commonsku AI Roadmap & skubot Agents | skucast Ep 369

"There's this sort of debate going on where it's almost pitting AI against humans; I see the opposite. I see that AI is actually creating an opportunity for you to be even more ingenious with the product that you end up creating."

— Bobby Lehew

"We think the potential value here of that co-pilot model is extreme … even for super senior people to move more quickly, to be able to spend less time on the configuration of the actual products and just let them focus more on the true value of what they're offering to their customers."

— Charlie Moscoe

[00:00:00] [Intro music]

[00:00:06] Bobby: Our product team set a goal this year that sounds audacious: eliminate close to 80% of the extraneous effort our users spend on the platform. Less clicking, less configuring, less chasing — which means more automation, faster projects, and less overhead for commonsku customers. Now, consider what that means.

[00:00:23] Thirty-eight million clicks happen on commonsku every month, and half of them are product configuration, because merch is a product-centric business. But there's a lot of great work we're aiming to streamline, and we have an A player we're putting back on the first string to help us. A few years ago, we benched our mascot, skubot, burying him behind the branding.

[00:00:40] But this quarter, skubot is on the A team. He's been given a job in AI. On today's show, Charlie Moscoe, our VP of Product Management, shares how skubot is a part of our vision for AI in the platform. We talk about AI in commonsku, automation, and our roadmap goals for the remaining year. We kick off this episode with Charlie interviewing me about what I've learned [00:01:00] about the industry and in my role here as Chief AI Officer.

[00:01:03] Welcome to the skucast, the podcast for innovators and maverick thinkers in the promotional product space. My name is Bobby Lehew. I'm glad you're here. Now, before we get to our episode, a quick AI tip and a question for you: are you tracking your team's wins with AI? We are. We have a Slack channel where we share what we're building with AI, week after week.

[00:01:22] It has done more for our team's AI fluency than any training program could. I hope it encourages you to start one. Here's a sample from ours, just from these past few weeks in July: Eric handed out some prospect outreach to an agent and got two hours back every morning. Dieter replaced 23 hours of meeting review with a 15-minute automation sweep, resulting in an 18-month roadmap of automation fixes.

[00:01:45] I wired a live onboarding dashboard with client health scores. Dickron created a coaching tool that scores sales discovery calls. Dave turned hand-built supplier performance decks into an automated analysis. Eku packaged a [00:02:00] build-your-own voice skill and shared it with the entire company. James distilled 2,000 customer questions into a searchable FAQ intelligence tool.

[00:02:07] Lindsey grounded Claude in our call transcripts to answer a 15-question customer inquiry. Allison tested our new onboarding experience by having AI play a skeptical, non-technical distributor. I pointed Claude at years of skucast history — over 300 episodes — and it compiled a catalog of more than 3,000 interview questions in minutes, organized by category.

[00:02:28] And through Claude Code, I put agents to work assembling our AI Promo Brief newsletter. Now, that's one team, one Slack channel, and a few weeks of sharing. If you're not doing this yet, carve out the time. It compounds the wins. Today's episode is brought to you courtesy of us at commonsku. Over 900 distributors powering 1.8 billion in network volume rely on commonsku's connected workflow.

[00:02:49] Process more orders, connect your team, and dramatically grow your sales. To learn how, visit commonsku.com. Now, here's my chat with Charlie.

[00:02:57] Bobby: Charlie, welcome back to the skucast. [00:03:00]

[00:03:00] Charlie: Bobby, always a pleasure. Happy to be here.

[00:03:02] Bobby: So this is gonna be a fun episode, because we both get to unpack the topic of AI, and we get to interview each other. So I'm so glad you could join.

[00:03:11] Charlie: Yeah, I figure we nerd out on AI all the time now. We might as well do it publicly for once.

[00:03:17] Bobby: True.

[00:03:18] Charlie: Cool. Do you mind, Bobby, if I actually start by hitting you up with some of my questions, as you've [00:03:22] sort of been in this new role for a couple of months? Really interested to hear kind of how things are going.

[00:03:26] Bobby: Yeah, go for it.

