Intro:    

Sid 0:00 

AI is everywhere right now. New tools, new headlines, new promises about productivity and automation. But most conversations about AI stay at the surface level. They focus on the possibilities instead of practical applications. So what does it actually look like when AI is implemented inside a real business in the contract interiors industry? What problem is being solved? How does it change the way teams work? And what happens when you start capturing decades of industry knowledge and turning it into something accessible, scalable, and intelligent? Today, friend, that's the conversation that we're going to explore with our guests about what happens when innovation moves beyond theory and becomes part of the workflow. We'd like to thank our presenting sponsors, Avanto, services and software that streamline how you operate and the collaborative network, a platform where leaders in the contract interiors industry unite.

Meet Threeform And Bitreel

Speaker 3 0:59 

Is your team overloaded? Avanto helps teams save time and money by using AI to handle daily work from order management and AP processing to task handling without adding headcount. It works inside the tools you already use, with no new dashboards to manage. Go to goavanto.com to see how we can run your daily operations.

Sid 1:18 

Welcome back or welcome to the Trend Report, your inside look at the people, products, and ideas shaping the future of workplace design. I'm your host, Sid Meadows, and I'm glad that you've joined me today for this conversation. And I'm excited to welcome back to the show Carly Slocum with Three Form, and welcome to the show Cosmos Kramer with Bitreel. Now, together, they've been working on some really fascinating applications of AI inside of our industry, and we're going to dig deep into this, what it looks like. And I think this conversation is a great example of what happens when forward-thinking leadership meets practical technology. Welcome, Carly, and welcome, Cosmo. Great to have you on the show today.

Karli 2:09 

Hey, Sid. Thanks for having us on.

Cosmo 2:11 

Yeah, thank you, Sid. Good to be here.

Sid 2:13 

Well, Carly, you were here at episode 112, which is a long time ago. So I don't remember what year it was, but it was episode 112. We had a great conversation then about something very forward thinking that you guys were doing at three form in virtual showrooms and showing the product. And now you've gone even further. So before we go further in today's conversation, Carly, take a moment and tell us who you are and what you do at Three Form.

Karli 2:37 

Yeah, so I am Carly Slocum. I'm vice president of product and marketing at Three Form.

Sid 2:44 

Well that at ThreeForm. That was quick, yes. That was quick. Cosmos, what about you? Tell us who you are, what you do.

Cosmo 2:51 

Sure. I'm Cosmacramer. I'm the CEO at BitRail. And yeah, we're a uh tech startup. We do two things, really visualization, and then on the data side, we have this AI knowledge agent that we're going to be talking about today. And uh yeah, we've been working with Three Form for a couple of years, and it's been great partnering with Carly and her team and looking forward to discussing what we've been working on.

Sid 3:12 

Well, I'm excited to be here. I was really excited to learn about it in our discovery call as we dove into what you guys are doing, and I'm really excited to share it with the community today. So as we dive into it, I mean, Caso, you got lucky in working with somebody like Carly and a company like Three Form. Talk about being on the cutting edge of not just products and innovation in the industry, but definitely forward thinking about what can they do, what should be done. And that was you know evident way back in episode 112 when Carly was here before about there talking about. So did you win the lottery by you know tapping into Three Form?

Cosmo 3:46 

I I think we did. Yeah, out of all of our partners, Three Form were the first people we started working on this, probably because Carly and I were just chatting about it. It's good. We we have a we have a pretty good dialogue, Carly, I feel like in the background of all the stuff we're working on. And yeah, I think it was nice to be able to go to Carly and say, hey, we're thinking about these ideas specifically for AI. Would you guys be interested in trying to see if there's real business value there? And as you mentioned in the intro, you said something that's practical. At BitRail, we're only working with folks in A and D. So that's very much our focus. And then within that, we're looking to derive as much value as we can from the tools that we're working on with our partners, and it just seemed like a good fit based on everything we've been doing with Threeform. And yeah, in true threeform and Carly Slocum style, she was she was game. So yeah, I think we did win the lawyer in some sense.

The First Problem To Solve

Sid 4:37 

Always nice to have a customer that's willing to walk out on the plank a little bit and try something new. Yeah, right. Yes. So, Carly, let's let's talk about this for a second. When you guys said, okay, this AI thing is here, there's lots of chatter about it. We need to look at how we can use AI. What was like the business problem that you were trying to solve when you started down this path?

Karli 5:02 

Well, I mean, there's a couple different ways that we're using it, and a couple of which specifically with Cosmo. And to kind of continue the love fest, he's giving me too much credit here. I think ultimately, you know, Cosmo and I got introduced, and I think we both had these ideas of what could be. And I sort of challenged him. I was like, listen, we make translucent materials, we sometimes put leaves in them, we sometimes put fabrics in them, sometimes we're doing completely different things with texture, and we really struggled to figure out how to embrace some of the newer technology to show our materials in a digital way without it just being wrong, right? And when you're working with designers, you really have to nail those aesthetics. And so from the beginning, Cosmo and our chatting, Cosmo was like ready to take on this challenge. So he he isn't giving himself an extra picture. But that was kind of how our relationship began. I was like, listen, if you can figure this out, we are we're set. So he was like, you know, confident in uh, you know, taking on that challenge. And he came back, I I don't even remember how long it was. It wasn't very long, probably a couple weeks, a month maybe, and came back and showed us some early prototypes of what he was working on. And then I was just really impressed. And I think, you know, when we talk about the partnership in general, I think what's made our partnership unique and what BitRail has done is they haven't come in and said, hey, we've created this thing and you have to fit in this box. And that has never worked for us, and it probably doesn't work for a lot of uh manufacturers either. And what I've loved about working with Cosmo and his team is they're very flexible. They look at everything as a learning opportunity. And I think through both visualization and this AI agent, which we'll talk about both today, I'm sure.

