Subscribe
Share
Share
Embed
AI First with Adam and Andy: Inspiring Business Leaders to Make AI First Moves is a dynamic podcast focused on the unprecedented potential of AI and how business leaders can harness it to transform their companies. Each episode dives into real-world examples of AI deployments, the "holy shit" moments where AI changes everything, and the steps leaders need to take to stay ahead. It’s bold, actionable, and emphasizes the exponential acceleration of AI, inspiring CEOs to make AI-first moves before they fall behind.
Forum3 (00:00)
This is AI First with Adam and Andy, the show that takes you straight to the front lines of AI innovation in business. I'm Andy Sack and alongside my co-host, Adam Brotman, each episode we bring you candid conversations with business leaders transitioning their businesses with AI. No fluff, just real talk, actionable use cases and insights for you.
Today's episode is a little different. We're planning on sharing an inside look of one of our AI First community calls, where leaders from across industries come together to compare notes on what it really takes to adopt AI in their organizations. In this session, we sit down with Keith Fairclough, the CIO of CABI, a women's apparel brand that sells exclusively through a national network of more than 2,000 stylists.
Key story stands out because the industry isn't traditionally known for tech innovation. And yet, Cabe has managed to build one of the most authentic AI-first transformations that we at Form 3 have seen. He walks through how they built AI literacy across the company, how they aligned executives and employees, and why focusing on culture over hype made the biggest difference in their transformation.
Then Keith shows what the AI foundation enabled. One intern building a fully functioning AI styling and virtual try-on tool in just a few weeks. A no-code training simulator created in a matter of hours. These are two real examples of what happens when an organization empowers its people to use AI every day. Let's get into it.
Adam Brotman (01:54)
Keith has graciously agreed to join. It was really kind of a we're we're working in the last week or so on on the agenda for today and completely organically. Keith was catching us up to what he had been working on and we have been working on some stuff together. And so I'm going to we're going to I'm going to kind of explain to you. I hope you'll find it interesting.
some of the things Keith has been doing, but how it sort of came about from his company, Cabi's journey of becoming an AI first company and Keith being kind of the AI leader there. And then some of these things that have come up, we thought, this would be really cool to show the community and kind of show, don't tell in terms of what it really, what we think what it looks like to be a great AI first company. What that.
sort of looks like and it's not as sort of shiny object like perfect McKinsey, you know, figure out your use cases and then write your roadmap like all that's good. But we've been we're believers in that. But I actually think you're going to see today what feels like a more authentic organic example of what it looks like, in our opinion, to be an AI first company.
Keith, thanks for joining us. tell us a little bit about Cavi, the company and yourself. And then I'm gonna kind of interview you kind of fireside chat style.
Keith Fairclough (03:15)
Sounds good. Nice to see everyone and meet everyone. My name is Keith Fairclough. I am the CIO here at CABI. For those that don't know CABI, and I didn't know it before I joined, but CABI is a women's apparel brand that's been around for 24 years. We are in the direct sales space, so an MLM. And our industry is not known necessarily for being the most innovative on the technology front. My background
was a management consultant at Accenture for the first half of my career. And then really a tech operations, a tech strategy after that before joining CABI. we primarily sell our product through our stylists. So we have 2,000 stylists who still do what we call shows. So think in person.
They have mini pop-ups. They don't hold inventory, but they do have their own clients. They have a sample line and we ship all of our clothing directly to our customers from our distribution centers. So with our PE firm, we have a mandate to modernize. We first had to get a number of things in place in terms of our data layer and our business intelligence. now working with ⁓
with the Forum 3 team really trying to figure out how can we leapfrog some of the competition and just leapfrog potentially things in our industry by leveraging AI.
Adam Brotman (04:31)
So when you first approached us, Keith, it was even before you had started in earnest your journey to re-platform your commerce engine for all your salespeople, which are your stylists, you were just in the sort of early innings of like, let's modernize our tech stack, let's modernize our approach.
Talk a little bit about like how that journey went. Don't worry about the form three part, but just on the cabbie side, like what was their like executive sponsorship? What was the role of the CEO? Walk us through sort of where there were pitfalls and challenges, but also where what surprised you on the good side.
