Inside 9fin's company-wide AI Working Week

Charlie Cowan: to this week's Kowalah Wednesday Webinar. Good morning, good afternoon, depending on where you're dialing in from. If this is your first Wednesday webinar, we get together every week at the same time to spend an hour talking about something that is top of mind in the world of AI, and especially in terms of how enterprises are adopting AI. And this week, I'm absolutely thrilled, to welcome Jamie Chung, who I'll introduce in a moment from 9fin, one of our spotlight webinars, where instead of me just talking through what's going on in the world, we've got some real perspective on things that have been going on in other organizations. Now, I'll introduce Jamie in a moment what we're going to be talking about. We follow a similar format each week. So we'll spend… I actually think we'll probably spend about 45 or so minutes talking about our main topic here today, which is about driving an AI week and AI enablement week across your organization. We're then going to spend maybe 10 minutes or so talking about AI in the news. This is our recurring segment where we look at what has happened over the past 7 days, and what a 7 days it has been. So we've got some new releases that we're going to talk through from, from ChatGPT, and just last night from Meta, formerly known as Facebook, with some new stuff, which I think is gonna be really interesting to take a look at. As always, I've got the opportunity to ask questions. If you're watching live, then of course you can ask a question in the Zoom chat or Q&A. If you're watching on YouTube, which we know many people do, then just ask in the comments there, and I or one of the team will come back to you straight away with anything that you need to know. With that, let's dive in and introduce Jamie, our star guest today. So, Jamie is the Chief of Staff to the COO at 9fin. 9fin is a data and analytics platform that's focused on debt markets, so providing information to organizations that are investing and making investments in, the debt market. Jamie, maybe I'll ask you to introduce yourself, maybe a bit of your role, maybe there's a little bit of extra context you've got around… around 9fin and where you guys are on your journey.

Jamie Chung: Absolutely, thank you for having me. Thrilled to be on this webinar. So, to introduce myself a little bit, so, as Charlie, you said, I'm Chief of Staff to the COO at the moment. My background has been in management consulting, and then I joined the world of tech scale-ups and startups, and I've worked in a lot of AI-native companies, ranging from, in different verticals, ranging from healthcare and pharma. To now, obviously, financial services. My role at the moment is, by definition, quite generalist and cross-functional, so I tend to work on a lot of projects that touch upon multiple teams and are a priority to the exec team. Oftentimes time-critical, strategically important. And need a lot of coordination and rallying around the same goal to move us forward. So that's a bit of my role, and I became responsible for our internal AI enablement program in the beginning of the year, because one of our company OKRs was around going all-in on AI, both externally, but also internally as well. So that's a little bit about my background.

Charlie Cowan: Amazing. And, just, you know, where, 9fin is on the journey, I think it was March of this year, raised the Series C, and about 500 people, I don't know if that's sort of the accurate number today, I know it's growing very fast.

Jamie Chung: Yes, at the moment, today we're around 600 people, and we have offices around the world, including London and New York and Belfast.

Charlie Cowan: Fantastic. And you mentioned there, earlier on this year, you sort of took on that, that mantle around AI adoption. What was your sort of personal, sort of experience and usage with AI at that point? Were you a dabbler, or, you know, where were you personally on that journey?

Jamie Chung: I would say I was a regular user of AI, but definitely, at a shallower level. So I was using chat quite heavily, but not, for example, within Claude, not co-work or setting up skills or anything like that. And also, I think… you know, when I stepped into 9fin, which was in November last year, I personally was also a little bit unclear on, you know, is using AI a good thing for work, or maybe not? Is it frowned upon? It seemed like it was becoming inevitable that everyone needs to embrace it, but I think at that point, I hadn't necessarily come across, you know, an organizational stance, not just at 9fin, but also other companies, where you know, we were actively encouraged to use AI or not. Now, obviously, the picture is quite different. I think everyone is expected to use AI and also become more and more and more fluent at it. So that was kind of my personal baseline, I guess, when I was starting.

Charlie Cowan: Yeah, fantastic. And then, maybe to set the scene of what we're going to talk through over the next, sort of, 30 or so, minutes, through June and start of July, you ran an AI week. and what we're going to walk through is a bit of the planning, the decision making, you know, what happened on those days, some lessons learned. But maybe to sort of set the scene of what we're going to walk through, what was the AI week? What was organized? How did it frame out in the company?

Jamie Chung: Yeah, so as I mentioned, we had a company OKR around going all-in on AI, and one of the ways in which we were going to measure that was by end of Q2, us, having… making sure that we had 50% plus adoption within the company, and a productivity metric by the end of the year. And my focus throughout H1 was very much on how we could drive up the adoption. And make sure that people were embracing, but also had the time and space to learn and experiment. So, in answer to your question, that was the broader context from which the idea came about. The idea was actually Steven, our CEO, who basically said: how about if we actually protected one day from BAU for everyone in the company to learn about and to experiment with AI? And… he… I think he was basing that suggestion based on the baseline survey that I had also run in Q1, where I wanted to understand where everyone was in terms of their baseline confidence levels, any sentiment that they had towards AI, what kind of tools they were using, how often they were using it, and where they felt they needed most hand-holding. And one of the key findings from that was people saw time as the biggest blocker. It was one thing for a company to ask its employees to learn and experiment, but then, you know, when business as usual continues so quickly, and you know, we are… we move quickly at 9fin as well, and so there's always things to be delivered and expectations, and when your time isn't protected from that. what I saw in the survey was that people found it incredibly difficult to actually set aside the time purely dedicated to learning and upskilling themselves. It was Steven's idea: how about if we took a day off from BAU and made that into a productive day where people actually learn and experiment with AI? I then iterated on that idea, because when I kind of tested the idea of taking one common day off across the whole team, it actually seemed pretty difficult from a business continuity perspective. We have, you know, teams that are doing time-critical work, that need to respond to market data and reports that are coming out, and it seemed a little impractical for us to actually dedicate the same day for everyone. So that's how, actually, the day got, prolonged to one week. So we…