Sid 6:58 

Yep.

Karli 6:58 

Both of those I think have been a learning experience for us both, but for which I we can go back and leverage as an organization. But then, you know, I'm sure BitRail can then go back and leverage and help others, you know, do some of the things that we're doing with these tools.

Visualizing Translucent Materials Online

Sid 7:13 

Where was the first place you started? So obviously you guys met, you talked about things, Cosmos came back and said, hey, we could do this, really creating a solution. What was the first thing that you took on that you actually got implemented?

Karli 7:25 

You referenced your previous show. We had just started working on our virtual showroom, and we were working on what we call visualization, which I think a lot of manufacturers are working on, so that designer specifiers can come to your website, go digital first, be able to see your materials in large full scale, see them in different hardware solutions, see them in different rooms. And we were struggling with that because of the translucent aspect of our material and because of the just huge variety that we have. And so uh Cosmo uh and I were introduced by a mutual friend of ours, and he was building out Bitreel, working with another furniture company, and I'll let Cosmo talk about that. But the um uh because I think that's gone well too, but I'll let him, I don't, you know, I'm not involved, so I'll let him talk about it. But um yeah, and it kind of just came this thing of like, all right, if you if you can do this transistent, you know, materials online and be able to visualize them appropriately, go for it. And he came back and I was just blown away, not only with like the quality of the material in a digital format, but also just being able to zoom in and be able to see everything and how quickly they were able to do it. Like I would have thought it would take us a couple of years, and it took us months to get there. Wow. That was another challenge I gave to Cosmo because we were we were launching a really big new hardware system called Array, and we wanted to show both our glass and varia offering in this system when we launched. And uh, we came real close, really close to nailing that deadline. Um, but I really only gave them like three months to get hundreds of our materials into the visualizer to put together five different configurations of our hardware system and and then to get it live in a very uh user-friendly way. So just really impressive. But yeah, that's kind of how we started. And then from there, and you know, we've gone to the AI agent sort of world too.

Sid 9:29 

So we'll get there in a minute for sure. So visualization tool cosmos, did you think she was crazy when she came in and gave you three months to do something? Not at all.

Cosmo 9:37 

No, I I don't think we we like a good challenge. And yeah, Carly's giving me credit. I should be giving credit to our team because we've got a bunch of it's basically all engineers, like our entire team is just technical people, and they love a good challenge. And this was a real challenge, like Carly said, it was really hadn't been done before, not just in the industry, but really anywhere, doing this through a browser. And you know, Carly said to us early that they wanted this to be accessible to designers and specifiers on the website, and it needed to be quick to load and you know, not slow down the side, things like that. So, no, I think we were we were excited for the challenge. We had a sense that it was gonna be that it was gonna be possible because of some other things we've been poking around with. Like our team's always doing RD, and we've done some stuff with semi-translucent materials, so we had a sense that it can be done. And our company was really founded on a bet. We made a bet when we started the company around visualization in general and what's called real-time rendering, and doing that on a browser. And the technology wasn't there when we started the company, but we figured within about 18 months it would be possible on the hardware that everyone has, and that turned out to be true, and so I was pretty confident placing that bet when Carly, when we first you know, started speaking, and she said, Hey, this is what we're looking to do. Um, I was pretty confident, but I think we structured it like let's do this as a test. Like, you know, we didn't want to oversell. But as Carly said, I think when we were excited to show her because when I saw what the team had put together, I was like, I think we're getting pretty close here. And we're still refining, it's great. We work with Carly's team who are looking at the physical materials, and then we can look at the virtual representation and make sure we've got fidelity between those two. And so, as she said, we you know, we did the first launch with a Ray that Threeform brought out late last year. And now we're able to say I've kind of rinse and repeat, and we're gonna work through the you know the rest of the products they have and and more materials. And um, I'm sure there'll be other fun things we'll get to, but it's been it's been great. I think it's just been a nice project to work on and yeah, challenges, but I think we're we're putting out a good tool that is is useful for folks in the industry and hopefully helping threeform you know get closer to their customers on the web.

Sid 11:45 

Yeah, thank you for that. Carly, this is live on your website. So a visualizer in the furniture world we call them maybe configurators, where you can pick your fabric here and there, pick your finish, and it comes up on the screen right in front of you, and then you can download it, request a quote, all that kind of good stuff. So is that what the visualizer is? And is it live?

Designing For Speed And Fidelity

Karli 12:04 

It is live. It's alive on our products array, is what it's called. And we've toyed around with is it a configurator, is it visualization? We tend to call it visualization because what we are trying to accomplish is allow somebody to see what they usually see in like a four by four or eight by eight sample size at a large scale. And our and our uh materials are basically sheet products. So they come in four by eight or four by ten. So sometimes, especially a designer, you know, you're you want to see what that pattern may look like at scale. You know, maybe it has a very specific texture and a color blended in with it, and then and then they want to turn around and they've got to show that to their end user. And so the idea being is that they can quickly visualize the different materials, and then they can download the hardware system and the materials and put it into Revit or whatever design program that they're using and really finish the configuration there. So determine like what do you want for dimensions, what does that look like, what's the true sort of shape that you're trying to create? Because really, ultimately, we could do that on our website, but then you'd have to redo that in Revit. So we're trying to find the best balance between what do you actually want to achieve on a manufacturer website versus once you've kind of said, yeah, that's what I want, where are you gonna take that from there? And we don't want to create a situation where you're doing duplicate work. So that's that's where we've landed, and that's why we call it a visualization tool.