[00:03:28] Charlie: Cool. [00:03:29] So out of the gate, obviously a big pivot for you, coming from the more content world and then stepping full force into this AI leadership role. How's it feeling with such a big shift?

[00:03:39] Bobby: It's fascinating. You know, people ask me, like, "How do you feel about working in AI now?" And I often respond, "It is both unsettling and exciting at the same time." You hold both those emotions all the time, right? Because there's so much happening in terms of the trajectory and the speed, but then also it's exciting, right?

[00:03:56] There are so many things. The tools are now so advanced that there are so many things you [00:04:00] can do that the challenge now is prioritization and implementation. But I'm getting ahead of myself.

[00:04:05] Charlie: Yeah, we'll dig into all of that. So I know you've been visiting with a bunch of distributors. I've sat in on a few of your sessions, hearing how promotional goods distributors are using artificial intelligence in their workflows today. What's your view of the state of AI in our industry?

[00:04:22] Bobby: It's wild to see how a distributor that might be, let's say, a two-to-five-person business is doing some pretty sophisticated things with AI, as well as an enterprise. So just because you're maybe a larger business doesn't necessarily mean you have a market advantage when it comes to utilizing AI tools. For example, we have seen distributors in that first category I mentioned — maybe two to five employees, six employees — using something like OpenClaw to try and [00:05:00] create an account manager. I'm gonna mention the edge cases first, and then we'll get to what I think is this sort of middle section of the business in the industry. There's a lot of experimentation. So you have distributors doing things like that, that are building solutions for clients, that are utilizing it for ops. But you almost have this situation where the edge cases are extremely edge, and what you're seeing them do is quite eye-opening. Then you have this middle experience where we're all still trying to wrestle with the tools to figure out which business processes we want it to tackle. I think the industry has migrated beyond ChatGPT and Claude and — insert your frontier model here. We've moved beyond it just correcting our copy. We're all now trying to figure out business processes, so that's cool to see that migration for the industry. There's this massive gap between folks that are maybe using agents and folks still kind of stuck in what I call prompting hell, with [00:06:00] iterations over and over of the same kind of experience and routine. But it's a massive gap. I think that gap is gonna close as more people are familiar with the tools, but right now the gap is pretty big. The other interesting part, Charlie — this will not surprise you as someone who has spent a career helping businesses implement — two challenges that I'm seeing in the industry:

[00:06:18] One is prioritization of which business problem you are trying to solve with AI. That's huge, right? That's been your life. That's all you've done, is try to figure out what business problems to solve with technology. The second one is the time and speed of implementation. We're seeing this with even just the frontier models.

[00:06:36] They're investing billions in these organizations that now come in and help enterprise adapt to AI. Because the power of the tools is there; just the functionality and deployment is the hard part. Not just change management — I think most folks have adapted to change. It's now just the implementation of the inner workings of the business. Would you agree with that analysis?

[00:06:59] Charlie: Yeah, [00:07:00] that mirrors both a lot of what I saw in the customer panels that I sat in on, and also a lot of what I'm seeing internally. It takes a lot of experimentation. It takes a lot of trying different tools to figure out what might work or not work for different circumstances.

[00:07:12] Bobby: Mm-hmm.

[00:07:12] Charlie: Plus that constant lens as the human reviewer — that role we all find ourselves falling into — checking the outputs of these things at every step along the way to make sure that they're aligned with the goals of the users in the business.

[00:07:22] Bobby: Yeah. Yep.

[00:07:23] Charlie: Speaking of what I've been seeing internally: on our development team, we've had very significant AI adoption. Our code is at this point basically all either AI-generated or AI-reviewed. There are agents involved in our whole coding journey. Just this morning I actually saw Dileshni, my partner, our VP of Development, having to re-up our credit limit for the month so that the devs could get back in there and keep working once we had burned through all of our AI usage for the month.

[00:07:51] So I know the latest Anthropic data said that coding is sort of the tip of the spear, the leading edge of what AI is being adopted for, but it only makes up about a third of usage now, [00:08:00] and we're increasingly seeing broader business applications. And I know a little bit of what that looks like in my world, but I was wondering what you're seeing in terms of promo world adoption and customer use cases that are effective for our users, where they're adopting AI tools and seeing success.