Charlie Cowan: Hmm.

Jamie Chung: Segment the teams by one week, and that way we could ensure that there was some level of continuity in the background. In reality, I think it spanned over something like two and a half weeks, because, you know, our team in Belfast wanted to go beforehand, and so they were kind of like our pilot group, and then the team in London and New York went in the second week, and then I think the US team had a couple of spillover days in the following week. So, that's kind of how it ended up being, but the idea was to contain it within one week, if possible.

Charlie Cowan: Yeah, you know, what you're talking about there, about time, is something that we hear so much, about, that things are moving so fast, we're so busy with the day-to-day, I just don't have time, to, sort of experiment with these things. And even where people are, I don't know, learning in their own time, or their evenings, or watching a YouTube video, they don't have the time to then reflect that back into their own work. If I have to deliver a report, or a contract, or a proposal, then in the moment, it is faster for me to just do it the way I've always done it, than re… reworking that. I don't… we talk about the J-curve, I don't know if you've ever heard of this, like, the J-curve when you start a new job, or you get a promotion, but actually your productivity dips, and then you learn the new role, and then up you go. And we see this J-curve with individuals and with companies. Like, the moment of learning this thing, you see a dip, and when people have got that pressure on, they can't afford to take that dip, and so they carry on doing it the way they did, unless someone says, here's some time. To take that dip, and then learn from it. You mentioned, obviously, Steven, the CEO, was the one that was driving it. Was there any, I don't know, sort of pushback, or how was that received from other functional leaders about taking a day out of their team's, work?

Jamie Chung: I think the functional leaders were all very supportive of it. I think it was helpful for, you know, everyone in the company to know that going all-in on AI internally as well was one of our big objectives for the year. So, it didn't come as a surprise to anyone. And generally, when I looked at the baseline survey as well, there weren't that many people who were skeptical or didn't want to use AI. vast, vast majority, I would say 95 to 90, you know, 9% almost, of people were… and this was an anonymous survey, so they, you know.

Charlie Cowan: happens in a…

Jamie Chung: the incentive to, like, make up answers, but what I saw was that the sentiment was overall quite positive. I know that this is not the case in all the other industries, I think it's the virtue of us probably being a tech scale-up, and maybe we have a self-filtering process for the people who also are interested in joining an AI native.

Charlie Cowan: Hmm.

Jamie Chung: As a company. So, yeah, so I would say that the vast majority of people were supportive, and the only remaining question was how to logistically divide up the teams so that business could continue in the background whilst a portion of the team was, you know, having a dedicated day to learn.

Charlie Cowan: Also, and we'll, we'll get into the, the structuring of that, you know, in one second. I'm, I'm just thinking, as we wrap up the sort of the, you know, the, the coming up with the idea. You talked about, I think the OKR around 50% adoption. Prior to this, had you done any other, training, or what had been done before the AI week, that led to, let's try something different?

Jamie Chung: Yeah, I would say throughout Q1, the main focus was on increasing, like, basic level of familiarity and comfort with our enterprise-level tools. So we had ChatGPT… we have ChatGPT and Claude, both in our enterprise suite, and so we brought in yourself, to kind of have one-on-one sessions, and I think we had a series of three of them, and opened it up to the entire company, shared the recordings, just to get people logging in, to get set up, to get familiar with some of the, I guess. interface and vocabulary, because those things seem small, but when you've never logged in, I think it can be quite intimidating at first.

Charlie Cowan: So…

Jamie Chung: That was… so we held a series of, I guess, one-on-one, lunch and learn sessions. And then that was great, but didn't necessarily solve the problem of protecting time. And so then the AI week was meant to then go, okay, right, so we've now learned the basics. There is every willingness in the company to support you do this. We were also giving people budget to try and experiment with different AI tools that they came across and would be cleared by IT and InfoSec and legal. And so we were providing the budget for it, we were providing the training for it, now the last component was providing the time for it, and so that was to be a complementary, effort to what we had done throughout Q1 and Q2.

Charlie Cowan: Brilliant. But let's, let's, move on. So now we've got the idea, got the mandate from the, the, the CEO, and now it's like, right, let's start planning it. We need to schedule, we need logistics, we need lunches. So, let's start with: how long was it from when Steven said he wanted an AI week, to when we actually had one?