Sid 13:40 

So does it go all the way to application, meaning I'm gonna put this three-form material on the elevator bank, I'm gonna put this on a piece of furniture or in a or front of a reception desk. Does it go all the way to that to that detail, or is it just showing you the big piece with the hardware on it?

Karli 13:55 

It's really just showing the big piece in the hardware. So and that's why when Cosmo is saying we're rolling it out to different hardware systems, we do provide our material selections in Revit. So if you're doing like an elevator cab, you can go do that. But that's not a standard hardware solution that we provide. So we're essentially doing it in all those standards. And then the Revit file has all the rules, right? So you can take it into Revit and then you can decide, you know, how big you want to go and everything else, and then see that pattern and repeat. But again, it's mainly just to visualize it on site, especially since if you're familiar with Reform, we can do like 30 million different combinations of materials. Like you can add a texture and a different color, and you can accomplish so many, you know, we have 250 colors you start with, but the color options are actually limitless because we can mix those colors. And then we have all these patterns that you can mix with color. So it's like it's actually impossible to sample all of them. And so it's a way to kind of see maybe what you want, try to get a better understanding of it. And then like we can of course always sample exactly what you're looking for, but it makes it so you can do a little bit more of that up front before you need to reach out to somebody and start that whole sampling process.

Sid 15:09 

So every furniture seller, every dealer, interior designer, every interior designer at a firm has looked at a four by four sample or a 10 by 10 sample and saw something. And then the actual product comes in and they're like, that's not what I picked, or they see the big piece of it and three yards of fabric or whatever, in your case, a big sheet of you know, your material, and they're like, that's not what I wanted. So the fact that technology has come so far that you can truly see the whole thing, the whole pattern, the implications of it, and in your case, the customization of it, it's really important for our industry so that less mistakes. Absolutely.

Karli 15:48 

Yeah, and that's a big part of it. And so Cosmo and I've talked about, I mean, translucent materials are unique. You know, the lighting in the space is absolutely going to change what that looks like. So we've talked about what do we do about allowing a user to play around with natural light and and different types of light that they may have. The other thing that I think is is really interesting is the way that Cosmo and I are building this is that three form essentially what we're making are like resin paninis, right? So we we have sheets of resin and then we have color films and we have organics or textiles and things that go between. And in order to properly depict this in a digital way, you have to actually build it digitally as well. So you can't just photograph the end product and it work. You have to photograph each individual layer and then Cosmo takes it and builds it digitally. So it's actually a very accurate representation of what you're gonna see in real life. Wow.

Moving From Visuals To Knowledge

Sid 16:48 

So the details are there. So Cosmos, I want to come to you because you know, Carly just laid out what I would view to be quite a few challenges. Taking the product and photographing it individually and then putting it all together so you can clearly see what it is. But she also talked about 30 million different combinations, 250 colors. That's a lot, and that's not necessarily uncommon in the manufacturing side of our industry with all these options and then the specials capability with it. What did you do to get the AI to actually realize this is what you need to be doing? Because we all know, and some of us have experienced it. Unfortunately, AI hallucinates with things. Like, how did you set the ground rules and stuff here so that it would actually work? So when the user, the designer in this case, the experience was exactly what Threeform wanted it to be.

Cosmo 17:39 

Yeah, it's a great question. And I would say the the AI agent um that yeah, I'm excited to talk about now is uh is a little different from the visualization. I think Carly and I have talked about there's certainly ways to connect the two and have one feed the other. But if for now that's separate, I will say you your question about 30 million different combinations and things like that, of course, is relevant to the AI side, and that's one of the that's one of the reasons we started talking to Carly about the AI agent, because three form, as you mentioned, is a lot of complexity, there's a lot of combinations you can have, there's uh different parameters for different products and different materials. So we felt that it was a natural fit to have uh an AI knowledge agent essentially helping. And I think first phase is helping the three form team internally. And one of the things we've been working on is yeah, how do you how do you provide all that information to an AI agent, whether it's 30 million combinations, how can you possibly essentially train it to understand three form and be um and be an expert on threeforms products so that it can help you know internal people that work to the company for 20, 30 years? So that's been the first thing we've been working on together essentially.

Sid 18:50 

Okay, so let's take this and break this down just a little bit. You got an agent, like customer service agent almost. What exactly does the agent do and where does it live if it's live yet?

Guardrails That Stop Hallucinations

Cosmo 19:01 

So it's it's it's not live yet. We're still in testing. And Carly, feel free to jump in if uh if you want to layer on anything. But I would say based on what Carly has been working with her team on, there's a couple of initial use cases that we're starting to look at, internal use cases. And what we've been doing on the BitRail side, I'll let Carly speak to three forms. The thing that we've been uh working to nail, and the thing that we spoke to Carly and immediately about was accuracy. And I guess to what you're saying, Sid, like producing hallucinations. Our thesis at BitRail um was that there's a big opportunity here for A and D and contract manufacturing, this specific part of the industry. One, because we know it's been working in this era and just focused on this for a number of years now, but also we felt like there was going to be a ceiling for them, these sorts of manufacturers using like a traditional like chat GPT off the shelf because it has those hallucinations, and also it doesn't know the intricacies of the company or the products. And so that was our discussion with Carly. It was like, hey, we can build this that will have accuracy in mind, and there's a few things we're doing there. We call them guardrail checks, and it's essentially it's checking if the if the uh LLM or large language model is hallucinating and it'll go back up and recycle through. So we can talk more about that.