Keith Fairclough (05:09)
Yeah, think we had some, I imagine like any company, had early adopters as soon as a chat GPT came out mostly on the marketing side. throughout 2023, it was everyone had their own personal accounts and I was pretty confident people were uploading confidential information and using it to their advantage for their role. And then, but at the same time, I knew that the experiment
I wasn't sure how I was going to implement it. Was it to build the, start the cool project and build something or do we focus on making the company a bit more AI literate? So we took the latter past and really focused on upscaling our entire organization. And if we could upscale the entire organization, the hope was that the ideas would start to surface from our team members versus from my seat.
I knew I needed executive buy-in. Some executives were using ChatGPT at the time and Claude. Others had really not touched it at all. So we got our executives into a bootcamp. So we were all speaking the same language. And it was really from that bootcamp that we got the buy-in. I think our CEO almost immediately was like, we have to figure out a way to do this. And it was at that point we decided to do in-person training for everyone.
I can tell you just a short story. When we did training, would say there's probably like 25 % of the people were still at their desks, just asking if they could zoom in because they didn't want to attend. They felt like they had too much to do and we forced them to get in the room and spend the day with the team. And by the end you had every hand up immediately saying they know how they can apply this to their role. And that's all we wanted. We weren't asking for projects. We just wanted them to figure out how they could take this new superpower and apply it to themselves.
And then from there, we could start to surface the enterprise implications.
Adam Brotman (06:54)
And where did you see like, so that was what a little over a year, like a year and a quarter ago that you sort of really kicked that off and got everybody on an enterprise version of chat, TVT and got everyone sort of trained and started to create sort of the kind of the governance in the, in the routine of everyone using it. What, how, in the last year, how would you say it's gone? Like what,
from your perspective, like what's gone well, what do you wish you could have done differently?
Keith Fairclough (07:24)
On the what's gone well, so we did a survey at the beginning. We wanted to measure sort of our awareness and adoption and then we did a survey about nine months later. But I think awareness and adoption, I wouldn't say we plateaued, but I think we're at a point where the majority of the organization leverages AI in some form for their roles, some more than others. Our early adopters, where again, we're marketing, our design team, the technology team.
We kind of lost the finance and business intelligence team at the beginning because none of the tools were really that good at data analysis at that point. And they tried to use it and they weren't getting the results they wanted, so they kind of backed off. That said, we're on the flip, the other side of that now where we're about to do a pilot with Snowflake and their Cortex AI product. And we're looking at some other proof of concept products to sit on top of our Snowflake instance.
On the things that went well, we did weekly open office hours. So you could come by and talk about AI for anything that you were doing related to your role. We did monthly GPT competitions. And it sometimes wasn't even GPT, it was just how you were leveraging AI in your role. And you won, I think we did $250 gift cards and the teams voted. It's actually a pretty cool team bonding experience. I think on the side where there were some gaps,
We didn't implement any type of evangelist program or ambassador program. It was really myself and a few other people really leading the charge. And the team jumped on as well, but it would have been nice probably to have a couple of people dedicated within each org focused on what are the things that they want to do and surface it to the top. It's taken us a little bit longer to get there. We're just starting to get there now.
Adam Brotman (09:04)
Did you actually, yeah, remind me, Keith, did you put a council or a task force together or talk a little bit about did that not work as an evangelist group or what happened there?
Keith Fairclough (09:15)
It was really small. It was myself, the COO, and one other person as the council. We did put together an AI use policy. We wrote it in a way where we wanted people to be smart about how they leverage data at the time, but we also wanted them to experiment. But our council was just three executives and continues to be three executives.
Adam Brotman (09:34)
Yeah, so you're saying in hindsight and going forward, that's something that like having a little bit more consistency and structure around evangelism and keeping people on it and not letting it after fee, that's something that you would have done differently. Yeah.
Keith Fairclough (09:49)
Yes, 100%.
Andy Sack (09:51)
Hey, Adam, I let me ask? If can you talk about like numbers in your own experience, how is your usage gone up and also organizationally? What have you seen in terms of actual usage, either per employee per month per week, however you measure it?
Keith Fairclough (10:08)
Yeah, we don't have all the stats off the top of my head, but we do. I can for myself. I use it. I mean, I use it like I use email or search the Internet. It's it's just part of something I use every day. So it is and it's true.
Andy Sack (10:25)
Was that true a year ago?