Jamie Chung: From memory, he said he wanted an AI day, and then I think it took me maybe a week or so to put together a proposal of, here's what I proposed for the AI working week. And then, from the point at which that was signed off by the exec team, I believe the actual implementation happened about three and a half weeks after that, so four weeks ago. So it was a fairly short timeline. I was conscious of moving quickly, because We wanted to make sure that, you know, we were getting people before they started going on summer holidays, and making sure that whatever knowledge and training we had provided in the lead-up to that wouldn't be forgotten, so we wouldn't lose momentum.

Charlie Cowan: That's cool.

Jamie Chung: It was a bit of a short period, and there was a lot of logistics that went into it, but ultimately, yeah, that was a timeframe, about three and a half to four weeks.

Charlie Cowan: Yeah. I mean, yeah, I mean, that's frantic, and you're talking about 500 or so people in multiple locations, sort of 3 where we were sort of in person, and then, other remote teams that were dialing in, so there's a lot of logistics going around that. How did you… you touched on this a little bit, about how one day became a week, but… And then became, 3 weeks that we actually ran that out. Maybe you could talk a little bit about that, just kind of the logistics and thinking about the different teams, and, how… how do you take 500 people and start carving that up into, individual people having this sort of personalized, experience?

Jamie Chung: Yeah, I think there are so many different ways of doing it, and it… to me, it was very clear, based on the baseline survey, that different teams had different levels of fluency, but also types of needs. So, I knew that, for example, our product team, when I chatted to the chief product officer. didn't really need a lot of hand-holding. They were kind of used to actually having their own internal learning cadence, and they had already been practicing taking one day off a month, to protect that time, and the chief product officer is in the room with the rest of the product team. you know, learning, coaching, trying things, you know, right next to each other, and I think that was very powerful. So, that was an inspiration for me, and I wanted to make sure that each team got what they were, looking for from the day. And what that meant was I needed to think about the trade-offs between having a cross-functional day versus a team-specific organization. And in the end, even though I think there is a lot of power to cross-functional events, because people… kind of come to learn that, actually, even though someone may be doing a very different job day to day, they have similar challenges and can, you know, do things with AI that is also applicable to themselves. But the trade-off that I made was, actually. I recognize that power, but I'm going to go ahead and organize it by teams, because I think that within the team, there will be separate themes, levels of fluency, and needs that the leaders have. So, I ended up having one-on-ones with all the functional leaders and exec teams, and understanding for them, A, what will create value for their team, and, you know, B, are there any constraints around, you know, the timing or how to split the team? So, for example, you know, content team is one of our biggest teams, and we needed to split them in half so that there was some level of business continuity in the background while the other half had that day off. And the question around, like, who, when, what, who gets to work on what, that was incredibly complicated, and so…

Charlie Cowan: Yeah.

Jamie Chung: doing that across the business, with, you know, anywhere between, I guess, like, nine to a dozen teams across the globe and across different time zones was, was challenging. And in retrospect, I wish that I had actually used AI more for it. Ironically, I don't think I was using AI enough to plan for this AI working week. But yeah, it was, as you can imagine, there were time zones, different team needs. logistics in terms of, like, what room people were gonna be able to go, and then obviously the trainers, who needed to be where, on which dates, that was all, very kind of customized. So that was a lot of… that was a lot of tailored planning. In the end, we ended up expanding it to the timelines that I described earlier, because for our Belfast Data Operations team, they… their members were traveling… going to be traveling quite a lot for a geographical expansion project that we had coming on, so they wanted to go actually before the main company week, and use that as a experience to give myself some feedback in terms of, you know, what could have been better, what worked well. So that was useful. And then, things like public holidays, you just don't think about taking those into account. but it also happened to be around Independence Day in the US, and I learned later on that actually a lot of people were going to be out of town because people were planning on working remotely in the lead-up to that public holiday, so then we have to shift some dates, and so all these little things that you just don't think about initially, I think come.

Charlie Cowan: Yeah.

Jamie Chung: Play a role in the logistics of it.

Charlie Cowan: Yeah. I mean, just on that, you know, segmenting, you know, we've seen it in a few different ways. So one is, team by team, like we did here. You know, another is, experience. So you ask people to sort of self-select, you know, are you a novice, are you medium, are you advanced? You know, our experience of that is. I mean, I think there's loads of studies about this, whether you ask people to, you know, rate themselves in terms of, I don't know, intelligence or something, the majority of people will rate themselves above average, which can't be possible. So if you ask people to self-select, they get themselves wrong about where they are, because they don't know what they don't know. And so we've often seen this, where people said, well, you know, I'm doing all this at the weekends, I know what I'm doing, you know, you can't teach me anything, and then you show them one thing, like, I don't know, voice, and they're like, oh, I didn't know you could do that. So actually, putting people into a team-based approach can work quite well. And then… Because they're in a team-based approach, and this is sort of lead-in to my next question about getting the functional leaders on board, if you take marketing, for example, you know, marketing are trying to do some very specific things, and that is around brand, it's around events, it's around design, it's around driving a pipeline. And so you're able to structure their day very much around the problems that they're trying to solve, rather than a generic, let's all learn how to build a skill, or let's learn what an agent is, but that agent is not actually relevant to what you're doing. And so, yeah, maybe I'll ask… I know in the run-up to it, we had individual calls with each of the functional leaders. Maybe you can talk a little bit about that, and why you wanted to do that, and how that impacted the week.

