Subscribe
Share
Share
Embed
A podcast about design, diving in with leaders IC and managers from the design world.
The Craft of Design
Gavin Elliott — Building AI-ready design systems at Deel
Transcript — final episode (39:47), including opening preview
[00:00] the quality of the output is really high. It's probably surpassed our expectations to be quite honest. So what does this mean?
[00:06] you don't want and what is probably not very good or useful when a manager does to you is essentially generate AI slow-free. Which is what would happen if someone just prompted an idea that they had and sent it your way, right? Like you really wanna craft it.
[00:23] Still needs human in the loop. Still needs human in the loop.
[00:26] And then we were able to pick from there, we could take 16 of them instantaneously and add them to the data grid and make improvements straight away.
[00:34] This makes it a hundred times easier, like faster than we've ever been before. And it It makes it more visual so we can actually track things.
[00:41] With the ability of the design system, and the new design system, which is more agent-ready, you can see where all of these mad things start attired together to put us in a world that we've never even, like six months ago, we wouldn't have been having this conversation. Like four months ago, even, we wouldn't have been having this conversation yet we're here now.
[01:00] so hello everyone. Today I have Gavin Elliott. He's a great friend of mine.
[01:06] But before we go there, maybe a little bit of background on your gap can you tell us how you got into design? Yeah, well, so 20 years this year actually. when I got into design. So my story is a little bit of a crazy one. I actually fell in the design after coming out of the British Army. I've been there for about three and a half years, but design had been something I always loved in school.
[01:33] So found my way back and into the world of web, not knowing really what it meant, starting off like most people do by doing logos and brochures and let their heads in that easy way. And then falling into the agency life, which a lot of us have of Dunn and Israel in the past, especially in those early days, and then getting through it, they're kind of a product SaaS startups, mobile agencies, and then all the way through to work in a government for about seven years, Shopify, big enterprise SaaS company, a little bit of a double back in the public sector again before joining Deal nearly a year ago, my deal of versories and on the thing in October.
[02:14] Tell us a little bit about maybe the difference between how things are operating in the public sector versus the private sector for people who don't know. And obviously like we're talking about UK private sector. So it's a lot of different.
[02:28] But the thing about government and public sector, they do very, very good at going really deep into specific areas, or you work in government, you will be doing a lot of research. You've got to get right up in front of the users that are going to be using the public services, you're going out doing kind of, you know, going to libraries, because it was the Department of Work and Penchance which I used to work for, you'll be going to various different offices, speaking to users, whether they'd be kind of my area at first was retirement, so you'd be speaking to people of retirement age, maybe speaking to people of working age and getting that level of knowledge about them on a very, very deep level, then help support them and all of their different needs that they have for the services that we need to provide.
[03:18] In comparison to private sector, especially kind of in that SaaS enterprise, the speed kicks up. You'll probably not go as deep, but the breadth and the scope of what you actually need to provide a design for or build, which was starting to gain a lot more these days, is incredibly exciting, very, very fast. And you get a little bit more initiative.
[03:46] It's like you can go and do the things that you need to do or want to do because you're literally two feet away from being able to kind of develop and go through pipelines and kind of get that deployment out to users. Whereas the government side is just slow. That's very cautionary. Could be talking about a project here and anywhere from seven months to go through what we would describe as a discovery and alpha, which is just the investigatory phase.
[04:14] If you said that to somebody in the private sector, that just the investigation is going to take seven months, it isn't going to cut it. So that's a dichotomy.
[04:23] - Well, it would be some of the things that you would have hoped like recently we had America.gov launching,
[04:31] they feel like they've locked the world like UK had that. I don't remember when UK.gov came live like many, many, many years ago, it took them a very long time get there. But of course, we can still improve and you've dabbled in it multiple times. Like what are the maybe top things that you want to see in UK.gov being improved?
[04:53] Yeah.
[04:55] I think part of it, especially having been on this side of the fence now, the speed to value to citizens of the UK is, you know, they deserve better because we are those citizens as you know, when we go and say something that's too slow, disconnected, you know, local government is still disconnected from big government and a lot of citizens of the UK don't realise that.