Keith Fairclough (10:27)
No, would say even a year, a year and a half ago is I would use it a handful of times a day for very specific things. Now it just becomes the default. If I have a question or want to brainstorm an idea or before I'm heading into a meeting, one thing I do all the time is as I'm heading into the office, I have a meeting first thing in the morning or first couple hours and I don't have time to prep. I'll just prep with it in the car. So I'll have a conversation. We'll talk about the topics. By the time I get to the office, everything's set up for me for my talking points for.
for meetings or discussions. From a team perspective, some of the things I do remember from our survey, we had about 40%, 50 % of our users leveraging some AI when we started to now, or about nine months later, which is about four or five months ago, we had like 96 % of our team leveraging it. Those that were using it three times per day are now using it six plus times per day.
the team members who've felt uncomfortable or just didn't feel like they understood AI significantly dropped within our organization. The teams that are talking about AI is nearly at 100 % on how they can leverage it within their organization. And that's what we wanted. I know it's not sexy. We've just wanted people figuring out a way to leverage it. And that was our goal on the awareness and adoption side.
Andy Sack (11:48)
And was there, was there, ⁓ can you talk a little bit about employee pushback or fears or concerns that got raised over time and how you dealt with them?
Adam Brotman (11:51)
Either.
Keith Fairclough (11:59)
Yeah, think there was a there's definitely I mean, and I still have the concern is always the concern about uploading our data in there. We have the discussion all the time about our data sets everywhere, right? It's it's in snowflake. It's it's in Salesforce. It's it's in other places. But there are some team members that were always concerned about what is the impact to my role. I don't think that goes away. Our messaging to our
team is that we look at it as like a superpower in your pocket to improve your own personal productivity. I truly believe that that's something that we focus on. It's the same thing we've been training our stylists to on how to leverage that GPT for the last year. We do monthly sessions to teach them the basics. And we use the phrase AI is here to amplify you. And I would say, Andy, that's probably the biggest one is like, what's the impact to my role?
But I do think that most of the team members, most, I can't speak for all, most believe that the investment that we're making in them, I really believe it's a skill that they're gonna have, whether they leverage it here or they use it to go somewhere else and have a successful career, I just wanna make sure that they have that skill on how to leverage it.
Adam Brotman (13:04)
So, okay, that's awesome. So that's a great kind of grounding for everybody of, know, Keith and Kavi and their AI journey. Recently, there's some things that came up with Keith that we wanted to share. know, Andy and Rose and I were like hearing that Keith, this was organic, this was not planned. Like he was doing a couple of things organically and we wanted to share those with you. And one of the reasons we wanted to share them with you is
that we're big believers, that the best way to come up with innovation, and we believe using AI is not to like start with innovation in mind or cost savings in mind or some specific shiny object application in mind. It's more of get the organization using AI, understanding AI with the right governance and the right mindset.
growth mindset and just leaning into AI but responsibly and then let every functional leader and every function almost organically come up with their own solutions and innovations because they are the ones closest to their work like Keith was saying. the example we wanted to, by the way, side note, Keith told me that like six, seven months ago, he got in front of all 2,500 of their Salesforce, their stylists, and he ran like a mini
chat to BT bootcamp for them saying, this is a free tool that you can use every day to just amplify yourself as you said, Keith. And I didn't tell you to do that. didn't suggest you do that. You just took the leadership of a AI first leader to just do that. And you did it in a way where you had brought the executive team and the headquarters along enough that by the time you got on stage in front of thousands of stylists and said, Hey, I'm going to show you about this new tool. And I want you to bring it to
to the table as appropriate. It wasn't like your head office was horrified. By that point, everyone's like, yeah, this is what we do. This is how we roll as an AI-first organization. You made the decision to do that. And I think that's very, very smart because it's a win-win for the company and the stylist. Now, you and I were talking about a couple of things that I wanted you to do some show and tell for everybody. One is, and you can pick whichever order you want to do this in, Keith. They're both sort of cool. I can't decide which one we want to do first.
you do it organically here for everybody, which is that you had two projects in mind in the last couple months.
So can you show everyone in whatever order you want the two AI things that you and your team have been building and they're ready to go into production?
Keith Fairclough (15:35)
I'm just going to show two things that the team's working on here. Like I mentioned earlier, we have a team that's working on a natural language bot with Snowflake in their Cortex product to figure out how our teams can have natural language conversations with our data. We have another team that leverages it for a photo retouching to reduce our costs there. But this one is a tool. Our style is basically provide
styling advice or free.