Jamie Chung: Yeah, I completely agree with you that each team will have a set of problem statements that they want to solve, and I think one example that comes to my mind is actually the finance team, where their AI fluency is actually higher than, I would say. In our kind of organizational map of AI fluency, and I wanted to make sure that that day, in order to be used most valuably, needed some definition up front. And so, rather than going into the day, blindly and just saying, okay, right, how should we use a day? I wanted the leaders to actually have a think about what they could use a bill time for. So first of all, the one question I asked them was. you know, what is your overall level of confidence in the team? And would you be looking for a structured training session, or would you be looking for a day to do your own thing and build what you want to build, but with some support from facilitators? Or do you want a hybrid of the two? And in the end, I think for most teams, they took that third approach of a hybrid, so in the morning. time learning the basics, and then afternoon time was protected for building. And I gave them some homework beforehand to think about, okay, if you could wave a magic wand, and you could automate something, or AI enable some workflow, what would it be, and what pain point are you trying to solve? And I think having that thought beforehand was helpful, because A I think it identified what technical blockers needed to be addressed. So, for example, if there was no MCP between, you know, two different tools, like Claude and another tool, then the they had to take that into account for what was achievable within the day. And similarly, like, if they were very sharp about the use cases that they wanted to tackle, I think they could probably come out of that session feeling really good about the fact that now there is a new workflow built, and that we can all benefit from this as a team. So, that was kind of the focus of my individual discussions in the one-on-ones, and I believe that that was essential for A, getting buy-in for investing the day into, learning about and building with AI, and B, gave me also the confidence and reassurance that the day would be used productively, because what each team lead wanted and needed was, was slightly different.

Charlie Cowan: I totally agree. Building up that direct buy-in from those leaders meant that as you're running up to the day, and then on the day itself, you know, they're in Slack, they're herding the cats to say, look, you know, we want you over here. And also, it was the… it was the carrot rather than the stick, so there were a number of teams who said, right, well, there are some people that do want something facilitated, but some just want to do their own, you know, work, and that's absolutely fine. The people that we want to be in front of are the ones that, you know, are asking for that help. And as long as people are taking that day out to do some, you know, building their own skills or whatever, then that's absolutely fine. And I think that, for a number of the leaders, was positive, that they didn't feel this was something being imposed upon them, but it was an opportunity, and they could… they could pick up on that opportunity if they wanted to.

Jamie Chung: Absolutely, yep.

Charlie Cowan: Anything, you know, logistical? Maybe we'll talk about the day in a minute, but… Lunches. I mean, I just… I keep saying lunches, because there were some amazing lunches, that was organized. I know you had Alex, your colleague, who was doing a lot of the logistics here. Was there anything outside of the AI side of things that, you know, was, you know, worked really well in the run-up that meant that when you hit the actual week itself, you were confident what was going to happen?

Jamie Chung: I think having a… like, a very, organized view of what teams are going when, in which rooms, which… with which trainers. It sounds simple, but because so many things were moving up until the last minute, that was, like, Bible to us, so every day we would post on the internal team channel. Hey guys, like, you guys are in this room and that room on this floor, and this team is in that quarter of the office, from… and your itinerary is here, here, everyone had their own separate Notion page, so Alex, who's our workplace manager, was so helpful in getting that all communicated, and he… also sorting out the lunches. I think the lunches sound really trivial, but I think it also adds to the sense of excitement around… it's not just some day where you show up and you get told, like, what to do, it's… it's a hybrid of learning, but also hands-on building, and also it's fun, because you get to… you get to have lunch and talk to your colleagues about what it is that you're gonna spend your time on, and I heard so many conversations in the kitchen about people talking about, you know, this thing that they learned, or how crazy it was that, like, this was, you know, the case, or what they were gonna work on, and so I think it sounds trivial, but I think it was helpful for creating that level of excitement, which I wanted every week.

Charlie Cowan: Yeah, it definitely shows the investment, that this is a thing, this is not just, here's some training, tick box, you know, compliance, you've got to do this, but no, this is an event, and it's something that, as a company, we're proud of, and I think, yeah, food definitely keeps people going there. So, then we come to the day itself, and, Of course, it was multiple days. I think we started off in Belfast for 3 days, then we did a full week in London, and then we went to the US. Let's just talk through… each day was a bit different depending on the teams, but generally, there was some sort of training and hands-on sort of enablement in the morning, and then generally in the afternoon was teams building things. Are you able to just talk through the, you know, the flow of that day, and how you felt that went?

Jamie Chung: Yeah, so the common structure was, and as you say, it really depended on the team, but for the vast majority of teams, they showed up around 9.30, and then the morning time was dedicated to, a facilitator from your team basically walking the participants through Common set of exercises, and slides, and prompts, and just getting a base… getting everyone to the same baseline. And then the afternoon was spent on build time, so everyone could identify a use case that they were excited about, and some teams decided to do that up front, did a lot of thinking up front, because there were some cross-team collaboration that were required within the same function. Some teams decided to leave it to the individuals, and the individuals said, right, I think this is going to be useful, so this is going to be my challenge, and spent the afternoon building that way. And then we wrapped up the day with, show and tell. So, between 4.30 and 5, we ran a half an hour show and tell where people could just volunteer and share what they were able to build, or something new that they learned. And then we wrapped up the day. So that was a… with a lunch break in between, obviously.