[05:20] So they'll not be able to perceive why they submit a form to their local council and then another government within, oh sorry, another department in government also doesn't have that knowledge. They say the giant numbers what should be real yet it's not. They couldn't be more disconnected if they even tried. So I think the speed to value for users and actually being able to like, you know, not have to spend three or six months on an alpha phase just to kind of test the water is like, take the learners from the private sector and apply them more quickly in the public sector.
[05:55] There's always going to be conversations about safety, security, privacy and everything within government. I still think to this day, and if not more so now, even with AI, it can still be dealt with and be managed accordingly and it doesn't need new slightly, it's been like how government employees should use AI, for instance. And some of the things on the list, you're like,
[06:20] come on, like how can the digital side of government do everything that it needs to do for citizens if they adhere to those rules? You know, it's just, it's putting its handcuffin folks. So I think the biggest one is that time to value, getting things out the door, testing even faster than we've ever done before, you know, in delivering the services that the citizens need within the UK, faster, better, quicker than we've ever done before to go out and ask.
[06:48] One of the ways that I've seen my mom used to be a CTO in all the biggest banks in Israel. And one of the things that I've seen that they've utilized to accelerate themselves is by off-shelf products.
[07:02] Do you think there's room or that for like an off-shelf like all governments you know use utilize this type of thing or is it a thing that the government has to build itself? So this is that's a deep question. So the government still do that. So UK Gov will and don't get me wrong, give us some context on UK Gov.
[07:22] There's central Gov, which is where Gov.uk still kind of sits. And then you've got all of the different departments, whether it be the department work and pensions, the driving vehicle license authority, you know, it's endless. There's lots of different departments and they all have their own budgets. They could buy software off the shelf. Yeah, I wouldn't say it's frowned upon, but because of the needs of government and from our perspective, accessibility has to be absolutely forefront.
[07:54] So they have to go through certain criteria, whether it be spend control, whether it be certain infrastructure that they can run on, is it within a single department state where it can be ran or can it be either offshore or localized to Europe and so on and so forth and the cloud, whatever it might be. So there are new onces.
[08:19] On one side, I do believe that certain systems should be wholly owned, wholly built by government. And in that case,
[08:29] public sector should absolutely learn as much as the possibly can and follow suit from what the private sector can and have done in many different verticals across the entire industry.
[08:41] When it comes to buying stuff off the shelf, there's still not opportunity, but it equally comes down to the rules and regulations around what they could and should be at the end of the day.
[08:55] And I think the biggest question is the what? Don't, you know, public sector shouldn't just buy just because. Like, if there's an opportunity to do something better or go more diverse, Do you need to try one specific system or could you trial three? Why aren't we given more opportunity to potentially smaller private sector companies from the public sector rather than just discounting everybody?
[09:19] I want to switch gears over to the world of the deal.
[09:25] What are some of the things that you are more excited about that we're working on that we plan to ship that really stretches what you are seeing nowadays without this like you know telling too much but like yeah.
[09:42] I think obviously because it's so close indeed to me is the design system work that we're doing. You know, when I joined a year ago, if I reflect back on just less than the year ago, whether the design system was even a earlier part of this year, with the way the world was changing, like everything had to change.
[10:01] I think with what we were doing with the design system, we had to know more about how it was being used. We needed to know more about where we were potentially headed of which that is like a role and target at the moment in terms of just worlds going and what we can do on a day to day basis to kind of go for that kind of big rock or end goal on North Star.
[10:20] So I think that for me because it and like I've said earlier this year design systems have never been important today than they probably have been over the last 10, 15 years and design systems have usually had that kind of same wrap of it's really important, it's really important and then it just becomes you know embedded in everybody's got a design system but now because of the fact that a lot of the private sector of Junoton, like Egenic Corden, or being able to move faster, design systems have literally never been more important.
[10:51] And what they do for agents and humans has never been more important.