Our stylist have a tool today right now where we help put together outfits. Our team gets together in the office. We manually put the outfits together. It's an Excel program. Originally, many years ago, was meant to be some... ⁓
machine learning algorithm that never came to fruition. And we upload these outfits and our stylists and our clients access these outfits to put stuff together for the current season. We wanted to do something different and we knew the technology would allow us to do that now. So we met with our merchandising team and they basically said, Keith, if you can create something that just at minimum gives us a, you know, the style boards you see on Instagram with like the white background and then the
flat lays of clothing on them, like a top, bottom, pair of sneakers, a necklace, a hat, all that kind of stuff. That would be great. And if you could take it to the next level and make it shoppable, that would be even cooler. So we brought in an AI and machine learning intern just through someone I knew in my network. He had just graduated and got his master's. And within three weeks, we started a
put some ideas on paper and he started going through. And before I do the demo, can tell you what you're seeing now is not what it looked like on day one. When the results, it looked like we were putting outfits together in pitch black darkness, nothing matched. And we just kept working through it and leveraging our content. is built leveraging Snowflake for our client data. It is built leveraging our Shopify product catalog. We use Hite Dream from a middleware perspective. We use Gemini, we've leveraged OpenAI and then we have
Also, we found an open source AI tool that has been trained to help with fitting clothes on people.
Adam Brotman (17:43)
So
just to be clear, this is one recent graduate, not necessarily like a team of developers. And what we're seeing here is the one person we're collaborating with you and some other people in the organization in three to six weeks.
Keith Fairclough (17:59)
The first draft, we got it working in about three weeks. We're probably on weeks seven now of it. So there isn't even a more newer version of this, but I'll go through this. So I can type in here. I need an outfit for my daughter's soccer game. It's Saturday. It's meant to be...
Cool in the morning, warm in the afternoon. Keep it casual and comfy. This is very high level one here, but I'm going to do generate looks. It takes about 10 seconds to create an outfit. We have it right now creating three outfits at a time.
Speaker 4 (18:27)
casual
Keith Fairclough (18:41)
And what it's doing at the most simple use case is it's always going to pull from our existing product. And when we looked at tools, there are a lot of AI styling tools out there. Don't get me wrong. There are a ton. Many of them focus on how do you put your product on models for photo shoots to reduce your overall photo shoot costs and put it on your PDPs or e-com site. Then you have the other end, which is the Google product and a few others where it's designed for the individual consumer. How do I
put an outfit on myself, but it's always different outfits, different brands. We wanted something that was just our brand that our stylists could access. And right now it's giving a high level outfit, but I can then personalize this based on who my client is. So it puts together these three outfits. I could ask it to do more if we wanted to.
Up here, it gives you a little bit of a styling strategy. It'll likely talk about my daughter's soccer game. Yep, my daughter's soccer game. Transferring seamlessly from cool mornings to warm afternoons. Down here, it'll give a little bit more. It's pulling our product names. So you'll the shawl collar cardigan provides cozy warmth. And it talks about why it put this outfit together. We are working with our merchandising team to continue to simplify the output.
Here you can, I'm on my computer, but if we were on a mobile device, you could share it through text. We have built the backend here, so when you share this, it's shoppable. So the client can click on it and it will be shoppable. And then that order would be attributed to the stylist who shared it. So she gets her commissions. You can download it, you can favor it. Our stylist, one of the services they offer is they,
purchase on behalf of their clients. So imagine you're going on vacation and you don't want to deal with shopping. You're saying, I need three outfits. Our client can put that whole order through for you, check out on your behalf and ship it off to you. So if they like an outfit here, they can click here. This is a rough version. It will connect to our Shopify checkout, but I could start adding the cart together and check out on Shopify on behalf of my clients.
Adam Brotman (20:48)
There's a question here. Is it creating the image based on the two or three items or do you have those image combinations already created in advance?
Keith Fairclough (20:56)
No, it's making it unsown.
Adam Brotman (20:59)
Got it. it's taking individual images that are associated with the with its result that it came up with. And then it's creating the combo image.
Keith Fairclough (21:08)
Correct. And we put some rules in there of how we recommend our products going together. We have a ton of metadata in there for the different products. We gave it some pairings that we thought worked to give it an idea of what outfits would go together. And there's a bit more complexity to it, but generally that's how we started to work through it. But it is pulling all of our product data from, and client data from Snowflake and from ⁓ Shopify.