Sid 20:19 

Okay, so I want to I want to pause there for a second. How do you make sure that it's checking it accurately?

Cosmo 20:24 

So essentially what we do is we have what's called a knowledge base, which is just a bunch of information that's coming from three-form systems. There's two streams, there's structured data and unstructured data. And then once we have all that data in the knowledge base, we let the agent go through and find what it wants to use as context and then answer the question. Before the question goes to the end user, who will be someone from the three form team internally, we run what's called a guardrail check and it runs through millions of rules. And all of those rules are populated based on three forms products. So everything Carly was talking about before, the layers of the panini that she was mentioning and which layers can go together and which cannot. We load all of that into a system, and then you end up with millions of these rules. And based on a very intricate algorithm, it goes and runs and it'll know which rules to check. It checks all those rules. If it passes all the guardrail checks, it'll continue to send the response to the end user. If it doesn't pass, it'll go back and say, you've, you know, it'll flag five rules and say, You've break, you've broken these rules with your response, please recheck and it'll go back and style through.

Sid 21:27 

Before I ask this next question, I'm never going to think of a panini differently ever again. I'm always going to see three, four material every time I get panini. So you talked about millions of rules. Now, these aren't rules that Carly, you had to write. These are rules that pretty much already existed in your business through the development of your product and your price books and your warranties, and you could use it here, you can use it there. So you didn't have to sit down and write millions of rules, right? You used what you already had, you just had to ensure that it was accurate. Is that correct?

Karli 22:01 

Yeah, that's right. We I mean we have a couple of different areas that Cosmos team are basically consuming for this AI agent. One of them is what we call our product database, which includes all what the Panini layup essentially, what how you make the sandwich. So that's one. But then the other is, and I'm sure there's so many out there that have the same problem. We have so much knowledge that lives in people's minds. We do a lot of custom projects. We really nothing we do is done over and over again, right? It's always some some sort of uh unique application. And so over the years, we've we've done our best to document some of that stuff. And so that a lot of that is captured in our text specs and our um installation documents, things like that that the AI agent has consumed. But we're also working on making sure that there's a lot of still information out there that needs to be documented. So that's obviously a part of it. The other piece that I've been really interested in is through this process, we found out that this AI agent can actually have a back channel to us to help us make sure that if we have documents that are like if we have documents saying two different things, it can alert our product management team and we can go make sure that those documents are correct. So there's a back channel that this uh agent can give us. So not only can it be forward-facing and solve business problems like we have so many different things we can do that the training time to get a customer service person or an estimator or a salesperson up to speed is very, very long. So the idea is maybe we can shorten that up, but also make sure that we're consistent on every single phone call that we're having, that you're not gonna get one information by a more experienced estimator than another. But also that back channel, like, you know, we can now be alerted if there's something we missed out there that we meant to update and we didn't do it, you know.

Capturing Tribal Knowledge At Scale

Sid 24:00 

So there's a couple of things I wanted to talk about. I want to go backwards for a minute. When you talk about capturing knowledge, you're what you're really talking about is tribal knowledge of your employees that have been there for a while. And this customer service person, this engineer worked on this project, and it's maybe documented in a couple of different places. There's probably, you know, dozen or more emails about the project that's got technical information in it. So, how are you gathering the tribal knowledge? Are you giving them a survey to fill out? Are you interviewing them and transcribing it and uploading it to the like how are you gathering the tribal knowledge? Because this is, I think, a problem a lot of manufacturers have is the knowledge of their people. When people leave andor retire, that knowledge goes out the door. So, how are Cosmos, how are you helping them capture the tribal knowledge?

Cosmo 24:46 

Yeah, there's a few ways. One of them is an interview. So the agent, in addition to what Carly was mentioning, where it can kind of back channel and send logs and information on conflicting data, it's also able to essentially interview those people to kind of pull out that tribal knowledge. And we talk about democratizing tribal knowledge as being one of the benefits of using this agent. So it basically goes through, we have different personas um for the agent depending on where you're using it and what the use case is. So one of the agent personas is this kind of interviewer that'll like interview someone uh on the technical side at three forum that is working on custom projects, for example, and it can go through and just ask a series of questions. It's just like having a conversation with a colleague, and during that process, it's getting some of the information, and then there'll be follow-ups that'll want to go and ask and get more information. Um, so that's one of the things we're doing to try to tease out that that knowledge from folks. And then the other thing we're finding is as we go deeper, and Carly mentioned getting access to like product databases and there's other PDFs and things like that on their on their TextBecks page. We're finding, I think hopefully in a good way, as we all work together on this, we're finding other sources of information that might be interesting to connect to. And one of the things we've found is that we're very conscious, at least of BitRail, of keeping kind of the context pure, which means you don't pull in a bunch of information that's not relevant. The agent might get a little bit confused. But what we've found through some experiments is you can pull in pretty much everything, even data that may seem out of date or you know, a little bit stale, and we can demote the focus. So the agent has access to it and it can kind of take it into account, but it won't consider it like source number one or source number two when it's responding. So that was the concern for I think a lot of folks was they were very worried about giving it access to a lot of information because they hadn't fully scrubbed some data, for example. We're finding that's not really an issue that you can kind of you know fairly quickly and easily demote focus on things that you wouldn't consider like gold standard and still make it available to the agent, because in some cases it can be useful. So that's been a helpful exercise. We don't have to be too concerned about giving it access to things that might seem out of date. So yeah, but we're still, I think we're still learning as we as we work with three forming this, we're just finding more and more sources that probably get you know helpful to connect to.