[10:57] Practically, what are the couple of innovations things that we've done in this world that allows us to move? So maybe also tell people a little bit about the Builder flywheel and kind of like how how we approach like building projects engineering product and design wise Yeah, okay, so I think one of the big changes especially the beginning of the year was you know Most most companies are organizations all the way up to enterprise would even smaller would have a design system to some degree and Agents and models would use that design system to try and create elements of design in the best way it possibly could, but in order to do that in the context.
[11:42] And there's been a lot of back and forth on what that context is and how it's used and what can, what could be achieved with it. Now with where we started at the beginning of the year, we had a very established design system, a lot of detail, a lot of components, I think we're probably sat around 150 to 160 components with thousands of variants and you know various different to be all called using it together.
[12:08] One of the big things and the important things that we, if we were a flag back on of where we were earlier in the air was we had a design set, sorry, we had a design team to see it 100. We had a wider org of many thousands all asking questions to the design system team about X component.
[12:24] How do you use this? I need something a little bit more. And at that time, back in, if you look back on January, folks are still working in Figma. You know, some in a very small amount were starting to kind of either, you know, work with one of the tools like Claude or cursor at the time.
[12:41] But the design system wasn't ready realistically for one, we didn't really know how it was being used, apart from a lot of the data coming in through Slack where a lot of those requests were being done. We didn't really know where it was being used for an engineering perspective and whether, you know, who was detaching from it the most, the reasons why, what was missing and all of those types of learnings.
[13:00] but we were getting an enormous amount of requests. And on the other side, did we have enough context within the design system to support the future and talk about agents and mongs and things? If we just look on the analysis side for just a second, we hope, we'll name them, Pantellis did an enormous amount of work to put a system together to do a large amount of analysis design system to know exactly where things were detached, knew exactly which components and patterns were being used and which ones weren't.
[13:39] So we could really hone in on this area, we can completely delete it. We don't even need those components anymore because it's not even being used in, say, 80 months. So we don't even have to focus on that, and that makes us hone in a little bit more. And then from there, it was like, right, so if we've got the core design system and then we're, for instance, we're used supernova and storage work, does the actual design system have enough context in the moment to make the right decision.
[14:06] And because everything was new, we've got designers using a various type of prompt to start things off. The prompt is then trying to connect through an MCP to DLY and then hopefully put something together.
[14:24] Depending on the prompt, depending on the model, depending on the quality of the design system, there's variable results from bad to medium to better, basically. But we didn't wanna just do medium and we definitely didn't wanna just do better. So we then, well, rather the design system team set about how to actually create our better. We had a North Star, we knew what we needed it to do, we knew what the problem was.
[14:49] When agents for instance were creating design based off of prompt, it was possible, but it wasn't good. So we started looking at how we can make that better. The time was like designing D-files that was creating or trying to provide more context in the moment through MCPs. It would get you so far. The design system team and Pantellus and Paolo and others came about with this.
[15:16] What was the theory at the time? Because we didn't really know if it was going to work called the crafting table. The crafting table basically breaks down all of the context that an agent would require, a model world require in order to kind of do good design in its various different snippets let's just say of Jason D.A.
[15:34] and MetaData. So you'll be going into like very specific content metadata. You'll be going in accessibility, you'll be going into rules, you know when something should be used against when something shouldn't be used and being very very specific across every single component within the crafting table and equally all the way down the tokenization and things as well.
[15:56] We've now done benchmarks because that was a big part of it as well in terms of valuing the output and within the last couple of weeks those evaluations and benchmarks are being run and they even just now so we're not at the north star yet, but the quality of the output is really high. It's probably surpassed our expectations to be quite honest.
[16:16] So what does this mean? This now means that the Builder Flywheel which is essentially is there a ticket that can be picked up from Gira, which isn't, it's not it's not that scary, it's just either a little bug, a little bit of a UX issue that an agent can pick up, understand it from a PRD or a ticket in Gira and connect in through with the design system and the principles, and everything else in the context, and do a very, very good job of being able to understand the problem, being able to come up with the design criteria for it, put it together, build it, and have it accurately created using the design system.