Adam Brotman (21:32)
And if the stylist or the client wants to see how the outfit would look on them, did you incorporate that into here as well?
Keith Fairclough (21:38)
We did. So you can take any one of these. Let's take this one here. Let's do this one. So we were inspired by the Nano Banana craze that was going around for a bit. And I'm sure it's still there. And we were like, how do we do that? Because they didn't have. I know they leverage. It's a Google product. So they leveraged their own tools. So we figured out.
We wanted to try it. We did try Gemini as the backend. We leveraged their APIs, but it just wasn't fitting right. So our AI intern went out and just tried to find if anyone had trained a model on fit. And we realized it's not going to be perfect, but You can, let's say I'll start with the top. And I'm going to apply the selected item.
Originally, we were thinking the merchandising team wouldn't like it. But after they saw it, they're like, even if it's not perfect, the clothing is perfect. It matches down to the button the way we've got it working. And they said, as long as it looks close and it looks like what it would look like on what a client will give the client an idea of what's possible, and it's our product, right? It's not making up a gray sweater here. That's actually our product. Then I could put the jeans on.
⁓ I can apply the selected item.
Three weeks to get the working version. This version is like six, seven weeks.
Andy Sack (22:53)
One person, one person. Fantastic. Yeah.
Adam Brotman (22:57)
And then, I'm not going to say I don't keep you didn't tell me how technical this person is. like the point is that like brand new graduate intern, you know, six weeks later, you've got this and you can imagine all the legs this has no pun intended for like the, know, incorporating it into their entire workflow because their workflow is salespeople. They don't have retail stores. They have these salespeople that have the clientele and communicate.
either in shows or after the show, messaging. And so this is just an amazing tool,
Andy Sack (23:30)
Was there a product doc that was created that was before this got built or did you just go straight to implementation?
Keith Fairclough (23:39)
kind of I knew what I had I knew what I wanted to build. As a matter of fact, when I brought it to Adam and Rose, they were like, Keith, you need to put together a brief. So I went back and created one, but we never really we just we literally just sat in my office and on a whiteboard said this is what we want to do. And then we just kept iterating from there. Nice part about this is it puts it on this background versus that background that joy had her picture was in an office. We wanted everyone to look like they were, you know, had a white background. So it looks like a
a PDP shot.
My AI machine learning intern actually did. He developed this. So he didn't use any of the cursor tools or lovable or anything. What I'm going to show next is something that I did on lovable. But he just built it from the APIs that were available and moved forward. So I'm sure he could have done it, but he's a pretty skilled developer. So he just went for it.
Adam Brotman (24:34)
Hey Keith, before we move on, I know you guys also were able to tie in like their existing, their existing kind of closets, right?
Keith Fairclough (24:42)
Yeah, that's a good deal.
Adam Brotman (24:43)
really interesting feature.
Keith Fairclough (24:45)
Um, our, if they want, if you wanted to do an outfit with it's all fall 25 product right now, cause that's what we want to sell. You could pick any product and then have it style outfits around that product or pick two products and have it style outfits around those two products for fall 25. But one of the things that our styles community, our clients love, we call it the closet. So it's different than your favorites. It's different than your order history. It's actually a, a, a slice of your order history. It's everything you've ever purchased.
And it's a flat lay in your closet, what size you purchased, when you purchased it. And we wanted to figure out a way, we take things from a closet of what someone's purchased and mix and match it with Fall 25 product to tell them, hey, you can create this outfit for things in your closet.
So it'll pull up her closet. This is everything she's ever purchased, and she has a ton of stuff. But let me find something that has some color so we can see how it works. And then I'll click Select, and I'll say, you create some outfits?
with this product for brunch this Saturday and then generate looks. It should pull that item as a primary item and then put some outfits around it. So our goal is to be able to give this to our style so that they can pull these things together. We are thinking about ways where we could
A client could say, want to spend $500. How do they maximize the number of outfits based on new and old? Lots of different things, but here it is. it's put some pretty cool outfits together with that product, even though it's, I think, orange.
Adam Brotman (26:28)
Yeah. What's been the reaction of because I know this isn't it's not in production yet or it's not live yet. You haven't launched it. But what's been the reaction with some of your stylists that you've shown it to?