Sid 26:58 

We'd like to thank our supporting Silver sponsors, KISP, who helps furniture brands visualize and sell products before specification. Navitas Credit Corp. Navitas makes great workspaces affordable, turning furniture projects into easy monthly payments. Web Configurator Services, the leader in AI-powered online product configuration and quoting. Selling contract furniture is getting harder. You're dealing with complex products, disconnected tools, and buyers who make decisions before you ever get to spec. KISP helps fix that with clean product data, visual sales tools, and expert visualization services built for the way furniture actually gets sold. Backed by 30 years of experience, KISP helps dealers and manufacturers visualize, configure, and get chosen early before specification ever starts. To learn more, visit KISP.com. So I think you might be looking at my notes because you actually led me right into a follow-up question that I had just a minute ago, which was about conflicting data. And I love the back channel. It's kind of like it's your own private chat and the AI saying, hey, you've got a mistake here. You need to validate which is right so that I know what information to give you. And I'm just going to be very practical, simplistic form. You could have a product that was designed and developed, and then you have an engineer that works on a special, or maybe you had a new engineer come in and make a change to it, and the original thickness was 38 of an inch, but the second engineer changed it to 316th of an inch. And don't you just be impressed with my knowledge, whether that's your dimensions or not? But but and all of a sudden now the AI is looking at a technical spec for the same product. It's got two different dimensions on it, and the probability in the human world of giving the customer the I mean, imagine you tell the designer, oh, it's 316th, and it shows up at 38, the problems that that causes. Now this back channel is going to say, hey, you've got two different bits of information here. Hey, three form, you need to go fix this. So I love that. Managing the conflicts because they exist, because we're humans and humans make mistakes.

Cosmo 29:06 

Yes. And we've also found that exact situation. It's been really nice. We've um we're kind of loosening the reins with the AI agent, giving it more flexibility so that if it encounters a situation like that, initially, now because we're focused on internal use cases, it's completely fine to tell the user that the agent's chatting with. Basically, what you just said, Sid, that it's, you know, it needs to go and it's flagging single conflicting information and it'll kind of respond. The good thing is um, in that case, in the past, I think what a lot of people were doing with these AI chatbots or agents, is it was you kind of considered it like a one-shot interaction where the user would ask a question and the AI would have to respond. Essentially, like they had one shot to give the answer. And that's why, you know, with hallucinations and things like that, it was a real problem because people would take the response as gospel and and run off. What we're doing now with uh threeforms knowledge agent is it becomes this very flexible process where the agent has a lot of a lot of leeway to go and do different things. It can call what we call tools, so it can go out and search the web, it can send an email, it can do a few different things. And then the biggest thing we're finding is it can actually ask questions, and then based on the responses to those questions, it can kind of deviate from its response. So in that case, what we're talking about with conflicting information, it might ask the user just because it's all internal now. Like, would you like me to reach out to the team and I can get back to you in a few hours or you know, whatever it says, or it can say I can give you these two bits of you know conflicting information and you can decide what to do with it. So we're just trying to make it really flexible so that we're not stuffing it into too much of a restrictive box because that's when I think the process can kind of break down. We've seen like the more flexible you can be and giving the agents more leeway, the better the result is for the user that's chatting with it.

Sid 30:50 

So I'm interpreting that as because we've all experienced this. You log on and you know you're talking to an AI chat and you want to ask a question, you went to help for a reason, and it gives you five things to choose from. And if what you need help with is not one of those five things, you are S O L. Like, you're not gonna get your question. So it's not like that. It's very called you're laughing because you picked it up.

Karli 31:13 

Yeah, and my favorite is when you click on one of them and it goes to a broken web page. It's like, okay.

Sid 31:20 

Oh, that's great.

Speaker 31:21 

We fixed this, yeah.

Rolling It Out To Customer Teams

Sid 31:23 

So this is funny. I had an issue with Zoom today and it disconnected my email. And so I went into Zoom and it said, you need to put your E U E W S URL. I'm just butchering what the acronym was. I'm like, what's that? Yep. Like, and so I go into Microsoft, I type in E whatever URL. I get three pages worth of help desk notification, you know, articles to read. I go to Google to like type in what this is. My wife's an IT person. I bring her in, like, what is this? Like, what is this thing asking me to do? Miraculously solved itself. I never found out what it was, I never found out where it was, and I never got the help that I needed because we've all been there with those levels of frustrations, right? Of things not working, especially a designer in your case, gonna be working on something at eight or nine o'clock at night and needs the information and they want to be able to get it without having to wait. So I love that you're not doing those boxes, and this is the parameter, and you're giving it flexibility. But Carly, I want to ask you who at Three Form is using this agent and how are you really going to, once you roll it out, how are you gonna use the agent?