[17:01] That whole flywheel now towards an output is doable. So now it's about how do you kind of turn the screw to make a better and better and better over time. And that means, you know, giving it more guidance, more direction, better metadata, better context, more rules or better rules, and then measure. To each one thing, measure it again.
[17:23] To each another thing, measure it again, see what the differential is. And then hopefully, you know, as we probably will be saying,
[17:31] I percentage marks, you know, if it's safe, we will start off at 60 now. It's highly likely you without within the next four to eight weeks, I will probably be saying nine years.
[17:46] Still needs human in the loop. Still needs human in the loop. Yeah, I was just about to kind of like go into this like where where do you see the human involvement necessary and
[17:58] where did you see the actual improvements like what are the things that are easier to improve and what are the things that are harder to improve in lagging in design system with code and it spits out the right thing.-Yeah, it's a good show. You've got to remember that an agent or a model is still just a machine.
[18:16] It acts in a very mechanical manner. So it won't have like a subjective response that we would like, for instance, it's doing an analysis or an assessment on the way something looks based off the viewport that it's looking at at the time. It won't go, or hang on a second, we actually, that would never be used in that viewport.
[18:36] It would always be used in this viewport or whatever it might be. But it's judging it going, but hang on the, for instance, the chipstone show, but it's obviously that the chipstone show because it's on the wrong view for it, where there would never show, but it doesn't know that. So it then don't grades the score. So that requires a human in the loop to go,
[18:56] that wouldn't happen, like that's subjective. So I need to be able to provide the context for that and we can add that into the system and give it that knowledge. There are like principles, like literal design principles that we have within deal. It will be interesting to see the effect that the design principles have on the Laflè wheel and how it brings in those principles, noodles them and then outputs them with the design system.
[19:25] It could get them right, it could make it them wrong. That requires human and loop because it needs an actual human to say that's a bad decision we would never do that because of X, Y, and Z principles model when we get a so far and because it's still a machine, it's thinking in a very mechanical way.
[19:43] Where is it just needs that human and human and we need to evaluate more as we kind of progress. So it's a very, very interesting time.
[19:53] So right now we have, like we talked about the development, the actual development process. And in In addition to the development process, there are other processes in the Builder Flywheel, like the Discovery process, the prototyping process for a designer before it does thing. Essentially, not everything needs to hit code immediately and gets deployed into production because we need to know if we actually want to even build this thing.
[20:23] I want to share something here to show you something.
[20:31] Let me do this.
[20:38] Over here, we have a conversation between a couple of people where we have a prototype that was done over here by Maria. And she kind of like goes and explains the specific project. And then the designers and sorry, the engineers and the developers decided, let's create like a code channel. So they invited Cloud to it. And then Cloud went and basically tried to assess the code.
[21:07] Maria tried to give it Figma, but that wasn't successful because apparently it cannot read Figma from within Slack. So blah blah blah, some more conversation. She basically decided, "Hey, I'm gonna give it to repo pieces." And it went and was able to get to the repo. And this is our playground repo, which means it's not like live in production.
[21:28] It's not like working directly with the code, but he was able to decipher and bring all of this together, as for clarification, Maria clarified, everything it needs to clarify. At that point, like the front-end engineers started joining, And then the back-end engineer started joining LongStoryShirt. We're liking final mini corrections. And this halting is done just like that in Slack, where a designer started essentially from a video and started in a secure place inside the ecosystem.
[22:05] This is one way of doing this. We also have heard recently, designers fun came up with this video. Claire Wu was talking about Claude being really good with HTML CSS. So she's just letting it explore like 50 options, pick like three, try to merge them, then give feedback, et cetera, et cetera. How does that impact how designers are working?
[22:31] How does that impact design system? Because like this is kind of in the design system, but slightly outside of the design system, because I just went and said, hey, give me 50 options. and I just picked in every one of the 100 and almost 20 designers that we'll have here, we'll be able to do that.-Yeah, I think, you know, it's interesting.