Keith Fairclough (26:38)
They love it. They absolutely love it. They've asked if they could just get it in the current form, just to give them ideas on what they could pitch to their clients. I do think it's helped that we started a year ago and getting them comfortable with AI. I mean, we were in a community that are not tech savvy at all, and just getting them to learn how to create their ChatGPT accounts and now having a good portion of our community who use ChatGPT all the time.
for their business strategy, for their show strategy, how they're going to approach the season, that kind of stuff.
Adam Brotman (27:10)
how frustrating was the iteration?
Keith Fairclough (27:12)
The iteration was frustrating at the beginning, not from a team member perspective, that our intern was doing great. We just couldn't figure out how to get the outfits to match. And we were trying everything. took transcripts from mainstage presentations that our designers have done. We took our training.
And then once we started really playing around with the metadata and sort of our styling rules and what should go together, and it started to come together. And that was probably the most frustrating since then. It's, honest, like I don't mean to exaggerate, but it's been nothing but fun. It's like, what's the next cool thing we can add onto this to be beneficial to us and ultimately use it as a tool to drive sales.
Adam Brotman (27:55)
Yeah, and Jordache asks, like, you know, since you're probably not going to charge for this tool, per se, is it, you worried about expense of API calls or hosting costs?
Keith Fairclough (28:08)
Not yet. We'll probably see what that looks like once we get it out there. you know, we are I imagine not every styles is going to use it every day all the time. So we will ⁓ it's in their back office. So it's it's gated. They have to log in to be able to get there and use it. We're not going to expose it to clients just yet. We'll have to figure out an efficient way to do that.
Adam Brotman (28:28)
Yeah, right. And Mark brings up a great point about out of way to talk to it versus typing in. My guess is that whatever implementation you do, though, they'll at least be the transcription microphone option in the interface layer. But maybe not. Maybe if you best stop that, you have to sort of build that in.
Keith Fairclough (28:44)
Yeah, we definitely could do that. We've been talking about voice activated for this. We're also ⁓ about to start a tool where our stylists put together invites for their shows. That's like an e-vite or paperless post or Apple invites. Instead of putting the invite together, we want to give them the way to just type in their invite and say, I want to do an invitation for this Friday at Carolyn's house. I want to invite everyone that Carolyn had to her show last year. Can you please put together the invitation for me?
and do the whole process and maybe not send it, but get it set up in case she wants to it.
Adam Brotman (29:15)
Yeah, no, this is great. Now let's let's transition real quickly switching gears we were you you have underneath this part of what's underneath this is your new commerce engine is powered by Shopify with API's and
And you were, you've got to get, you've got to train a whole Salesforce on how to use this brand new interface of a commerce engine, which is like the main tool they use to do, to run their business. You guys were going to do a bunch of like videos that you had created. What are those called? Like loom videos or whatever. And, and like some PDFs and, you and I were brainstorming. I was like, you should just like use lovable and like vibe code training.
video site or something like that, some kind of training site. And you were like, I hadn't used lovable yet. Adam, can you show me how you use lovable? And you got on you and I screen shared for like 15 minutes, I showed you some stuff that I was doing with lovable. You went off and in like, what three hours worth of your time you you you kind of advanced past what I was even thinking for lovable. Can you show everyone what you did?
Keith Fairclough (30:17)
Sure. And that is a good segue because it was about three to four hours, I would say, of just learning the tool and figuring out how to do it. I'll show you two screens here.
Adam Brotman (30:29)
Before you get into that, Adam or Keith, could one of you guys just describe what lovable is for people who might not be familiar with it?
yeah, yourlovable is a platform that uses generative AI and natural language to create code and create applications. So you don't need, the AI is actually doing the coding and the designing for you. You don't need to be a developer like this intern Keith had. You can be someone like me. I don't know how to write any code at all. And yet I can create apps.
just by speaking natural language, like as if I was prompting chat GBT, but doing it on lovable, lovable, especially made to help non-technical folks, for example, just naturally talk to the AI and the AI will build you an app as if it was like a development and design team both. And then it'll host that app for you if you want to let other people use the app that you create. It's amazing for clickable prototypes, but
Keith Fairclough (31:11)
time.
Adam Brotman (31:26)
more and more people are using it as beyond just prototypes, but actually for production grade applications that they can get out in the world. And Keith is going to show an enterprise version of an example of that.