Karli 32:31 

Yeah, we have a couple ideas. So we have a test group that's working using it right now. That's a combination of several of our product engineers, both on the standard sort of typical products we put out, but we also have another group of engineers and they deal with like the highly, highly custom jobs. So they might be even doing like, you know, full custom hardware. So we have all of them working on it, and then our product management team working on it as well. The goal being is that as soon as we feel like it's at a pretty accurate point, then we're gonna roll it out to all of our more customer-facing teams. So I mentioned our customer service, our estimating teams, and then our sales teams. The idea being is that in the past, if you get asked a question that is really kind of out there, or maybe it's not just part of your daily business, you'd have to go find documentation, or you find somebody who's been here a decade and ask that question, right? So you're not getting an immediate answer. And then therefore your customer's not getting an immediate answer. So the goal is that you know, we all kind of know the standard answers to our questions, like what gauge our materials come in, all that kind of stuff. But then as soon as it gets more complex, you know, it's hard to keep all that in your mind, right? And so the idea being is that we can arm people with that information, that also there will be consistency in those answers across the board. Because again, more complex, the more likely you're gonna get different answers from different people. So that's first, and then just get that company wide. But then second, it goes back to what Cosmo is mentioning. Once you can visualize our materials, I think it would be ideal for a customer to be able to say, okay, I like this, but can I do this? Can I do that? What if I wanted to put it into an elevator bay, like you mentioned? What does that mean? What gauge do I need? Does it meet the fire code? All that kind of stuff. And instead of digging through documents, it will just tell you and it'll give you the source for it. And so that's ultimately uh where we want to go is both a tool that helps internally, but also one that our customers can interface with. But I will real quick mention, like, because I think to me, what you're talking about is what's really cool about technology is your example of when you get on, you get the five sentences, like, you know, do you want to know our return policy or whatever it is? And then you're like, you know, you used to call somebody, right? And then you start using chatbots and you pre-program these chatbots to assume they know what you're looking for. And it never seems to be what you're looking for. Because it's just web pages that you can already find. And now I feel like what we're doing is we're kind of doing our best to emulate what that human answer would be because it was it's way more educated. You have more of a conversation, but then you can do it at any time of the day, right? You can you can be on your computer at nine o'clock on a Friday night if if that's when you choose to work, and then you can get those answers. And so it's it's a little bit like this interesting cycle that technology takes you on.

Sid 35:28 

Or your customer in the UK who wants to specify your product is trying to find out information at eight o'clock in the morning, which is you know, one o'clock in the morning your time, or two o'clock in the morning your time, and nobody's working at three form, they don't have to wait five, six hours to get information that's gonna be there. So go ahead, Cosmos.

Cosmo 35:46 

And I was gonna say, and I think this is this is a competitive advantage for threeform is I think uh pretty soon those designers or specifiers, they'll they'll expect a fast answer. And if they find a manufacturer that doesn't have that capability, they're just gonna move on to the next. I think it's potentially like where we're going. They're not gonna want to wait around and you know, wait the two or three days. If three form can give them timely and accurate information about products they want, it might, you know, it might win a deal over a competitor that has you know very slow feedback and things like that. So I feel like that's that's kind of an interesting element. Is like seriously.

Sid 36:18 

I think we're actually think we're already there. Okay, yeah. Yep. Because I had Crystal Lacera with Edwards Mulhausen, a design firm in Austin on a few weeks ago, and we were talking about how you call them the A and D community, and I asked her for some tips, and she said, if I ask you for information, be speedy to respond to the information that I'm needing. I'm gonna give you 24 hours. After that, I've already forgot about it and moved on. So I believe the expectations are honestly already there. Right now, they're being put on people. Like, okay, I got to respond to this, I got to get this done. What you're doing is enabling a tool, in the case of an agent, to help answer the question. Because we've all been there. The designer says, Hey, no offense, designers, we love you. I want to put this custom fabric in the middle of these two. What are the requirements around how thick it has to be? What is it, blah, blah, blah, all the whatever things that Threeform would ask. And the salesperson's got to call customer service, got to wait for them or email them, wait for that information to come back, and then he or she's got to get it, look it over, and then take what the factory sent to him or her and then put it in a readable document and send it to the designer. And what you know now is going to take a matter of seconds, I think, when you guys get it launched, was taking two days to get that information back. So definitely speeding up the the go-to-market strategy and how quickly we can respond. And the sales guy, me, there is there is success to be found in the first to be able to respond to a customer's request.

Karli 37:47 

Nailed it right there, Sid. No doubt.

Getting Leadership To Say Yes

Sid 37:49 

No doubt. It's all about sales. It's all about that revenue. We've got to accept that revenue. So I am just fascinated at what you're doing and how you're doing it. I have two more things I want to talk about before we wrap up. And the first one is to do with leadership. In our industry, we are fast innovators, but we are slow adopters. So, how did you get the leadership at Three Form to buy into literally, I said in the beginning, walking out on the end of this plank saying, Let's try this. What was it that made them want to do this, Carly?

