The Rise of Vibe Coding & Replacing SaaS Tools - Inside Scoop e4

What you'll find in this episode:

→ Why a non-developer ops lead became a vibe coder in under a month
→ How SDH built Swirl, a custom agency hub replacing four separate systems
→ Real numbers on what it took: 65 hours, ~2,500 Lovable credits, ~$45k in first-year savings
→ Where vibe coding gets frustrating, and why the simple fixes are often the hardest
→ How a senior dev and design team keeps security, backups, and technical debt in check
→ What this opens up as a service SDH can offer clients

Chelsea Pagliuca is SDH's project and operations lead, working behind the scenes to keep projects on time and on budget, the team well-resourced, and clients happy. Over the last month she added "vibe coder" to the list, building most of Swirl herself with Lovable and Claude.

Ben (00:02)All right, welcome to Inside Scoop. We've got Social Design House's very own Chelsea Paliuca with us here. She is our operations lead at Social Design House. I'll let Chelsea do a quick intro. I'm Ben Visser, the founder and creative director here at Social Design House. Today, we're going to talk about vibe coding. We're in our studio. We run a design and development studio. And we've been tinkering around with replacing some of our external tools, some of the SaaS products that we've utilized over the last few years that we feel like we could build or maintain ourselves. some of these new AI tools have given us the opportunity. So in order to understand, in order to experiment, we've kind of taken on a couple of internal projects just to see what's possible. If you don't have curiosity, if you're not exploring or experimenting, then you don't know what you can and can't do. Chelsea, you want to give a‍

Ben (00:53)quick overview of kind of your role within SDH?

Chelsea Pagliuca (00:56) Yeah, absolutely. Like Ben said, my name is Chelsea and he did pronounce my last name correctly. So kudos to him. And I am the project and operations lead, also a Vibe Coder apparently as of the last month, which is exciting. So I work primarily behind the scenes on most of our projects, making sure that we are really staying on time and budget and that all of our team is well-resourced, making our team happy and also making sure the client's happy. So really kind of bridging the gap. between both sides of a project. And then I also work our operations side of it as well. So once again, working with our people, making sure their goals are being met, that their development path is created and that they're following it. And then also working on the billing HR side as well. So kind of touching all areas of the business. So I was not touching the technical development creative side until now.

Ben (01:48)Yeah, so I think one of the really interesting parts of what AI and some of this vibe coding in Lovable, which is what we're going to talk about today, has done is it's activated the opportunity for folks in Chelsea's type of role, who were commonly non-developer, non-engineer, non-designer, but more task-oriented, resourcing. It's given them the opportunity to activate a side of the brain that makes total sense, right? Chelsea probably has the same mentality as a developer or an engineer, but she's just using it in a different way. What AI does is kind of remove some of that, those layers to get into being able to code. It removes the barrier. And so now we can kind of activate a different part of her brain to provide some new tools for the studio and build things that she uses day in and day out. Let's maybe set the stage a little bit. As far as what tools we use as an agency, I think it'd be good for people to understand kind of our stack from a SaaS software perspective, like what we're using to operate and run the studio today. And then we can talk about kind of what our concept or idea was with, you know, this new tool that we're building, that we're kind of like creating on the fly. Run us through kind of, you know, day to day, what we use every day, which you use as a project manager as operations.

Chelsea Pagliuca (03:10)Yeah, so teamwork is our big one. That's our project project management system that we use. So that's what our clients see as well. That's where we track all of our time, all our tasks, plan our projects. So that's what the whole team is in every day. We also use HubSpot as our ⁓ CRM and then we also use Quiller for our proposal creation and management system. And then we do run on EOS and so we also have 90.io as well for our leadership team.

Ben (03:38)Yeah, so we've got essentially four systems and a few others. We have some tools to run our SEO. We have tools to run our social media. Those aren't necessarily the items. Those are pretty specialty niche products that we'll probably maintain. Will we think about potentially building our own versions? Yeah, sure. Why not? But for now, we saw opportunity within initially the project management software. So for me, as a business owner,

Chelsea Pagliuca (04:02)Mm-hmm.

Ben (04:04)I started with kind of this EOS 90 methodology, right? There's meetings, there's ROCs, there's, you you've got your goals and your KPIs and your metrics that you measure. And so we talked about, hey, could we replicate that and then merge it? Because we were tracking ROCs in 90 and we were tracking tasks and teamwork. Like, do they just need to live separate? Is there an opportunity to merge those together? You know, that's kind of how we got to the point. where we were like, maybe we could build something. So I took a little bit of time, very little time, and just put an infrastructure in place, like a concept, where I was like, could this work? And so I built out just a small framework that would replicate what we were doing with 90, replicate what we were doing on a real high level, rough level, what we were doing in the project management space to see if we could potentially merge those two concepts together and just add some leadership layers. and then allowed Chelsea to kind of dig in and build around. Chelsea, can you kind of explain how teamwork works, how we use it in the studio and maybe some of the pain points that you saw and kind of got us to the point where, hey, this would be nice to be able to build my own or to just do all the things I want it to do and not be beholden to a company spending up features.

Chelsea Pagliuca (05:21)Yeah, absolutely. So I actually like teamwork. I'm actually a big proponent of it. Over the last four and a half years, we've tested out at least five other project management systems and have always come back to teamwork. And they do a great job of building more features, keeping you informed. But the biggest issue is obviously it's not built for us, right? It's built for agencies as a whole. It's not tailored to social design house. It also has different levels and we're you know, not paying for the enterprise level because we're small. And so there's things that it has the ability to do, but we don't have access to that functionality. there are definitely, there are definitely pain points, like you mentioned one that's really simple and seems kind of silly, but there's no calendar view in teamwork for projects. So you can see your own calendar, but a client can't see their project on a calendar. They can just see it on a Gantt chart, which is great. And a lot of people like Gantt charts, but I would say that ⁓ most of our clients and our team as a whole doesn't look at a Gantt chart and is like, I completely understand this and completely understand where I fit into this. You know, they're used to looking at calendars. So that seems like such an easy fix. And it was in lovable, very easy to build, you know, just a one little prompt, but in teamwork, it doesn't exist, you know, and you can reach out to their team, which I have and request features, which is great. And they do have new features every quarter, but. You know, that's not one of them that hasn't been one of them. There's also other little pain points, like our clients use teamwork allegedly, supposedly, but it's something else. Yeah. It's something else for them to log into. And really the messaging system within it is the only thing that they use and they can do it via email, but it's just always been clunky. I don't think I've ever had a client say, wow, I love using teamwork and it's super helpful. There's just not.