Keith Fairclough (31:38)
And what we were trying to do, we've been working with Shopify for about a year now, is they try to get into the direct sales space around how we can leverage their technology within our industry. And they've done a lot of unique things and opened up things for us to be able to test. One is the point of sale so that we can deploy it to all 2,000 stylists. And the only thing is there's some unique things that we have to do on the POS in order to set up a...
a sale appropriately so that the stylist gets the right attribution when she makes the sale. That's the complexity of doing it versus a typical retailer. And we can't deploy this until we go live. So we needed a way to train them. And a loom video is great. Live training is great, but we wanted something that could click around. And I didn't want to do corporate LMS. I didn't know what to do. And it just happened to, when I connected with Adam, I figured I'd give this a shot.
So this is an actual screenshot of the point of sale. And I uploaded the screenshot and I just started chatting with it. I told it what I want to build. I wanted to build a fun, engaging training tool that goes step by step on each of the steps that a stylist will need to do to set up a sale in the point of sale. And it didn't do a great job just by me giving it a screenshot. I told it to highlight this EOS tile here.
And then it would highlight down here.
I had to actually get away from the screenshot and I took the screenshot and I brought it over to another chat. And I said, recreate this as a clickable prototype because it didn't like the screenshot. And once I got it into that format, it was almost smooth sailing from there. And when I say it's all natural language, it was all natural language. It was just me chatting with it and constantly troubleshooting. But I got it to this point where I can now, I'm to go to the shortcut, but I want to put together a retail order.
and up here on the left, has a training progress. What to do, first click on the Kavi Studio POS tile on the main screen. It gives you little bit of a hint. I told it I wanted a pulsing thing so it knew where people needed to go.
Adam Brotman (33:39)
I gotta interrupt you and then look at the confetti there. Like, like I gotta just say, you're not coding, you're talking the lovable in natural, you're typing in natural language lovable. And you're saying, I want you to make this application that will show the exact screenshot of my thing. I want you to have it pulse. I want you to have confetti. I want it to do the next, and you're just telling it and it's designing and coding this.
Keith Fairclough (34:03)
Like last night, I found a couple of errors as I was getting it ready for today. And I just told it here on step five, there should be no pulsing. I mean, as simple as that. And then it goes off and does a mini build. And I know I'm doing a simple use case because it's just a prototype, but yeah, it just takes you through. Confetti, I had an animated character there the first time. You pick your customers. now that it's working, I can almost do anything.
They pick their client. go back here.
Adam Brotman (34:33)
I'll out practically, Keith, to point out you were in a situation, I love you, you were smarter than me about this, which is not a surprise, but you, I was like, ⁓ why don't you create, if you've got these Loom videos, why don't you embed them in a site that's got like a trackable gamified thing for all your stylists? And you were like, heck, if I can just wave a magic wand and vibe code a site, why don't I just take all my screenshots and then create a wizard?
that's going to like and because you couldn't give this you can't give access right now to your thousands of stylists to the actual new site because it's not live yet and it's not ready but you could totally demo this in this way for them. So I thought it was so smart that you took the screenshots because like one of these lovable is really good at is taking like you can give it example screenshots and say make me a site like this or I like this page like this like how would you redesign it. You just took
the actual screenshots are created a wizard that was super smart of you.
Keith Fairclough (35:32)
So it's been, this is my first go. I I created one other one the other night. And now that I know what it takes, I've contracted with someone that I know who is a UX designer. He does a couple of other things and I've just asked him to create 10 of these for different training scenarios. So hopefully I'll have them within the next week.
Speaker 6 (35:55)
Yeah.
Adam Brotman (35:56)
Amazing. thank you for showing both of things, Keith. Andy, just really quickly, what's your like 60 second sort of summary of like things that struck you that we want to like impart on the community today?
Andy Sack (36:09)
Before I answer that question, one last question for Keith. did you talk about the discussions at CABI related to your investment in AI and ROI?
Keith Fairclough (36:22)
Yeah, we, I think it's more been on the investment side and less on the business case at the moment. We know from, we do also have support from our board. So one of the areas that our board has really wanted us to look at is what we use the phrase democratize BI. We don't know what the benefit is going to be, but we have tried for, since I've joined and since we ⁓ implemented Snowflake, have Pablo on the top end.