[22:51] I got a question on LinkedIn, I think it was two days ago about how we should design this. And this was from a more creative designer as well. How should designers utilize AI in today's world? And I did say to them, it's part of the creative side of design. I was like, it can be absolutely be an ideation partner.
[23:12] So it can go away and give you as many possible variations or different ideations of a design. Fantastic. I think when you get into the more enterprise aside, the ideas are great, but the product still has to be consistent because there's more up play than just an ideation. You could have tens of thousands, or hundreds of thousands of customers, and they need to have a consistent experience.
[23:40] We still want that creativity from all of the design team, but just remember to boil it in the reality of where we are in the design system. We need, you know, you as a designer in a different part of the org, need to understand the limitations of the design system. We don't, because we are within the design system team.
[23:57] So we know that a component might not be as good as what it could be, but we equally want all of the designers from around the org to have a piece of the pie. Like they, you know, they see the paint, they go through the paint of, you know, trying to utilize a component that might just be a bit too niggly, might not fully meet the user need.
[24:15] So go away, idea, bring it back to us, show us how great it is, you know, validate it with users, fantastic. Let's get into the design system. Now, whereas it would have been like, hard to get a new component into the design system, hardly time-consuming, let's just say, it could take a couple of weeks, not anymore.
[24:36] So, you know, the ability to then be able to kind of integrate things into the design system can be far faster than ever has been, which makes it even better because once we've done that work, you know, you don't have to wait three or six months, so you might have done how to do last year, you're talking three or six, this is an exaggeration, but you talk about three or six days, you know, of getting it literally into the design system to be then utilized by everybody else.
[25:04] But you as a designer and an outside part of the org can absolutely go and explore and validate it with users and do some research with a Pock because we've got the infrastructure like the playground. How about it?
[25:17] Go for it. I think that's probably always been one of the like limiting factors of a designer from not within the design system team because they've always kind of felt like They're at the mercy of what's in the design system. Largely, yes. Big components, data grade and things like that. You are pretty much beholden to it. But I think now, because of the opportunity with all of the tools that we do now use,
[25:46] and the speed that you can move, whether it be testing a new design, speaking to users, getting user feedback, understanding, and analyzing a certain design or a component.
[25:59] Designers now have way more of a piece of the pilot that have ever had before.
[26:07] So let's just have the opportunity today to just go and have essentially like a components in the design system that are not officially yet in the design system, but then you can accept them into the design system and we can, you know, create like skill or like a theme that basically makes. When you decide, hey, this is the thing I want to use, it essentially bakes it as a component that later on can surface to you and you can say thumbs up, thumbs down or like, hey, I need like you guys to change this or just change the things that you need in order to merge it with other components and then it will update the code.
[26:43] In order to get all of the work done together to this place where we are today, we intentionally backed off from that workflow so we could focus on everything else in order to kind of get a small agent ready than we've ever been. But yes, that is a massive opportunity for in the future without without a doubt.
[27:02] And to be fair, all of the work that we've done now then helps us in the future to do that type of workflow. And again, it could be through video Slack. You know, we're a remote company. We've got designers all the way from the, you know, as far as these can go all the way at the late west coast.
[27:21] So that can happen.
[27:26] In addition to that, you have your leading about 10 people at this point of time. Yep.
[27:32] You still have the time to contribute as I see as well. Quite often, I would say. Can you maybe share like one or two, first of all, ways that you do this and maybe share like a project. Yeah, yeah.
[27:50] But there was absolutely opportunity to kind of jump in and become a little bit more of an IC, but just do more practical work because I think stepping too far away from the practical side of the work can sometimes move you
[28:05] too far away from the tools and actually doing the job at hand. And I love doing it. So for me, in my process, part of my, well, the large amount of my role is or gate and directing, leading, managing folks and teams. But equally, I'll try and at least spend one to two hours a day of doing like, I see work on various different projects.
[28:28] Largely, that's been massively sped up or,
[28:33] you know, productivity gains through using the tools that we have in our disposal. Claude reuse within deal. I've been using it for not just Claude, but I've been using the Janet Corden for the last two years. I'm all in. I know what I'm doing. I can do it quick.