Ben (06:55)Right,

Chelsea Pagliuca (07:13)it's not client facing as we would want it to be. It's not pretty. And we've tried to come up with other solutions to that and we have, we've built our own client portal on Webflow, but even that it's not within Teamwork, so it makes us have to sign in somewhere else. So anyway, something simple that Adam, our account and strategy lead was always annoyed by is that clients would ask for a feature or... know, give feedback within a message. And there's no way in teamwork just to simply create a task from that message. You would think there would be, it's very simple. You'd have to, you know, take the time to copy paste, open a new window to the project, add a task. And even though that's not, doesn't take a long time, it's just a little annoying. So there's been things like that where, that in swirl, our version, our agency hub where we have studio, which is, would replace teamwork. It has all those little things. So everything works a little more seamlessly together. So we build all of our projects in there. It's the timelines, tasks, logging time, comments on tasks for projects, which that's also something I think our new system will allow our team just to have a little reset on that because we also use Slack, right? Which is great. And we don't want to replace Slack. I love Slack. It's awesome. But trying to, how do we integrate that in with our new system just to make sure stuff isn't falling? behind or falling in the cracks. So as we built Swirl, trying to make sure that it's connected to the different Slack channels for each project and things like that. there's, yeah, think the bottom line is we can't tailor teamwork to our team, but we can create something that's tailored to our team. And that's really powerful.

Ben (08:51)I agree with that. And I think that's a big thing to talk about is, know, six months ago, we may not have been able to do this. This wasn't a, you know, a dot on the radar. was, it was something, it was systems that we had carefully curated over time that we had built integrations with that we had kind of worked our system around. And I think what's happening now is we have the ability to kind of flip that script and say, I'm not necessarily beholden to, you know, a system somebody else made. And the reason they can offer it at a reasonable price point is because it's made for the masses, right? It's got scale and you can offer a $20 a month seat per user because the same system is being managed and run by thousands, hundreds of thousands of other people. Makes sense. With Lovable, we now have the opportunity to kind of build what we want. So we didn't start this. I don't think we started this thinking, hey, we're just going to replace everything. We're just going to rip it out and replace it. It was more... We started this as an experiment because I feel like right now is the best time in the world to be curious. It's the best time in the world to pressure test a system. Can it do it? What is the output? Is it production ready? Can it handle what I want it to do? And I think everything we threw at it, it kind of kept coming back as, hey, this isn't really that bad. So I put the foundation in place. And ⁓ Chelsea, I'd love to. ⁓ just pick your brain on how you mentally came into this. So I kind of threw this at you. said, hey, I put this framework together. Here's my high level concept, right? I love this opportunity to take these four systems that are all kind of separate, that have kind of rough, like loose integrators that kind of work, kind of don't, and pull it under one system, one shared company database, one shared, you know, signed proposal to active client flow. that could replace what when I brought that to you, what was your thought when I said, hey, give this a go like ground level, because there's a lot of people out here that I guarantee are like, I want to try it. I don't know what it can do. Will it work for me? And you know, we're a small business, so we're nimble. We have the opportunity to be nimble and work quick. Not everybody has that. But I think our experience is valuable to others because because they can see what we went through. So walk me through kind of the beginning process and kind of how you got started with this.

Chelsea Pagliuca (11:08)Yeah, I think at first I was really excited to try it because this was an opportunity to try new technology and it'd be paid for by the business. So wasn't like I was paying anything out of my pocket. So wasn't losing anything. I also think as a project manager, we are always telling people what to do. And there's never an opportunity to actually do it yourself. That was something early on in being a project manager that really bothered me because you were beholden to someone else finishing something. Now that's very much second nature to me, but it was like opening the floodgates of, hey, just do it. And it was really exciting and do something. You can learn something along the way. I love learning. I know you know that about me, Ben. Like anytime there's an opportunity to learn, I'm going to do it and put the time into it. So AI is like the floodgates of learning and education are completely open. And it's like drinking from a fire hose. and it can feel really overwhelming, but it's so powerful. So when I first started, I was a little nervous. I was just gonna break something and that nothing would work. And that it would just, it would take me forever to get momentum and to actually build something. And I was shocked by how quickly I was proven wrong. Like within like two prompts, like something's created and is usable and actually can replace something. I mean, that's crazy. And obviously, I've done way more than two prompts now and spent a lot of time tinkering as Ben likes to say, we tinker with a lot of AI here. And, but I think it's all been worth it. I've learned a lot. I've been used, I've utilized Claude with lovable. I should say that as well. So I did try, I understood my limitations as someone who's not a developer. Like I think there's like two sides of that is like one, this, can do so many things now with these tools, but also understanding. that I don't really know the why behind the what of what I'm creating or why Claude tells me to do something. And so understanding the value of our development team and asking questions and understanding too that I'm never going to be a developer, at least in the next 10 years. I don't have a desire to do that. So there's a lot of jargon that's used or that's in my prompts. I'm like, I really have no idea what it's saying, but it's okay so far. It has turned out okay. So that's been also kind of interesting because there's a level of ambiguity, I guess, when I'm working this because I couldn't tell you exactly what's happening behind the scenes. Kind of trusting the product a little bit.

