowerPoint because I think- Oh. No . CodePen really sucks. ### Speaker 1 · 59:27 Yeah. ### Speaker 9 · 59:28 I, uh, I've stopped trying. ### Speaker 1 · 59:30 Yeah. No, I, I had to. ### Speaker 2 · 59:32 We're gonna, um, we're gonna- Claude is very good. ### Speaker 1 · 59:34 Claude is good. ### Speaker 2 · 59:34 I just wanna say one, one last thing before a short leg stretcher and over to you, Jocke. But, but going back, going back to what you said, uh, around, uh, re-recording meetings. Uh, and also I think, uh, we are still as a s-society and lots of organizations around AI now is like looking at content. Is this AI? And if it's AI, it feels bad some, in some way. And that is something we need to have some sort of discussion around as a culture because probably all digital content will be AI made. So are we still gonna think that that is bad one year from now? Are we still gonna think even though we're all. I mean, we're moving into everything's gonna be recorded in meetings if we have a Teams meeting. Are we still gonna not-- or are we still gonna be hesitant to speak up? So it's, there's something around enabling or hindering as a culture. I understand, like, speaking up in a meeting if it's, like, recorded and it's transcripted and, uh, we, we miss some and people are hesitant. But how do we handle that? ### Speaker 2 · 00:40 Because that's not AI problem. It's like, it's a culture- Yeah . issue, right? Great. Leg stretcher, some, uh, coffee. Uh, is it okay if we do a quick one, maybe five minutes tops? So back at nine thirty-three.