[28:53] And it could be, you know, we could be working on settings, we could be working on data grid component, we could be working on a left navigation, or rather we, I could be doing that stuff. You know, a lot as a design leader within Deal, I only have enough time or guidance to be able to kind of get it possibly towards the 60s or 70 percent, but at least then it's, you know, being done and it makes me feel more comfortable with the work that I can do and then I can continue to have strong conversations and understand what good looks like and set a craft bar and everything else that comes with it and then the rest of the team can take it on.
[29:30] They'll be busy with something else whilst I'm doing that so I can support work and get work done and keep things moving and go for that.-You mentioned something really, really important here. You said 60, 70%. And I think that is a key difference because what you don't want and what is probably not very good or useful when a manager does to you is essentially generate AI slow-free.
[29:56] Which is what would happen if someone just prompted an idea that they had and sent it your way, right? Like you really wanna craft it. This is like a good opportunity, maybe to look at something, maybe you can pull up the stuff you did on search or...-I agree.-Yeah, let me bring something up.
[30:19] In fact, I'll show you. To be fair, I'll show you, I think just took so much of the grub of the links.
[30:27] I'll be able to show you the 60 to 70 % and then where one of the team then took it from me continued through the like the 100 % getting the shift, which is data grid and it's no small it's definitely not a small piece of work.
[30:42] So data grid, the biggest, most utilized component inside of deal. It is everywhere and has lots of different users across different parts of the product. But it's because it's so big and so meaty, it's needed a lot of love over a bit of time. So when, earlier in a couple months ago,
[31:07] I went through, you know, the team were working on various other things, but I wanted to move along, data grid to get it started, basically, because we need to do a bit of an overhaul.
[31:18] Utilise one of the tools that we've got, all of their intelligence and channels and so on and so forth. I pulled together all of the questions, all of the problems that folks have with data grid. Analyze the whole lot, found out where the gaps were, spoke to various different members of the team that needed certain things, and then pulled them together into a request list, which have a little bit of a small number to go with.
[31:47] And then started working through like what types of decisions could we make in order to make data grid better? Some of it is down at typography, something, you know, some of it's down at Padden, sometimes it's down to a kind of margins and functionality and features and placement. For instance, on the current data grade, the toolbar, the data cards, the actual grid itself are all separate, but we wanted to kind of pull it together and bring it in.
[32:12] So we've tightened a lot up. There's various different requests in here, but I tried to document because I knew for a fact I was going to be handed this off to somebody else. So for me, just to kind of do some design work can kind of vibe code, it would never have been any help to anybody. So as a case of what could I document a note for somebody else to then take it on?
[32:31] All the way through to the design decisions that I'd made and the problems that we were trying to solve and why some of them might have been rejected and so on. Any of the motion tokens that we've included, any of the actual code for certain parts, any engineering bugs that start a short, that I just, I couldn't personally fix in the moment, which another designer could offer from an engineer, for instance, the requests of which there was many and then a comparison depending on what you were looking at.
[32:59] And then Rosina, one of the team members, she's a staff designer on my team. We had a conversation about what piece of work she's gonna pick up next. And I said, look, Z, I've been looking at data grid thoughts. It's just like, yeah, I might as well. There's a lot of things I'd say, right, I've got this huge, like, prototype set up on our internal infrastructure.
[33:21] It's probably about 60 % there. There's lots of niggles with it, but it needs to be, you know, developed even further. We need to continue some of the conversations that we've got with just designers from around deal. There's some things that they're going to request. You need to go a bit of back and forth with them, get some feedback, and then go from there.
[33:39] She's just like, "Fight, no problem at all." So she creates two two day grades one day a set. She's gone in even further now and essentially taking what I've had but we're finding it even more so that a lot of the functionality issues which were you know needed support have been now sorted. She's spoken to the likes of to your mess within deal and refine them even more and then set about creating a separate prototype which isn't on the infrastructure that I can show to have a conversation with the design system team to say, look, in order to make some swift improvements, which ones can we do?