Charlie Cowan: Hmm.

Jamie Chung: was a common structure in the day.

Charlie Cowan: Hmm. Yeah, and I think I was part of some of the sessions, certainly for the sales teams here in the UK, and then in the US, and then I also ran the session for engineers as well, on, on Claude Code. And… what I find about the, sort of, the morning session of the setup is that, so many people don't know what they don't know, and so even some of the people are like, look, I've got this, like, why can't I just start building something? Can we start talking about some connectors, maybe. Oh, I haven't got my, you know, my Gmail connected, I haven't got my Slack connected. Oh, I didn't know there was this connector or that connector. So just, like, plugging them in so that they're clawed, or if a company's using ChatGPT, making sure that they're, you know, aware of, the other systems that are there. Then there are little things like, you know, in Claude, you know, what's an artifact? You know, what's a skill? What's a plugin? just this basic stuff that when, in the busyness of the day, people are so busy, oh, I use ChatGPT all the time, I use Claude all the time, do you use skill? What's a skill? Okay, right, let's, you know, let's introduce some of these things. And so, while sometimes the morning feels a bit of, you know, a bit more training-like, actually it frees people up, because then in the afternoon, they're like, oh, I saw a live artifact, and now I've got an idea of what I'm gonna build. The other… the other thing that sort of builds on that, I know there were a few teams, and we won't, name them by team, but there were some teams who, in the… in the pre-planning, the leader was like, look, I think we've got this, you know, I think it's fine, like, we know what we're doing. And then a couple of their team members had been to a… it was like an AI 101 that we did on the first day, and so some of these people had been to that, and they were like, I've just realized I don't know a lot of this stuff, and actually we should do that. And so we put in some, sort of half-day sessions for some of those teams, because they now knew what they didn't know, and I thought that was a good sort of validation of the, this learning, even when people don't think They need the learning, because they don't know what they don't know.

Jamie Chung: Yeah, I agree, I agree. The organizational structure I had described earlier, which was, like, different teams going on different days. was definitely the case, but to your point, on Monday, I had carved out that day as an optional kind of one-on-one, brush-up for anyone who felt that they needed to be brought up to a certain baseline before they participated in the team day. And I think, as you say, what ended up happening was those people went back to their teams and said, oh my gosh, today was so useful, that their leaders then went, oh, maybe can we get a condensed version of that in our morning session, so that was also part of the curveball, which made it challenging, but also more rewarding, because I could tell that, you know, the people in the Monday session were actually, yeah, excited about what they learned.

Charlie Cowan: Yeah. Definitely. And, I guess that feeds in… we've got a question here about, did everyone attend, or did some self-learn? So one of the things that we… we put in, we call it a Kowalah Bar, but you could call it, like, office hours, or a drop-in session. So for some of the teams that said, actually, I think we're fine, I think we've got this, we can do it ourselves, they had the ability on a couple of days to just drop in, and there was someone where they could bring their, their current situation, their challenge, their problem, or here's an idea that I've got. How did that play out? How did you communicate that? Did that, how did that work?

Jamie Chung: Yeah, so we… so we had this concept of Kowalah Bar, of, like, facilitators available to answer any questions. I would say it was… it really depended on the day. On some days, I think the trainers were flooded with questions, and I think in the end, we basically just allocated one facilitator to one team, because it was more efficient for them to, like, actually be in the same room and walk around. On other days, I felt that it was relatively quiet, mainly because people were kind of heads down and trying to figure out things on their own. But I think that… that ability, the availability of additional facilitators was definitely helpful, because the last thing that I wanted was for people to get stuck in the middle, and then just not have anyone to help them problem-solve that, or troubleshoot it, and then just give up. And…

Charlie Cowan: Hmm…

Jamie Chung: just say, okay, you know what, like, I tried, but it didn't work, so I'm just gonna stick to my current ways of working. So, the fact that there were people who were walking around, who were available to help troubleshoot, I think for me, gave me peace of mind that anyone could reach out to those people.

Charlie Cowan: Yeah, I spent a bit of time in one of these Kowalah Bars, and some of the team from finance came in, and we see this from all of our companies, I'm not sharing anything that's, that's secret, but you think about, like, a month-end process. It's the person saying, this is what I'm trying to do, this is what I do each month. And it… how might I do that with Claude in this example? And so, it's a bit of show and tell from them, of like, this is what I do today. you know, what might I do? Oh, well, you might create a skill that might help with that, or you might, you know, are you aware there are some plugins that you could use that would help? So it's just a little bit of this, you know, unblocking of, you know, questions, and having that, that works, works quite well.

Jamie Chung: Yeah, yeah, definitely.