[00:08:14] Bobby: Yeah. So anecdotally, I can share what folks are doing. Seventy percent of the business case — 70, as you said — of utilization of AI right now is not coding; it is actually solving problems in the business. One of my favorite stories as I've been talking with distributors one-on-one and in our roundtables: I was talking with a distributor who's probably around the age of 60. In a very sophisticated business, by the way. This is somebody who's built a successful business, a profitable business, great team — somebody I admire. And he said, "I was months away from selling my business." And he said, "Because as the CEO, I was wearing an ops hat, I was wearing an HR hat, and I was wearing a sales and leadership hat" — like four different roles. And he said, [00:09:00] "Because the implementation of AI focused on the right business problems, I've now been able to offload some of the admin of those roles that I do very poorly," 'cause he said, "I feel like I'm just doing all four roles poorly." And he said he's been able to offload some of the admin and point AI at those. And he said, "I'm no longer thinking about selling my business." I mean, it's just one story, but it's an interesting experience that's talking about the scale and scope of what the technology can do for us now. Anecdotally, again, what I'm seeing: folks build interesting landing pages for customers using vibe coding with Lovable or something like it. I remember one of my favorite stories is someone who's building landing pages for customers after they make a presentation. So it might be a big project that they're trying to land, or a big prospect. And on the landing page itself, they have everything they talked about in the meeting, and then they have the commonsku links to presentations and shops that they have as well.

[00:09:50] So it's cool that they're building these bespoke landing pages for customers using AI, with all the details for the customer there. So that was pretty cool. And of course, [00:10:00] on the business ops side, the use cases are myriad. From folks that are using tools to manage the hell of their inbox better — you and I saw some of those examples — to folks... I think it's fun to watch younger and, dare I say, smaller distributors experimenting with tools at the same level of sophistication that some of our enterprise customers are doing.

[00:10:20] That's kind of fun to see.

[00:10:21] Charlie: Yeah, that was one of the things that really struck me in one of our recent roundtables: the diversity of tools that the customers are trying, and all the different use cases that they're using them for.

[00:10:30] Bobby: Right.

[00:10:30] Charlie: One that came up there that I found really insightful, and that I've been adopting myself, is not only using AI to record and transcribe your calls for you, but then also to do some of that post-talk analysis on what your follow-up actions should be, how you could potentially have handled that conversation better or differently.

[00:10:45] Bobby: Yeah.

[00:10:46] Charlie: Just really getting into the guts of not only the mechanics, but also helping you be that second voice, that second thought, to help prioritize your own tasks.

[00:10:55] Bobby: And Charlie, I can say, the three of us — me, you — and I say three because Ritz is our production [00:11:00] manager on the podcast, and she's back here in the background. Hi, Ritz.

[00:11:03] Charlie: Hi Ritz.

[00:11:04] Bobby: And we've all been deep in it, I would say, for a year or two, right? Maybe three, maybe longer. We've been pointing it at business problems to solve for quite a while. And there are a couple of interesting things happening in the maturation of adoption with AI. You can get into this cycle of iteration and sort of flounder in this constant iteration — that's one of the danger zones you can get into. Another is — and this is probably the hardest problem for businesses to solve, particularly in this industry where things are moving so fast — adapting to the time required to learn the tools. That's why implementation is now this massive industry that's growing in Silicon Valley: because the time to adapt, or the need and the speed to adapt, is pretty big. But those are some of the quagmires with AI. It shouldn't slow down our adoption, though. I'm getting ahead of myself, but I just got into what can slow things down.

[00:11:53] But I'm seeing that too — that we're now seeing the deployment of AI in a sense that you've got to be able to focus [00:12:00] on a business problem, solve that problem, and move on to the next one.

[00:12:02] Charlie: Speaking of the quagmire, the proliferation of tools: as the editor of our AI Promo Brief, where you're keeping a finger on the pulse of all the new things coming out all the time, what macro trends are you seeing in the AI space recently that you think will have a major impact on our industry and our customers?