Ben (13:36)Yeah, 100%. I think that's a valuable conversation to have is, know, Chelsea, you have the benefit of having two seasoned developers, a Slack message away. You have three designers that you can lean on for UI, UX patterns, for design systems, you know, I kind of help facilitate the foundational creation. So I think, you know, as a studio, as a design studio and an agency that has

Chelsea Pagliuca (13:45)Yeah.

Ben (14:03)built brand identity that has built marketing websites for the last 16 years, that has built lead generation funnels, done CRM integrations as a studio that has built product and wireframed and seen the manual process and the creation of those. I think there's an extreme value to be able to lean on people with expertise. to kind of explain, as Adam says all the time, you don't know what you don't know. And with this stuff, it's so true. And things are happening, but you don't know how they're happening. And the benefit of having a team behind you, like ours, and we've facilitated some build outs for other clients as well. It's been really fun to take our expertise to empower a VibeCoder like a Chelsea, to make sure that security concepts are thought of, to make sure there's backups in place, to make sure these core fundamental

Chelsea Pagliuca (14:33)Yeah.

Ben (14:58)know, development processes are being followed, but not necessarily having to get in the nitty gritty of the build out and the code structure and the filtering and the data. You know, like that barrier is kind of gone, but there's still like checkpoints where we need to come in. Like right now I need to go in there and probably, you know, UI UX test this thing, experience test it and, you know, put my layer of design polish on top of it. But the beauty is Chelsea has been able to. create probably the most polished proof of concept I've ever seen before, like the amount of functionality, the amount of understanding to go, how long have we been working on this, Chelsea? ⁓

Chelsea Pagliuca (15:32)It's been about a month, like right around four weeks.

Ben (15:34)Four weeks? Okay, so what we've built in four weeks, I probably would say would have been easily a 12 to 18 month project. And that would have been really, really working fast with multiple engineers, a couple designers, multiple processes. And I say this as a studio that makes money and runs a business building these tools. So like this... fundamental shift in like what you can create is really impressive from a non-technical person. Now that said, right, there's negatives, there's cons to all of this, right? I don't understand the code base. I don't know where my data is housed. Like we do like as developers, but Chelsea building this, she's like, what is a super super base? What is, you know, I got AI, I don't even know what model I'm using. Not to discount what you're doing, but it's, you trust the tool. You said it earlier, you're just kind of trusting the tool and you're trusting the AI to make the correct decisions for you. And sometimes we know we got to challenge that every once in So the beauty of having a partner or facilitator is you can kind of, whenever you get a little nervous or you question, you can raise your hand and say, hey, is this OK? Or how should I think about this? And there's some value there. I think that's a good segue into talking about, we talk about how crazy this is, how amazing, how fast we work. But can you talk about where you question yourself? in this process you've gone, am I doing this right? Is this really the best way to do it? Like, were there certain points that you can think of where you just kind of questioned your validity of doing this? I mean, I'm sure there's got to be like this extreme imposter, and I have it too, this imposter syndrome of, know, am I an expert? Am I really an engineer? Because we're doing all the things that those folks are doing, but what are some of those pain points or like areas where you question yourself?

Chelsea Pagliuca (17:26)Yeah, I think for me, especially at beginning, it just felt so powerful that I just like continue to throw things at Claude to prompt and lovable. And it was just like a snowball, like so many features, many, like, you know, dreaming big. so things, you know, as a non-developer, like not understanding how those things connect and if things will break and Claude eventually was like, ⁓ we need to have this bulleted list of things not to regress every time we send a prompt. And I was like, ⁓ okay, interesting. Didn't know like that there's, know, that this is connected to this. And like you were saying, like to a database and all these things. So, I think at some point I was like, man, like this just might be too robust. Like everything it's like a house of cards. I think we've used that analogy for other sites that have been built on AI before. Like, you know, if, if I don't, you know, say this, you're not regressed with this prompt, is this all going to fall apart type of thing? I think not having that. base knowledge of a developer. think it's like pretty consistently. I probably feel that at least once a week, like, like, is Taylor going to get in here, our senior developer and be like, what is she doing? Like, what has she done? Like I have to clean all of this up. And thankfully, I don't think that's going to be the case. You know, it remains to be seen. It's not officially launched yet, but I think asking Claude the right questions and understanding that Claude also is not infallible. Like it does not, it does not. Understand what I'm asking it all the time I've learned to make Claude ask me clarifying questions every time I send a prompt I make sure to Second guess it a lot of times like you do have to treat it as you would like another human team member like pushing it to you be better honestly and just like that whole idea of trust but verify like every time and But it can be a lot right because I might send like a paragraph to Claude of my Opinion of a page that was built or something and it'll send me a ton of prompts to send a lovable But if I don't read every line, which I don't always because it's just too much I found that there's often little mistakes in there like okay Claude. That's actually not what I meant So there's a lot of it's like a lot of writing and reading which is more up my alley So it's kind of interesting so I do think it plays to my strengths like vibe coding does because it allows me to really just communicate in words which is, you my background is communications. That's also what I do a lot of my job. And so it is kind of interesting how I do feel like someone who's in, I know it wasn't really your question, but I do think someone who's in more of a project or operations role is actually really place your strengths because it is all about communicating with words. so, I mean, words are powerful and this is just another example of that. So I think actually it's a big win for communications.

Ben (20:08)That's a really interesting perspective because I do think, and we mentioned at the beginning, that this has really unlocked a different set of people. It's empowered a different set of people to be builders, to be engineers. And that's incredibly intriguing to me. It might be scary to some from a jobs perspective, and that's completely valid, 100%. I respect that. But you also got to see it as, know, golly, like, People who didn't have access to build are not creators. And what are they going to create? Like, how are they going to approach these things? You know, I did that AI hackathon up at Winthrop. And seeing how young developers, young engineers were attacking these problems using these tools was a completely different, you know, methodology than what we would have done in the past. It just had, you know, it was much looser. It was much more, you know, energy and... you know, what you could output quicker, how you could test proof of concepts faster was really, really intriguing. Let's talk about how you kind of got started because I think a lot of people gets, they hit a roadblock of like, there's like the speed bump in front of me. And I'm just a little like, ⁓ how do even just get going? You know, I kind of teed you up a little bit barely. It was, it was very small. It was a very small infrastructure that I provided. like, when you go to add a feature, when you go to like do what's next or

Chelsea Pagliuca (21:21)Yeah.