Charlie Cowan: And then I… I didn't know if, oh, yeah, so show and tell, which I think is my favorite bit of the day, because, you know, at the start of the day. people are, you know, we've got this AI day, you know, this training, you know, I don't know what I'm embarking on. And then at… when it was, like, sort of 4 o'clock or 4.30 or whatever, and people come together, you know, the smiles on… on people's faces. So maybe, can you just talk about how the… how the show and tell happened? How did you organize it? You were the one, sort of, comparing it, and, yeah, what worked and what didn't work.

Jamie Chung: Yeah, ideally, I would have loved for everyone to come together in person, and to kind of actually do a show and tell for half an hour. In reality, I think because of the space limitations, sometimes it was just easier to do it virtually, so it was a mixture of in person and virtual. And, really, it was very open format. I basically just opened it up to the floor and said, you know, this is mainly for you to share something new you learned, or something that you built, so that other teams who also participate in the day can get inspiration, or maybe learn from what you did. And there's no expectation that anything that you built is, you know, perfect, and… running flawlessly. It's more just to showcase the inspiration and the idea. And so, teams actually came online and shared that. I think some of the examples that really stood out to me, just because of the immediate impact that it had, was when our data operations team engaged in the AI Day, they set up this, like, onboarding platform for their new joiners, and we were as I mentioned earlier, just about to go into very rapid geographical expansion in Chile and Malaysia, and so we had New joiners in both locations who are coming on board, and… we needed something to help them ramp up quickly. And whilst, without that day, I'm sure they would have had a standard training kind of material and practice that they were going to put in place. Using that day, actually, two members of our team were able to set up a platform that had practice guides, training videos, practice questions, and I think that was, you know, really impactful, because they were able to then use it straight away, I think within a week or two. With the new teams on the ground, and it helped cut down on ramp-up time, and it also got really good feedback from the team members. So, I mean, there are so many examples that came out of the working week, but I think that, to me, stood out because it was one example where, you know, someone, created something that had… that was immediately applicable to so many people on the ground, and then they were able to use it, and that had real business impact of cutting down on ramp-up time. And so, that was very exciting to see.

Charlie Cowan: Yeah, yeah, really good. one thing that sort of warms my heart when we do these sort of, show and tells is people that, at the… in the morning, would have considered themselves, you know, non-technical, you know, I'm not a developer, you know, I don't really know what's gonna happen, and then… once they've learned something about maybe Claude design, maybe they've learned something about live artifacts, maybe they've learned how to connect up their Claude to a few other systems. And then they go, well, here's this thing that I do, you know, every week, or as a team, we do this, and I've built this visual thing, and they cannot believe that they've built it. And for them to, you know, stand up in front of the team or on a Zoom and share that with people, it's… you see the pride in what they've created. And the thinking of, well, if I've done this in half a day, what else is possible?

Jamie Chung: Yeah, definitely.

Charlie Cowan: Yeah, yeah, and I, I think in the, in terms of the show and tell, and we, yeah, we see this across, across companies, there's, there's a bit of, You know, keeping an eye out for what people are building, and maybe you have to coax some of the quieter members of the team to show what they built, because they built something crazy, really good, and not just the people that want to put their hands up, so that's definitely something as a… sort of facilitate to sort of keep your eye on. And and then, yeah, keeping some kind of log of what people have shown, because you want to… These ideas are just the… the seed of something that comes into something later on. So, yeah, whether that's recording or using Granola or something like that, just to keep a track of what people are, are creating. And then, so that brings us on to our, sort of. You know, after the day, and… this is something that I think, you know, all companies need to be focused on, that, you know, doing one day, or over three weeks, or a training session, or whatever it is we're doing around change enablement. It's what happens afterwards, because if we don't change the people's behavior and the ways of working, then all we've done is take a day out. So, it'd be great just to get your perspective on it. Before I ask you that, I will just say, this is… I think the front line of what companies are trying to figure out right now, is How do we… like, track and monitor, like, the success of not just these programs, but AI generally. If we think about, Well, I used the finance example earlier around closing the books. at the end of the day, the books close themselves, or they don't close, and, you know, how do you monitor what the impact of that is on the business? You know, it is hard to do. So yeah, it'd be just great to get your sense of, you know, what happened after the week, and how you're thinking about it.

Jamie Chung: Yeah, absolutely. I agree with you that this seems to be what everyone is struggling with at the moment, so no one has a perfect answer, right? I think in a perfect world, every automation, every workflow, every agent can translate directly into a top line or bottom line, and you'd be able to demonstrate the ROI, like that. Obviously, in… the real world, where there are lots of moving variables, and especially for a fast-growing company like 9fin, where our scope of work is also increasing, and domains and geography, etc. It's… it… it is difficult to do that. what we're trying to do at the moment is understanding the impact of our AI investment in a couple different buckets. So one is you know, time saved, like, that's… that's, I guess, the easiest metric to kind of measure in some ways. You know, if a process got cut down from 5 hours to 1 hour, like, what does that… what does that mean? How many hours were saved, and what does that actually then enable for that individual and for the team? The second bucket is higher throughput, so, you know, something could take the same amount of time, but actually, rather than processing, I don't know, 500 data points within that same time, you're able to do 20,000, and therefore there is a X, you know, percent increase in productivity that way. The third one is around, being able to do… take on net new work as a result, so things that were just purely not possible, before you automated a certain workflow, you know, there's a question not marked around, like, so what are you able to now take on as a result of that? So they're just purely net new work. And then, the last one is around higher quality and error reduction. So, we have an internal example where, you know, someone in the content team actually created a way to automatically check all the editorial content that gets written, with an internal AI tool, and just to make sure that it's, in line with our style guide, for example, and there are no grammatical errors and things like that. And something like that can, shave off a lot of time, right? And so, what that enables is higher quality output, but also enabling our content team to actually go spend time on other activities that are a value add to the business. So that's kind of the bucket through which we're thinking about it. In terms of what happens immediately after the event itself, I think people probably needed some time to settle in, you know, what they learned, build and refine what they started to build on the day, and then think about, you know, and I think it probably went through some iterations as well afterwards. I know many, many examples where teams are still using what they built during that week, or an iterated version of that. And we're trying to create, like, a centralized skills library where that is all documented, and we're trying to measure the impact. It's not a straightforward exercise, as you can imagine, but that's kind of how we're thinking about the aftermath of the investment.