[34:22] So she, you know, so you know, I created this entire prototype and down the right hand side, she had a new rail that basically switched on or switched off every single different thing that we've added into the data grid. And then we were able to pick from there, we could take 16 of them instantaneously and add them to the data grid and make improvements straight away.
[34:40] The next 10 might be a little bit more problematic or challenging, but we can do them and then we can backlog them and work with engineering. We literally just went through and went tick tick tick tick check check check right fine fine fine okay they can move into the data grid and then the utilized.
[34:57] It then enables having much better conversations around the org and it enables having really good conversations with engineering folks. So a lot of the functionality will now be to become the data And I didn't have to take it that far. You know, I didn't have to get it to 8, you know, 8, 100 % and kind of hog the whole thing.
[35:13] That wasn't the point. The point was to get this, a he-muth of a component started to get the wheels turning. So somebody else could pick it up and then keep the momentum, movie from faster and then get it through the delivery, which we've done.
[35:30] So that's just one example as well. is others whether it be search or navigation, which is a big one within the platform, obviously and work with all of the different areas, but having a prototype on internal infrastructure that you can share around and even have it somewhere where you can do user research with as well.
[35:51] Right, so this has had a couple of different core names, but let's just call it Jimmy Jango for now. [Sighs]
[36:01] There aren't as far as we know it. Tools available to do full end-to-end flows for mobile applications to have a level of detail that will be helpful towards the team.
[36:20] The mobile app for deal, even if you just look at Worker, is vast. Many different parts to it, many different use cases, many different things that could show up depending on the user and so on and so forth. And basically one of one of the team had an idea as to wouldn't that be great if and then we looked into external pieces of software that could actually do this.
[36:43] There was one probably a year or so ago which might was loose but wasn't exactly it wasn't exactly what we needed. So now, like this shows all of the different flows and you can go on for days by the way, they say every single screen and then dig into it a bit further. But if for instance, if I click on map, we can now see all of the different flows across all of the app.
[37:08] So if I just click on one, it then brings it up and then we can start leaving comments and say on this screen, there's this problem and then leave a comment and we can tag people and then they can come in and then we can go on with different one and then make comments again. And in the background, the scans going on to do to update this, I think it's maybe by daily, it's if not daily, to know exactly where we are at all times.
[37:33] And then if somebody does a report and issue or a new problem with the flow or something doesn't feel right, we can go in like this is the different types of users. So we can go in and say right for a contractor within Deal, these are the flows that they have. And then dig into them. Instead of having to go into Figma,
[37:51] good luck on even for the best Figma file, it's always a nightmare to go through. And that was one of our biggest issues, especially when we move more towards, you know, prototype and R using a playground for this design work within mobile. This makes it a hundred times easier, like faster than we've ever been before. And it It makes it more visual so we can actually track things.
[38:13] We're now looking at extending Jimmy Changa with more detail again. So imagine if we can get user research results or insights, which can apply to specific part of the flow and alert us where something might not be quite right or it could be a pain point, based on various different data points. We can instantly, from a single notification goal, a specific part of a flow, get the context of the whole thing, and then find an avenue to fix that problem that users might have.
[38:49] With the ability of the design system, and the new design system, which is more agent-ready, you can see where all of these mad things start attired together to put us in a world that we've never even, like six months ago, we wouldn't have been having this conversation. Like four months ago, even, we wouldn't have been having this conversation yet we're here now.
[39:08] It's incredible to what the team have been able to do. And you know, this is largely down to Shahi and the engineering team have put this together and then it constantly just keeps getting developed and more developed. And the design system team in power on Patelus and now Alex and other folks that have been involved from the engineering team.
[39:26] They've put us in a much better stead for the next months ahead. Yes. So hats off to them all. Okay.
[39:37] Thank you, Gavin, for this lovely conversation. I'm sure we'll have you here again very soon. Thanks.
[39:44] No problem. Thank you so much. Speak to soon. Bye folks.