Ben (21:31)Do you have like a sketch? Do you plan it out? Are you just prompting with Claude to get what's next? Do you kind of have a mental list or a written list? How are you managing feature ads and some of that, you know, product management? Because essentially you're a product manager, you're a developer, you're a QA. Like you kind of now fuzzed all those roles into

Chelsea Pagliuca (21:53)Yeah. Yeah. Yeah. How I did it versus how I would suggest doing it is probably a little different. ⁓ Because how I started, I was just like, like, how does this work? What can I do? And so I kind of feel like I just word vomited to Claude about what I was looking for type of thing. And that probably wasn't, that was not the best way to go about it. wasn't very strategic of me. think I would have, I

Ben (21:54)So how did you get going, I guess is the question. Even better. Even better. This is what our yeah, this is great.

Chelsea Pagliuca (22:20)And I have since then taken a step back and like written out workflows and tried to actually like, give Claude more of like a visual of what, what I want something to do within swirl. ⁓ but the great thing about Claude is that I actually read screenshots really well. And so I have, and we'll continue to do so. I've, sent screenshots of teamwork and I was like, this is what I don't like about this page. This is what I do like. And then it would make suggestions based on that as well, which is really helpful. And then also had Claude, Claude does a great job pulling from Reddit. And so I had Claude pull from Reddit from like project management or project manager channels and ask, what does everybody complain about and with project management systems? And what does everybody say is the best about different project management systems? And so it pulled like a bunch of features that people consistently liked and people consistently didn't like. And so that was really helpful in thinking through like what we don't have, that what would be cool and what we do have that maybe is just bloat. And it just helped me think through it. didn't necessarily obviously just take those two lists and like run with it, but just getting different opinions and seeing what's out there was really helpful and really powerful. So I've continued to do that. Clawed design now also came into play the last couple of weeks. I haven't used it a ton, but I did use it on a couple of pages. the pages and asked us to give me a prototype to see. And once again, just shared with Claude and like chatted through, this is what I like about this. This is what I don't like. Also pushing Claude to say, hey, you're a senior UI, UI UX designer. How would you review this? Because Claude responds differently when you tell it to be like a different persona. So yeah, I think I would have taken a step back and thought through things a little more rather than just. I was kind of throwing spaghetti at the wall at the beginning without realizing how powerful it be and how quick it would be, right? Cause those features were built out very quickly. Um, but I mean, it has worked, but I would, I would suggest taking a little bit of a step back, but at the same time, like to play devil's advocate, not to be too scared of just trying it because it is. Yeah.

Ben (24:25)Yeah, there you go. Would you see value, and this is maybe something we talk about, like to be, I think one of the most powerful things is to build a proof of concept, to see if something can work quickly and then say yes or no, and then either iterate and push farther or say, no, this doesn't work onto the next. Would you ever like build a proof of concept, invest a small chunk of time to make sure it works and then just start over and then, you know, just say, Hey, I'm knowing what I know now based on what I've built over the last four weeks.

Chelsea Pagliuca (24:34)Hmm.

Ben (24:57)could I restructure or reframe this in a cleaner, more logical manner and just rebuild it from the ground up? You know, kind of take all that spaghetti that we threw at the wall and that has created probably technical debt. Let's not lie, just because this is AI doesn't mean there's not technical debt floating around in there. know, broken things, missing stuff. And you know, that's where you got to do code audits and security audits. do feel like... Like does that freak you out? Like if I was to tell an engineer like, rebuild this whole thing, like, all right, cool, we spent four weeks on this, but let's start over. Like how does that make you feel if I were like, just start over? Or does that like give you, does it excite you? is it like, I don't know, I'm curious.

Chelsea Pagliuca (25:35)No, know. I mean, no, think that now knowing what I know now, I'd be like, yeah, sure. No problem. Because I could create an MVP in a day. You know what I mean? Like, it's not like I'm spending four weeks creating the MVP and then having to start over. Like you just have to literally or even hours. I don't even think you would need a whole day to be like, okay, this can work. Let me step back and think about

Ben (25:56)Yeah, you could almost like reverse engineer it out of what you already have, tell it to, you know, identify all the problem points, the issues and say, strip all that out, fix all that. And then let's rebuild this sprint by sprint with the highest level of database consideration with, you know, all these things like I had didn't think about at the beginning. So I think there's something powerful there about the cycle of the build process where, know, you might build technical debt, but like the amount of work it takes to just start over and to improve. to be like, here's V1 of model one. Like when you release a car, it's, you this is the first model year. And then to create the second model year, takes another year, 24 months or, you but now it just takes a couple of days. So there's, you know, this cycle has gotten a lot faster. And I think the opportunity to clean up, fix, like your brain just has to work like that. It's not this traditional like long-term cycle. Like starting over, isn't like a freak out point. It's a, you know, it's the opportunity to say, Hey, nah, we can just build this better. And you know, based on what I already know. But it took you building on features and adding things to find all that. So that's an interesting place to be, because that hasn't been the norm for the last 16 years I've built stuff. It's just not financially viable to start over. But now it's like, yeah, I can start over. If anything, it might be better to start over, because I can clean it up. What we're talking about, these are some aha moments, but what frustrated you? What kind of

Chelsea Pagliuca (26:58)Yeah. Right. Right.