Charlie Cowan: Yeah, I, It's so true that it's so difficult at the moment, and especially with more and more people focusing on the cost of tokens, the cost of AI, that is making people think more and more about what's the R in ROI, so what's the return? How do we measure this thing? I was listening to a podcast episode a couple of days ago, my favorite podcast called Lenny's Podcast, which is great on, sort of, product management and careers and growth, and he had a guy, a guest on from Andreessen Horowitz, the venture capital firm. And we're talking about exactly this, this subject about, you know, the return on investment. And he was saying, you know, when you think about Sundar at Google, do you think Sundar wants to run a more efficient $4 trillion business? No, he wants to run a $40 trillion business. And so, when they're thinking about AI, it's about how do we grow, grow, grow. And I think this is something that I definitely share with clients, that it's… the focusing on the efficiency and the productivity is… it's the easier thing to do, because you're dealing with a known known. I know we're spending this much on HR, on finance, on sales, and therefore, if we could spend, you know, 10% less, 20% less, then I know that I can put that into a spreadsheet or a slide deck. What I can't do is say that, well, now that we've done this, our content team or our marketing team are going to be, you know, higher quality, more creative. In a year's time, they're going to come up with a new campaign that they wouldn't possibly have thought of before, and that's very difficult to put in a spreadsheet, but ultimately, that is where the gold is, because people They feel more capable, they feel more aspirational, they're taking on… Bigger challenges that they wouldn't have done before. It's just very difficult to put that into a… into a board pack.

Jamie Chung: Yeah, agreed.

Charlie Cowan: Well, and what I would say, on Friday, we're doing a wash-up for some of the US sales team, and for me, that's the proof in the pudding, is that some of the sales leadership were like, oh, you mentioned that it was Independence Day when we ran, Around the first session, and they said, well, some of the people weren't able to make it, and so, can we get them the same, experience. So I think that, definitely, definitely shows that on the ground, it, it lands, and people benefit from, from this kind of experience. Jamie, thanks very much. Any, I guess, wrapping up, any lessons learned, or anything that you would share if you were, speaking to yourself 6 months ago, what would you, what would you say back to yourself as, heading on this, this journey?

Jamie Chung: I would… even though we were able to, pull off a pretty successful event, I would, just for the sake of my own well-being, I would give myself a little bit more time to plan the week. Also, because I think that helps, just… align on the logistics side of things, like people traveling into town, you know, remote employees flying in to be there for the day, and booking rooms, and so on, so I would give myself a little bit more time. And what I would definitely do, again, is spending that… investing a little bit of time up front in that one-on-one with the different team leads to understand what they're looking for and what would make the day most valuable for them. So I would… I would repeat that, for sure.

Charlie Cowan: Fantastic, great advice, and thank you so much. Like, these, these Spotlight webinars are so helpful, because it's, like, real tactical advice that someone can, put straight into action, right now. And so thank you very much, Jamie.

Jamie Chung: No problem, thank you for having me.