Karli 38:25 

Yeah, I mean, I think my my best advice always around this is that if you start with a problem first, and then you look at what can solve that problem, that's your best use of technology. I think when you see people fail is when you're leading with technology and you don't quite know what problem you're solving, and you don't quite know how it's going to help your business. And I think that, you know, not just AI, but when you look at any sort of uh technology developments over the years, those that have been the most successful have looked at as a way to solve a problem, whether it be capacity, productivity, information access, whatever it may be, that's when you have a winner. And so I think when you can go to your leadership team and say, we all know this is a problem and this is what it's causing. Maybe it's costing a lot of money. Maybe it's uh maybe you're losing jobs because you don't have something at hand. But then you show like the solution, it likely in these days has some technology component, but when the solution shows how it would solve that and how it would potentially save money or get you more jobs or whatever it may be, it's pretty easy then to get leaders, I think, to buy in on. If you just walk in and say, hey, we gotta jump on this AI bandwagon. I don't know what's going on, but we gotta we gotta figure it out. Like, yeah, you you should be exploring it, but I think leaders would be like, what's the investment for? What are we gonna get out of it? You know? So it's gotta be more than just like, let's just everybody's doing AI, we gotta do it, you know?

Sid 40:01 

Yeah. Well, people will do that, just that phrase right there, everybody's doing it, we gotta do it. But you hit the nail on the head. What is the problem you're solving for? And what's the result of solving that problem? How is our business going to be impacted positively? More profitability, lower cost of ownership to our customers, quicker response time, better quality product, whatever it might be, you solving the problem first rather than leave with the solution first. So that's perfect. Customers, what was it like walking along Carly here and listening to the feedback that we were getting from the leadership team? And again, from your side as a consultant here, as a service provider, getting them to buy into adapting a new technology.

Cosmo 40:43 

Well, I think it was good. We had the track record with the visualization. So when we had the AI discussion, I think it was a natural progression in some ways. And like Carly mentioned, I think, you know, bringing it back together, providing information at some point, maybe rolling that in with like visuals, helping them package things up, to share to their end and customer, whatever it might be. But yeah, I I we just had a sense that the three-form leadership team of which you know Carly's on is fairly um willing to try new things with the right framing, which I was saying, based on you know, what is the problem to solve. And I think we also felt confident because after working with them closely on the visualization stuff, meeting a lot of folks on the team, when we ourselves were thinking about AI and the problems it could solve for three form, we had a leg up from someone that was just coming in cold because we were kind of we were already seeing kind of under the behind the curtain, so to speak, in some areas. And we were like, when Carly was, you know, chatting with me about it and she was bringing up use cases. Oh, we could maybe look at this or we could look at that. I had some context because I already knew a lot of the things she was talking about from the visualization work we did. So I felt pretty confident that we could we could nail some of those use cases with this AI agent because I I understood a little bit of a little bit better uh about their business, you know, based on the work we've been doing for the last year or two. So it was it was great. But yeah, I I was I I had a good feeling we were going to do something just because now I've gotten to know Carly and the way that she kind of gets projects approved. And yeah, so it was good, but I I I was fairly confident we'd get something going.

Sid 42:17 

So, Cosmos, you know Carly better than I do. I know her decently well enough to get a hug at Neocon. I know her that well enough for that, right? But dude, I would hate to be the guy or the girl sitting on the other side of the table telling her no. Because I can guarantee you she's not taking no from anything along the way there. She's smiling, she's plushing a little bit in the buttons. I don't know.

Karli 42:40 

But I don't know. You might be giving me too much credit there.

Sid 42:44 

But so Cosmos, I want to wrap up with you, and then I got a question for Carly, okay? Then we'll wrap up with her. I'm a manufacturer, I'm a brand. I've got ideas, I understand some of the problems. I can come to you and say, this is what our problem is, these are some of the ideas that we have, and you're going to help create a custom solution to work for that brand rather than put them in the box and say, you have to work inside these parameters in order to work for us. So you're truly going to build a custom AI solution slash tools that will help move a business forward.

Cosmo 43:18 

Yeah, that's that's exactly right. I'd say we consider ourselves like boots on the ground. We're not that industry agnostic. Like there's plenty of tools that are like, oh, we'll just do retail and they work with every company in retail, for example. We're very much focused on just A and D. And then, yeah, we like to get in and get and get our hands dirty because I think with these AI implementations, at least what we're finding, the biggest piece and the initial piece is connecting all the data sources. And for that, we need to be the experts helping the three form team. They know where their data is, but we need to be helping them with all the infrastructure, getting the stack set up, getting all the data floating. And to do that, you really have to be involved. It can't just be, you know, a SaaS platform that, you know, is self service. Like you've got to be a partner. We've got to be a partner. And so and and we're used to doing that and we really enjoy that. So I think we're a decent fit to work with companies like Three Form. But yeah, I'd agree. I'd say like, yeah, we're kind of in partnering hand in hand and and then we'll kind of stick with it where as Kelly mentioned, yeah, we'd like to be flexible ourselves and and not just have something that we're like, all right, this is exactly what we're gonna do, and we're not gonna deviate from that scope. But we're very much like, this is like possibly where we start, and then let's just see where we get to.

Karli 44:27 

I know that was a Cosmo question, but I couldn't co-sign that more because I think we've all experienced where you go in and you get sold a piece of software, but they just want to pigeonhole you into that software, and it never works. Like it never works exactly how each individual business needs it to be. And what has been so great about working with Cosmo is he comes in in like blank slate, and it's like you can define it the way that you need, and then it and then it gets built on that. So it's avoided all these conversations where it's like, well, well, we can slot that idea into here, and it's only gonna be like 70% of what you wanted. You know, like that's that's how I feel like it's been historically in my experience. So it has just been a complete game changer working with Cosmo and his team because of their approach. It's very different.