Ben (27:21)triggered you a bit or what have you found frustrating where you've hit a wall? We're kind of like, as the kids would say, we're kind of glazing these tools saying, they're so great. They're magic. Blah, blah. And to a point, that's true. But there's also challenges, right? There's also, you have to have, I think that engineer and developer mentality to be stubborn, to kind of hit your head up against a wall to fix stuff. Like where have you gotten frustrated and been like,

Chelsea Pagliuca (27:31)Yeah.

Ben (27:48)I don't know what to do now or like I thought this would be easier moments. Can you maybe talk to me some of those?

Chelsea Pagliuca (27:53)Yeah. Yeah. It's funny. You say those, thought this would be easier moments because I do feel like some of the easier prompts or updates were the things that Claude or lovable for some reason did not understand or struggled with. ⁓ which is funny. There was, one just color change on a dropdown that was trying to change color. Not a big deal. I'm sure our developer could have done it in a minute, right? It's hope. would think something super simple for whatever reason, it just took I don't know, like five prompts for us to get there, like with screenshots, with explaining... It's just fascinating. Like AI is not human, right? They do not have a human brain. Very clear. So just having to push on some of the little things just seem so silly. And there's been times that Claude has said something and I'm like, no, that is wrong. Like completely different than what I just said. And it's like, oh, you're right. So some of that is... It's just kind of funny. Um, and I think too, it's made me be like, okay, I need to read like every line of these prompts because stuff is getting thrown in there that I didn't say, or I don't want Claude and lovable both also assume things a lot. If you don't, if you're not very clear, like they'll just kind of do what they think is best, which may or may not be best. So some of that I think is where the developer, our developers will definitely come into play because they may have different opinions of how to handle something that I have no idea about. um, yeah. It's been interesting.

Ben (29:22)Yeah. How do you, so now that you're, you know, VibeCode your engineer, developer, product manager, like when things break, how do you fix it? Right? Because naturally you'd say there's a bug. I create a task, I assign it to a developer, the developer looks at it, the developer fixes it, tests it, flows it back to the team for additional QA, they approve it, and then it gets pushed live.

Chelsea Pagliuca (29:34)Yeah.

Ben (29:48)What is your kind of bug fix or kind of troubleshooting process currently if something doesn't work?

Chelsea Pagliuca (29:53)Yeah. Yeah. I mean, I typically just screenshot it if possible and explain the problem to Claude. And typically it, I usually ask Claude to act as a senior developer and treat Lovable as the junior developer and give instruction of how to fix it. And I will say like 90 % of the time Claude did a good job of suggesting something and Lovable was like, yeah, it's great. I'll do it that way. And the fix is there. I haven't, I feel like they've fixed almost all the bugs that I've found. But yeah, mean, it's super, super quick and easy. Within like two minutes, a bug can be fixed.

Ben (30:27)Do you feel like you can communicate with the dev team better going through these processes or understanding how software is built? Like, do you feel like that barrier? Cause I know like you've always been task oriented, project management, operations, behind the scenes, HR energy. Do feel like on a project level you can communicate better?

Chelsea Pagliuca (30:37)Yeah. Yeah, to an extent. mean, I still think there's just a lot of jargon that I'm like, I don't know what that means. Or like, I know to be dangerous, but also enough to confuse everyone. Confuse myself. So yeah, yeah, I think maybe as I continue using it. Yes, I think that's true, especially, especially as our team jumps in here and sees things like I want to I want Taylor to communicate with me about what's happening so I can understand.

Ben (30:48)Yeah. Right.

Chelsea Pagliuca (31:11)Like, OK, maybe next time we build something, we should do it this way and better understand the process.

Ben (31:16)Yeah, I was just curious about that. It just popped in my head because you're learning new skills. You're in this world way more than you ever would be. And you are reading those prompts that have this jargon in them. And you might honestly know more than you think as you get going. So let's talk about, obviously we didn't start this whole project like.

Chelsea Pagliuca (31:22)Yeah. Yeah

Ben (31:40)Rip out and replace everything. But as we kept going, we were like, but what if, what if, you know, could these layer on could, you know, when a lead comes in from the website, it automatically creates a deal in our CRM from the CRM. can copy them meeting notes into our proposal system, which will then dynamically create the proposal. Once that proposal is accepted, the natural thing would be to create a project in teamwork or a studio as we're calling it. So then that same database column just gets flipped to client mode rather than CRM mode. And then it, based on the project requirements, spins up tasks. And then, you know, I think we saw that flow. We saw the opportunity to build that natural one, you know, one after another, which before you could kind of do it with integrations, but it wasn't nearly what it could be, right? It didn't automatically take the users from the CRM and add them to the project and the project management, right? There's, all these little things that we can now do. So I think integration and, know, kind of single source of truth is some of the most powerful parts of what we're building and where I see the most value is right now I have a HubSpot instance with a company and client list. Right. I have emails, I have companies, I have companies assigned to clients I have. teamwork where I have a company and project list, which those two don't talk to each other. have Quiller who has proposals assigned to clients and companies. So, you know, like each one has the same data, but we're not sharing it. And if we are sharing it, it's through a loose connector or integration that might not necessarily work how you want. So, you know, I just say that to kind of... you know, mentioned how you can stack on and how to how we're thinking about these tools kind of internally as well. Where have you seen the opportunity to create the most kind of efficiency out of these tools? Cause you're within the studio, your role is about process. It's about efficiency. It's about resourcing. It's about task management from an agency perspective as a design studio, as an agency, as a company. And this obviously helps our clients too, because now we can do more for them more efficiently and produce better work. for them, like where do you see the biggest opportunity from an efficiency perspective that kind of layers on to those ideas?