We tried to get the team to leverage data more outside of those that know how to leverage data and that we just not had success there. But we know we want to be a AI first or even a data first company. And we just haven't had success there. And then with the tools that are out there right now, we know that we will get a benefit from everyone have the ability to walk into a staff meeting on Friday, having looked at the last week's reports.
I'd be engaged with it for their role as a marketing analyst or a designer and what the reports mean to their business and what they could do to help drive change. we haven't put a number to that, but we want to figure out a way to make that happen. we are moving forward with that proof of concept to prove it out. And I think at that point we would define ROI. I think for right now we're still at
Andy Sack (37:34)
Is
there an intuitive sense amongst the management team that this is worth it?
Keith Fairclough (37:40)
Yeah, 100%. Yeah, I think we talk about it all the time. For sure.
Andy Sack (37:44)
I mean, my reaction is like six, six, three weeks, six weeks, one, one person in intern, master's intern. And that's that style of state. It's just, mean, having built software in my entire career, it was like those numbers are just staggering and like the demo of it was awesome. that's what stood out to me. Yeah. It's just like, I mean, when I think about it's a moment to step back and go, Oh,
There's all this talk about the AI bubble and overhype. I see that and I'm like, my God, this is underhyped. It takes a demonstration like that to a community where I'm in this space and I'm still blown away.
Adam Brotman (38:32)
were you a Microsoft shop? And is for those that are Microsoft shops, does being restricted to co pilot? Would that be like a barrier to adoption? what are your Are you guys Are you guys a Microsoft shop?
Keith Fairclough (38:44)
We are, but very few use Copilot. We use ChatGPT. So we have an enterprise version of that. I have Copilot. ⁓
Andy Sack (38:53)
We're
going to clip that and send that into Microsoft.
Adam Brotman (38:56)
Yeah. ⁓
Keith Fairclough (38:56)
But
we use ChatGPT as our primary AI tool. That said, there are people that leverage Claude. Some of our developers have Cursor. We have Gemini. People use it for different things, but ChatGPT is our core account.
Adam Brotman (39:13)
Yeah, Paul, that's it. was a great question and it's actually something comes up a lot like we in our boot camps with our clients. We're not we're we're very pro Microsoft and and we're actually believers that copilot is going to become net positive. You know in the long run, but right now like we tell our clients hey, you know it's fine that you've got copilot. It's fine that your Microsoft shop. We highly recommend that you pick a different sort of major tool as your baseline tool. In fact.
We recommend that people use multiple different tools in a responsible way through a task force, a council, a use policy, but through education and empowerment of an organization, we are big believers that responsible use of AI can happen even if you're using ChachiBTE Enterprise or Gemini Enterprise or even some of these other tools. And if you're just restricted to co-pilot, you're going to end up...
really wondering what the big deal about AI is, to be honest with you, as opposed to like, if you really empower your teams to, you you can still use Copilot, but have a different baseline tool and not just VT, like notebook LM is incredible. know, Claude code cursor with Claude underneath it is great. Claude itself is amazing. I guess they just released a new Excel extension, you know, all these tools. I mean, even, even,
Grock heavy.
Transformation is hard. transforming into an AI-first company, is it going to be a competitive necessity over the next two years? It may not be there now. Now it's an advantage. It's going to become like a disadvantage if you're not doing it. And it doesn't happen in a month. You got to get an organization that's all sort of thinking and rowing in the right direction, being curious, having the right experimentation, having the right mindset. And so I love that.
This stuff came out of that. I think Keith gave an example of the snowflake example. He gave an example of talking to all of the folks in the organization about Chatchabee Tea and the sales side. I think those are more organic. that was my takeaway is that it's not a coincidence this happened. It wasn't like start with the initiative in mind. It was start with the foundations in mind, the mindset, the culture, the training, the tools.
and then let the organization sort of
Forum3 (41:35)
Thanks for joining us for this episode of AI First with Adam and Andy. Conversations like this one are why we continue to bring leaders together. When organizations share what is actually working in the real world, it helps everyone move faster and make smarter decisions about how to adopt and apply AI. Thank you all for listening to AI First with Adam and Andy. For more resources on how to become AI First, you can go to our website, forum3.com.
download case studies, research briefings, executive summaries, and join our email list. Also, we want to invite you to connect with us on our AI First community, a curated hub and network for leaders turning AI hype into action. We truly believe you can't over-invest in your AI learning. Onward.