Sales Training With ChatGPT Scoring

Sid 45:19 

Well, I think every uh everybody listening has experienced what you just described. Oh, we can get up to 70% of it, and then we can custom code it for you later, but you can live with the 70%. We'll make this work and square peg round hole, right? So it's really I I'm fascinated by the whole thing of AI. We've had a lot of conversations on the show about use cases of AI, the practicality, the guidelines, governance, and we will continue because it seems like everybody that I talk to or interview wants to talk about ideas or things around AI, and I'm all about it. Let's bring practical use cases to the community here. And with that, Carly, I'm gonna wrap up with you with the kind of the last question that I got here, which is you built without Cosmos, if I recall correctly, you built the sales training AI because your salespeople were complaining about having to do role play with their bosses. And so tell us quickly, what did you do and how is this thing working?

Karli 46:15 

Well, I would just let this go without talking about it. I've been adjacent to the project, but essentially, yes, we started to do a role play with our sales reps about two years ago, universally hated, right? Just, you know, absolutely did not like it. And and we all understand why, right? We don't have to go into all that. But it was actually my boss who he's a big fan of Chad GPT. I mean, he talks to Chad GPT every day, calls him calls her char.

Sid 46:40 

But anyway, he has his name. Whoa, whoa, whoa's the name to him.

Karli 46:44 

Chat GPT Char. He's gonna love the night, yeah, Char.

Sid 46:48 

Oh no, because I'm one to name mine, and I'm trying to figure out how to name it.

Karli 46:51 

I know I actually named mine and then I forgot. Now I go back to anyway. But so I think he was trying to, he was essentially just trying to solve this like delta between, you know, the reality is that when a sales rep is giving a presentation to an end customer, it is different than when they're doing role play. And then how do we give really meaningful feedback? And so he engaged with our sales management and some other uh people, figured out how to just essentially train Chat GPT into what we're looking for, what we think of as a successful sales call. And then the sales reps can now just record themselves doing a real sales call with their customers, just record it, upload it later to ChatGPT, and it will score them based upon how they performed on a couple of different factors, right? On how well they represented three form, how well maybe they represented a most recent launch, um, what was their closing ask. And they're way more comfortable with that because it's not like a peer-to-peer thing. It's actually a more organic recording of what they're really doing in regular life. And then they can truly take that feedback and think, oh, you know, maybe I'll focus on doing this better next time. So it's been way more well received by our sales reps, and I think more effective in general because of the way that you do it and the feedback that it gives you. Very easy thing, honestly, to set up as long as you're kind of knowing what you're looking for, and then you can kind of train. And again, we just use chat GPT for this, and you can train it on what you're looking for.

Sid 48:23 

Now that's really cool. I mean, that's a great use case of ChatGPT. And you're right, salespeople hate role play, so much judgment wrapped around presenting to your boss. You have this different level of nerves when you're presenting to your boss or your boss's boss versus presenting to a customer that you do all the time, right? So I love that just record it and upload the recording, which there's so many things today you can use to record, and then ChatGPT can analyze it for you. So I used ChatGPT recently, who by the end of this week will have a name. So I said to it, act as a sales and marketing coach. And I started inputting the problems, challenges that we were having going through, and feedback we were getting. I actually put transcripts in of sales calls as well as transcripts in of meetings that we've had. And dang, he's reading. He's calling me out. He's calling me out. Like said, you know better than that. It was funny because I'm reading, I'm like, oh man, he is right. Yes. Like I exactly what I did. So I think it's funny, but you have this non-human thing over here that can actually really tap into just reading stuff that works.

Karli 49:30 

Well, and that's a big part of it. I think it's human specific to role play. You don't want to hurt somebody's feelings. And you may notice that they did five things weird, but you're only gonna say one or two because you're just like, I can't do too much, right? And so you take all that out, and it's just it's a it's a true analysis of it. Yep. And it's going to hit all five points that you need to work on. And yet, and then, but as a person that's receiving that feedback, I think it's more well received too, like versus a peer-to-peer. So yeah, it is nice.

How To Connect And Final Thanks

Sid 49:59 

Totally agree. So, Baldwin, I can't thank you enough for this conversation. So enlightening to see again how three form is continuing to lead the way in our industry with innovative ideas and products and helping to really shape and push our industry forward into this new wild west of AI implementation. So, Carly, thank you for coming on again. Cosmos, thank you for being here and sharing what you guys are doing at BitReel. I will ask you, Cosmos, we'll start with you. If our community would like to get in touch with you, what is the best way for them to do that?

Cosmo 50:28 

Yeah, they can go to uh bitreel.com, b-i-t-r-e-e-l.com, or uh feel free to email me, cosmo at bitreel.com.

Sid 50:35 

We'll make sure LinkedIn profile, website address, all that good stuff is in the show notes for Cosmos and for Carly. So, Carly, same question to you. If our community would like to get in touch with you, what's the best way to do that?

Karli 50:45 

Yeah, uh come to our website. It's 3-form.com, so the letter 3-form.com. And then it's Carly Slocum spelled with a K. So you can K-A-R-L-I, you can find me on any platform, really.

Sid 51:00 

If you do reach out to them, please let them know you heard them here on the Trend Report. That's why you're reaching out. You wanted to learn more about them and connect them. This was fascinating. Appreciate you both coming on today. We'd like to thank our community bronze sponsors, Catalyst Consulting Group, ReSeat, and Staffing Plus. And to you listening today, thank you very much for being here today and for joining us on the Trend Report, your inside look at the people, products, and ideas shaping the future of workplace design. Go out there and make today great, and we will see you in the next episode. Take care, everyone.