Chelsea Pagliuca (34:03)Yeah. Yeah. I mean, I think like what you said, There's all these manual tasks that have to take place along the process, like at every single phase. And now some of those tasks are most of those tasks actually are going to be just automated, which is amazing because, you know, those tasks might take five minutes, but it's not really five minutes because there's the time that you have to switch your brain to doing that and maybe pause something else. And so really those tasks become, you know, 10 or 15 minute tasks because of all the movement. And I say that like from experience, because I do most of those manual tasks for the most part. So I think I had Claude run some numbers and even just with studio or teamwork replacement, it can save up to like 40 hours a month of just seeing all these little tasks of creating projects and adding tasks manually and adding time estimates manually. with our, with Swirl and having everything in one place, the projects, I mean, having those create automatically and having tasks create automatically. I mean, that that's at least a couple hours, depending on the size of the project. so, I mean, that's, that's awesome that that makes it so. I can look in there and do more quality control of like, okay, is this really the best timeline? Are these the tasks we really need rather than that being tacked on to the already two hours I spent on it, you know? So it, allows the human people involved rather than the AI agents to. QA it, which I think has more value often than AI. So I think it allows us to be even more efficient. And we have the opportunity to consistently improve upon the processes a lot quicker than before. you know what I mean? Like, cause last year we did a big process overhaul, like looking at all our processes within the agency, especially related to projects and which is so funny because some of it's antiquated even now just less than a year later. Yeah. And I've already had to think and step back of like, okay, I got to rethink those because I'll be uploading those into swirl. So that way that's where it's pulling from, like knowing what our processes are typically. And so just thinking through how that looks different with AI, how it looks different within swirl. But yeah, mean, efficiency thing. I mean, that is

Ben (35:56)Wow. Yeah.

Chelsea Pagliuca (36:20)you know, the love language of a project manager. So it is very exciting to me that they all talk.

Ben (36:26)Yeah, mean, yeah, and that's just something once again would not have been viable for us to spend our time to dedicate our developers to. And that kind of leads us into kind of time and the investment it's taken so far because this isn't free, right? I think a lot of people look at AI and I hear people say, the beauty of it is I can do it myself. It's going to be free. There's not a whole

Chelsea Pagliuca (36:38)Mm-hmm. Yeah.

Ben (36:49)to it. Obviously, we've built a relatively complex setup for smaller, quick one-off type things. That's true, right? 30 minutes, an hour of investment. But I think as you get into building a tool that has layers, that has real value within the studio, that has implications, if it doesn't work, it's going to mess up our processes. It's going to slow us down. It's got to be right. So you're, still have to QA, you still have to security check, you still have to do all these steps to make sure this thing's functioning. That said, you you can do it and 20 % of the time, you know, and somebody who, you know, like a Chelsea, you don't have to pull a developer off a project or they can continue to do client work while we focus on this other stuff. So there's a beauty there. Can you talk through kind of how much time Like you've spent, we touched on it a little bit with the four weeks, but like from an hour's perspective, how much time per week over these four weeks you've spent, how many credits within Lovable, because there's cost to using these credits. We're asking it to do a lot. So that compute adds up. Can you talk through kind of some of those numbers so people get a kind of an understanding of what something like this would take? And we're not done, right? These numbers were probably 75 % there. So these numbers will change.

Chelsea Pagliuca (38:02)Yeah, right. Yeah. Yeah. Yeah. So it's been around four weeks. I've spent around 65 hours working on it. And honestly, you didn't ask about this, but I think five coding is really addictive. some of those hours were just like late at night. I was thinking about things that I wanted to add. It's exciting and fun. Right. And so it feels I think I told you before, Ben, it almost like gamified work for me a little bit ⁓ because it wasn't something I knew how to do. And it was

Ben (38:29)Yeah, I know.

Chelsea Pagliuca (38:32)like instant gratification. but yeah, so it was around 65 hours. And then we use just over 2,500 credits for this project. So that's about like $780 of lovable credits. And, but like Ben said, we're not done yet. Like our dev team hasn't hopped in there yet. So I'm sure they'll, they'll spend time fixing anything that I messed up on. And we still have to, of course do, testing with the rest of the team as well.

Ben (38:57)Right, there'll be a little onboarding. I imagine, you know, we'll probably be in that a hundred hour range, Chelsea's time, a little bit of dev time. We're going to be probably a thousand bucks, 1500 bucks in credits. You know, we'll buy them in bulk right now. It's about 30 cents, you know, per 500 or something like that. It's or 30 cents or I can't remember. There's a, there's a correlation there between the credit cost first, all that, neither here nor there, but I'm guessing we're a thousand to 1500 bucks.

Chelsea Pagliuca (39:26)Yeah.

Ben (39:26)And the cost savings there, you for us as a small agency, do you have those numbers in front of you? I got them kind of rough in my head. Kind of like what we're going to save just from a subscription standpoint, because there has to be a correlation, right? Like there's a lot of tools out there that are really great. Like they're well put together. Like you said, teamwork is a good solution. It was a great value and price point for what we needed. And I don't mind spending that money, right? Like I'm not going to replace my email.

Chelsea Pagliuca (39:33)Yeah. Yeah.

Ben (39:56)I'm not going to replace, there's certain parts of HubSpot like the email send and all that stuff. I'm probably not going to touch that. I'll probably just more replace the CRM because it makes sense within the funnel that I have now or within the system that I have now. But I do think it's smart to think about what it's going to cost internally from an hours perspective, what it's going to cost from a credits perspective and build out perspective and then ongoing. Like if you're utilizing AI within your tool, if you're processing data within your tool, like there's cost to all of that, right? So as you scale, depending on how big your company is, you know, it could really, you know, balloon. So you just have to be conscious of that. What are some of those numbers like financially that we've used?

Chelsea Pagliuca (40:25)Right. Yeah, yeah, with the tool stack, it saves us about 5,000 a year, which isn't, I know, a huge number. ⁓ But I think the time aspect is what is more valuable here based on how much time we save. running the numbers first year that we launch it, it's about $45,000 that we're saving both time and tool stack cost. And I assume... hope that we would probably just add value to that based on features that we add and as we continue to perfect it. And so, I mean, a three-year total value would be over $120,000. So, great review.