Charlie Cowan: And so now, I will, move on to our AI in the News segment, and Tim, you can feel free to comment on anything as we go through. I've got three things that I think are really standing out, right now. So, just before we did, the webinar last week. ChatGPT, well, I said OpenAI launched a GPT-6 Astra, which is their new top-class model. And I certainly had an opportunity to play around with it at the weekend, and the main thing that I was seeing on my X feed, which is just full of these things, is ChatGPT's computer use, where it can basically take over your computer and start to work with applications on your computer. And what I was seeing so much of was 3D modeling with an open source tool called Blender, which allows you to create these 3D models. And I was having so much fun. I'll just go to my X feed here. All I've done is search for Astra and Blender, which you can have fun doing yourself. And then here is this guy who's built this really complex, model of a steam train. Great, 3D, and then as I scroll down, you're gonna see exactly the same thing. I know what this person's making a bat. Great. I've seen people… this is a guy building out a data center, with all of the right plumbing and everything. I've seen people making, PCBs, sort of like microchip cards. This was someone that, rendered their own internal, house. Pretty cool. And then this was, a doctor, an MD, who was, yeah, a single prompt telling Astra To make a video of a tendon transfer surgery that we do as hand surgeons. And so this has really opened up my… oh, yes, I'm down one more, because I'm a Formula 1 fan, for Lewis Hamilton's Ferrari. So this is really, you know, opening up my ideas of what is possible with a model. Yes, these are all done with Astra, but with Fable 5.1 on the Claude platform, you're doing similar kind of things. It just so happens that the latest one to come out is Astra, and so that is at the forefront. But this whole idea of using AI to start building these 3D models, CAD designs, is really at the forefront. Anyone that's still thinking that AI is just predicting the next token, and it's just a next word generator is missing a trick, because these things are able to be very, very creative. The next, release from, actually this was last night. And this is also ChatGPT, and they've released their new image model, which is Images 2.5. So, you may be familiar with just coming here, creating an image. I think I gave a little slide, here to show something that I thought was quite fun. it's rendering… well, A, it's a much better model, but it's also able to render about 50% faster, and it's much more capable at taking inputs that you provide it. So here, you provide an input of a dog, and say, right, I want to dress this dog up, like, I don't know what that is, Captain America or something. And then, so you can see, it's rendering that. So you're, able, if you're in a physical. sort of manufacturing business, maybe you've got retail stores, maybe you're in fashion. You're able to take pictures of models, pictures of clothes, and get them, you know, transformed in a completely different way. The other thing that I will have a go at doing right now, is they've launched a thing called Sketch. So, up until now, you either had to provide a input, image, or you have to just ask it, whereas now there's a thing called Sketch, where you can now sketch out, what you want. Now, I'm not on an iPad, so this is going to look really bad, but if I attempt… to draw the koala logo. Oh, this is not going to go very, very well. Okay, this is bad. And then a little circle there for the nose of the koala. Okay, that's horrific, but you can imagine if I was doing this with a stylus, it might look a little bit better. And then I can… it's gonna upload that. Can you turn this into, a logo, symbolizing the face of a koala. There we go. And so, I'm thinking of, you know, my daughter, who's passionate about fashion design, I can imagine showing her this this evening, and on her iPad, she's going to be drawing all of these different, you know, designs, and then ChatGPT's going to turn that, into either, in this case, just a logo, but you can imagine saying, I now want you to turn this into a, you know, a catwalk-style design. That's gonna crack on. So, in order to do that, all you do in your chat GPT, it's rolled out everywhere today. Just look for Sketch, and I think there's going to be loads of good business use cases for that. And then my final update, for this week, this also came out last night, and is only available in the US at the moment, but I think it's going to be really, really impactful, and you're going to hear a lot about it. So, Meta, formerly known as Facebook, obviously the owners of Facebook, Instagram, WhatsApp, have had a challenging journey into the world of AI. Zuckerberg has thrown an inordinate amount of money at the problem. They made amazing acquisitions of companies like Scale.ai. They are right at the forefront of some of the open source and open weight models. And last night, they've launched their first personal AI agent. So what is a personal AI agent? Well, you might have seen Grokbot that we talked about a couple of weeks back. This is where you are chatting to an application on your laptop, on your phone. And the agent is not running on your laptop, it's running up in, in this case, Meta Towers. In Grokbot, it's running up in Grok towers. And so you're chatting to your personal agent. Now, why I think this is going to be very interesting is that, if I click on here, and, not to make a… a huge generalization about my wife. But this is the question that my wife is always asking. I get… we've got 4 children. I get so many emails from the school about different children, who's got sports, who's got this, who's got that, I just, like, cannot stay on top of it. My email is just full of school admin. But on top of that, it's all the other stuff that's going on in people's lives, whether it's homes, mortgages, insurance, pensions, or all of that. And, what I think is really interesting about, Meta is that it's going to be plugged into the digital life that so many of us have, whether that's WhatsApp, whether that is Instagram, whether that is Facebook, and therefore, it's connected into your social network, your social graph, it's connected into your email. I think this is going to be something that is going to gather a huge amount of traction beyond a Grokbot, or a Claude-managed agent, or a ChatGPT agent. Where the majority of people are not using those tools. So, we'll share out, some of these slides afterwards so you can get an idea, but in essence, the meta agent called Muse. You're going to be able to connect it up to your email, your calendar, your Facebook, your Instagram, your Peloton. It's got its own browser, it's going to be able to go out and find, you know, cinema tickets and book them for you. It's going to have a feed based on what's going on on your life. You're going to be able to set goals and track what is going on. So, less of a business use case, but I definitely think it's one to watch, and I wouldn't count out Meta in terms of what they're doing on, in the world of AI. So with that, that brings us to the end of today's session. Thank you again to Jamie. Next week, we're back to our usual time. This says 2 o'clock BST, but we will be at 3 o'clock, our normal time next week. And we're going to be talking about Claudeforce. This was announced just a couple of weeks ago, a beautiful union between Mark Benioff and Dario. from Salesforce and Anthropic, there are a number of ways that you can get your Salesforce data into Claude, and we're going to be talking about that so that you can really start to use Claude as your, sort of, sales cockpit for your AEs, your sales managers and BDRs, and so on. So with that, thank you very much again, Jamie. I really appreciate you joining, and thank you, everyone, and we will see you next week.

Jamie Chung: Thank you. Bye.

Charlie Cowan: Bye!