Ben (41:12)Yeah. And for a studio like ours, like that's significant, even just subscription costs, you know, we're in that three to 400 range for HubSpot. What is teamwork? Like three to four grand a year? Yeah. Three grand a year. 90 is about, in all these tools, I, you all these people have built great products. But yeah, it's small business. You do what's best for the studio. I think as a, as a business in general, that's where everybody's heads at is like, where can we save? Where does it make sense?

Chelsea Pagliuca (41:17)Yeah. That's round three. Yeah. Mm-hmm.

Ben (41:41)You know, for us, we have a team who can validate this stuff, who can utilize it. So yeah, I think 90 was like 600 bucks a year, which is super great. That's a value, right? But if I can, you know, I can roll that into something already have, and it's tightly integrated with my management, like that value there saves time, right? Rather than having disparate systems, we can now get the rest of the team, you know, within that EOS structure without having to have multiple accounts over, you know, on 90 and on HubSpot and on Teamwork. And on, know, so just, kind of compounds. So, you know, that lifetime value does add up over time. And I think there's value in being able to build bespoke. Now there's negative, right? There's a flip side to that. You are now, you now manage this product. You now control that database when it breaks, you now have to fix it. So some of that will be offset by hours internally. I'm comfortable with that because I have a team that can handle it. So, but. I think a lot of people need to consider that when they're saying, hey, I'm going to buy up code something. Because we do make it sound like this is great, this super fun, we're excited about it. But there are implications. There's time implications. If this thing goes down and it's a critical piece of software within your studio, within your agency, within your company, who's going to fix it if they go on vacation? You have to consider these things because now you can't just holler at support and say, this is broken.

Chelsea Pagliuca (43:06)Yeah.

Ben (43:06)or, you know, and then say, yeah, we got it fixed or this is why, or there's an out, like you now internally have to handle all of that. So that has to be okay, right? Like you have to, there's a certain level where you have to consider if the value you get out of it, the money saved is worth the additional, you know, technical debt, project management and infrastructure that you now have to manage. So I think that's something really important to consider that a lot of people just don't think about because they don't. think about it as an engineer or developer or designer. They think about it as, just ripped this, the rift on this real quick and now it's out there and we're going to rock and roll. And then you get to a point where it's like, goodness, now I have to like kind of work my way back. And you realize why some of these design principles exist, why there are, you know, know, scrum masters and product leads and engineer teams and all these different things. Like you see why that stuff happened. But you come to it from the reverse side. So. Just some good stuff to kind of think about there. And it's, I think it's good just to have some numbers, right? Like 800 bucks in credits so far, four weeks of work, 65 hours of Chelsea's time, our studio bills $150 an hour. So there's costs there, but lifetime value and ability to build exactly what we want and have it all integrated. feel like it's worth it to me. It's worth it. And to me, it's also an experiment and it's also an opportunity to try and learn. build. And I think that's fun. And it's, you know, kind of think open Chelsea's eyes to what, what could be and how we could, you know, maybe help clients. Because this comes down to the more efficient we are as a studio, the more value we can then put into the work itself, into the creative, into kind of differentiating. I think it kind of opens up the door for us to be able to do more for our clients and, you know, a tighter timeframe. And I've never been the one to like, pitch speed, but this does, it kind of takes, it removes some of those barriers to allow us to like create faster and to not have to tinker around.

Chelsea Pagliuca (45:08)Yeah, I think it even adds value like from a client from for prospective clients, because we're not only speeding up our process of how we set up a project, we're also speeding up the sales process because our proposals are going to be made quicker. We're going to have response times and be faster. So really the whole client cycle should be seamless for our clients using the system. And they also are going to be able to see.

Ben (45:20)Yeah. Yeah.

Chelsea Pagliuca (45:33)more of their project within the system. Like it's more transparent the way that we've built it than what teamwork has ever allowed. So I think it's just going to be a better experience for a client too.

Ben (45:42)Yeah, that client portal was created specifically for transparency for the clients, but it's a manual update for us. It's a manual. The value was worth it. Like the lift was worth it to allow that transparency, but this will be one to one, like all pulling from the same data source. So what the client sees is live, you know, a dash of current hours. And, you know, if we want to drill down, we can, we can have different layers to it. So yeah, that's super fun. Where do you see, you know, talking about clients, you know, we talk time investment, like So how do you see what you've learned? Because we're not the only people doing this, right? There's other people out there exploring and experimenting. How do you see clients, you know, applying some of these new tools and where do you see kind of SDH fitting? Because now like I'll pull Chelsea in on a call and be like, Hey, you know, you're vibe coding something. Yeah. Let me get Chelsea in here. Cause the way she's approaching it's probably gonna be similar to how you are. So how do you see client applications on it from a service perspective? Like what we can offer.

Chelsea Pagliuca (46:35)Yeah. Yeah. I mean, I think there now is opportunity to be a consultant regarding these tools to clients. because like you said, people are going to use them and they want to use them. And I think it's a great opportunity for people to use them and potentially have cost savings. So, I mean, we're not knocking anyone who wants to do it themselves and not have us create something from scratch. I think it's, we maybe take the seat of someone who's a teacher or empowering. our clients to do something and try something. And they're kind of like me. I mean, they're going to be doing the same thing as what I'm doing potentially. And so they would have the backing of socials development team and the knowledge base that we have in the design team as well. I think, I mean, I think there's huge implications for a service here that we can provide to clients. And I think it's exciting actually. Like I don't. I know AI can be scary, especially for an agency. I mean, we've talked about that at Nausium this year, like everyone else has talked about, but I do think that there is opportunity there. And I know that for me, I probably see that more clearly because of my current job and how it's different than a developer would see it. But I actually just saw a meme this morning of two guys on a bus and one, like they both have little thought bubbles above their heads and... They both said AI is taking my job, but the one like was crying. And then the other one was looking out the window excited because he saw opportunity there. Right. And so if you just have a different mindset of it, you know, like think about how your role will change or opportunity to change and do better, be better, learn more. think, I mean, I think there's great opportunity looking at from the sunny side.

Ben (48:01)Yeah, I've seen those, Yeah, looking at opportunity and I've talked to you all internally about that idea of this Jevons paradox. I'm not a big paradox guy. I'm not super heady, like intelligent person. You know, I'm not like, I don't get like real deep on this stuff, but my buddy Dave, he brought this concept up to me and I thought it was really intriguing and it gave me a little solace as I thought about who we are as an agency, what we can provide, where we kind of plug in and provide value. And that's going to naturally just be different, right? It's just going to... It's not going to be the same as it was the last 16 years. If we're doing everything we do the same way next year as we did it in the years prior, we're doing something wrong because there's so many more opportunities. There's different technology. It'd be like iPhone comes out, and I remember that. I'm showing my age, but I remember when that iPhone, it changed everything and it created so much more opportunity because now I needed to have a responsive site. Now I needed to have a website that worked on mobile. you know, it kind of created, you would think it would create less work, but it creates more. So this idea of Jevons paradox, like if you look at it from an automotive perspective, you know, the original car was very inefficient. It probably, you know, five, six miles per gallon, that could be incorrect, but you know, it wasn't efficient, right? It ran, got you from point A to point B, but it wasn't efficient. Over time, these cars get more efficient. We can now run hybrids and they go 60 miles per gallon. So naturally you would say, Hey, know, the use of fuel is gonna go down because we have created a massive efficiency in the market. So, hey, you know, now that I have a car, I'm not using as much gas, gas gets cheaper, but the inverse is actually true. Usage of gas and fuel went up as cars got more efficient. Like, why is that? It's because people now use it to drive farther distances. They use it more on the regular because it's cheaper to use. And so, know, Jevons paradox is this idea that you create efficiency, which you would think would reduce usage, but it actually increases. So if you put that in terms of AI, it's the same thing, right? We've created all these efficiencies within these little pieces of the puzzle, but because we've created those efficiencies, people are going to use technology more because it creates even more efficiency. So this, this idea that it's going to take all of our jobs, everything's to be gone. That's true in some instances, but I think the usage of it all is going to, you Escalate it's not just gonna it's gonna not gonna take you know these technology jobs down if anything It's gonna escalate them because everything is now layered on with a layer of technology So that concept that that whole Jevons paradox was an interesting read for me is an interesting conversation to have it gave me a little a little calm in the fact that hey just because like we're creating all these efficiencies and removing these kind of deadweight tasks does not mean that nobody's needed if anything it might be might be the inverse I could be totally wrong but

Chelsea Pagliuca (51:02)Yeah. Yeah.

Ben (51:03)It was a fun thought. It was a fun thought to talk through. So what's left on Swirl? That's what we're calling it internally, Swirl. It's got our four different pieces. What's left? What's our launch date? I love to do, we'll do ⁓ maybe a two month recap or something after this where we talk about how it's going. think it's like sunshine. We're pumping sunshine now. Everything's great. Once we get this in, working within the studio, maybe we do a little bit of a recap of what worked, what didn't, where we saw issues. So what's, what's left on the list and what's kind of the rollout plan.

Chelsea Pagliuca (51:34)Yeah, so I just met with our dev team this week to try to get them spun up on it so they can get in there and do their magic. And so I know their first steps is simply to sign in lovable and kind of see what's going on under the hood. So they're doing that within this next week. And then hopefully based on that, we'll have a little bit more clarity of if they have a lot of tasks to do or if there's not a big lift. But then once they start looking in, we'll pull Ben and Adam in to start doing some testing and then pull in the broader team as well. Hoping honestly, by the end of June to launch that of course, client work takes priority in our studio. So some of that is a little bit dependent on client workload and things like that. it is, mean, like Ben said, I think he said 75%. I think that's really accurate. Like it's close to being there, just needs some testing, some finesse. Obviously there's...

Ben (52:11)Yeah.

Chelsea Pagliuca (52:27)a little more pressure with the studio aspect of this because clients will have a view within Studio. So we want to make sure that is seamless and makes sense so they can log in and things like that, that it's working as expected. So yeah.

Ben (52:39)Yeah, 100%. Well, that's great. Yeah, we'll do a recap of kind of where we're at. You know, we did this just to share process, to share how we're using these tools, how we're experimenting, exploring, you know, what I've seen so far. I have no doubt that I think this will be either A, a valuable experiment, or B, just a valuable tool in general that was worth the investment in hours, time, and understanding. Because without trying, without being curious, You can never have confidence in saying, hey, oh, I've done that. Oh, I know it can do that. Or yeah, you can do that, but that's not going to do it very well. So I think being curious, tinkering. I love that word right now. Be a tinker. You got to tinker with everything. Try a tool, sign up for something. I think that's huge. That hackathon mentality is what I've been spreading, is that you just got to play. Right now is the time to play. We'll check back in Chelsea. I appreciate you chatting with me about kind of this tool we're building. Cause it's been fun. It's been fun to see you get excited and engaged and build something and be like, is all new. This is, know, like this isn't the norm or hasn't been the norm, but it's now the norm. And so how do we respond to that as an agency and as a studio and help provide ultimately it's how do we do better work for our clients? Provide more, give more value for, you know, be good stewards of billable hours and time. So it's exciting to build a tool that can help with that. So any final thoughts?

Chelsea Pagliuca (54:05)Man, try something out this week. That's my final thought. Just do it even if you're scared. You won't break anything. That's the, I think that's the final thought. If you break it, you can fix it. Yeah. Yeah. Yeah.

Ben (54:15)Yeah, yeah, yeah, there's something comforting there and uncomfortable. You know, there's two sides to it. That guy in the bus looking out the window. Yeah, same thing. So, all right. Thank you, Chelsea. See you everybody.

Chelsea Pagliuca (54:24)Yeah, yeah, yeah. Thank you. Bye.

‍