[00:12:19] Bobby: Two big ones keep coming up over the past three months. One is that the design tools are getting more and more sophisticated. If you tried to build a presentation with Claude or any of the frontier models in the past — I don't know, three months ago, six months ago — it was disastrous, right? It wasn't that sophisticated. I would challenge you to try it again, because the design tools that are coming along are phenomenal. There's a lot still to go, but I think that's one big trend, and it's not just us saying that as an opinion. What you're seeing with Figma and Canva — you're seeing this race toward providing AI tools that are sort of compressing the creative process a little bit. They're sort of [00:13:00] collapsing the steps involved in that, and you're seeing decent output. I wouldn't say great output in some cases, but decent output. I mean, we're a visual medium, so the presentation side's always gonna be a big part of what we do as an industry. That's one big thing: the increase in design tools that have come along, and we'll talk about ours in just a minute with the AI mock-up tool. The other thing I've already mentioned is just implementation. I've even talked to people who are very sophisticated with AI and who say, "You know what? I'm still gonna hire folks to help me figure out how to implement things," because there's such a vast need to adopt and adapt. Implementation's probably the biggest trend that we're seeing as well. There are some folks in the industry that are doing a little bit of this, and so you're starting to see that even in our industry too. I'm not the only one that studies these trends. What are you seeing?

[00:13:45] Charlie: Few big trends I'm seeing right now. The first is model proliferation and targeting specific models at specific tasks. And this is something I'm working on both with my team internally and as we build more things into the platform — that different models tend to excel at different [00:14:00] things.

[00:14:00] Bobby: Yeah.

[00:14:00] Charlie: And sometimes by chaining them together in unique ways, you can pull off things that are either just better outputs, or that you can do much more affordably than you can if you just throw the frontier model at everything. Which has been a big learning for us as, like I said, we adopt AI very broadly, especially in the research and development teams — the costs have the potential to really run away from us. Making sure that we're keeping an eye on it and being strategic about when we deploy a Fable or a Sol or a sort of leading frontier-class model, versus when we may be able to delegate some of those lower-level tasks, once some of the strategic work is done, to more affordable or more specialized workers.

[00:14:37] The other big one I'm seeing on the design side is just the adoption of Claude across all of my teams for prototyping and for trying out new concepts. If one of my designers is working on something in Figma, they can now very quickly pop over into Claude and throw together a working version of it, and then we can put that in front of customers and see how they react to something that [00:15:00] looks and feels a lot like how the final version might look and feel.

[00:15:03] Which means that when we go to actually build those features, we have much more validation on not just what it looks like, but how it'll work and how a user will actually interact with it. So we can preempt a lot more of the feedback we would usually get when it's in limited release or full release, and do that much earlier in our design cycle.

[00:15:17] So I think that should help us really accelerate a lot of the things we're bringing to market.

[00:15:20] Bobby: Yeah, that's cool. I wanna also mention, for those that haven't done this yet: one of the big unlocks for us as an organization is our Director of AI Operations, Eku, leads a weekly with our team to just share what folks are doing with AI, and it's just a simple unlock for you and your team.

[00:15:38] I hope this idea helps you as you're listening — to just carve that time out in your week, or the following week, to let folks share what they're doing. Sometimes you'll have incredible story after story, and sometimes you may not. It's just that you're doing it every week to give space to that, because in the promo industry, everything's moving so fast, the demands are so high, teams are so busy, that if you don't carve [00:16:00] that space to do that, then you're probably not going to accelerate that adoption.

[00:16:04] But it's been a big win for us.

[00:16:05] Charlie: Yeah, absolutely. Speaking of the team — as you've been collaborating a lot more tightly with us over in the research and development world, Bobby... You know, we used to joke about how long it had been since you logged into the platform, and now I'd say times have very definitively changed.

[00:16:20] Having moved into this new seat, now that you're coming to our technical reviews, our design reviews, trying to help be that voice of the customer directly with our developers and our designers: how did you picture the platform side of the business before you stepped into it? And how different has the reality been now that you see how the sausage is made?

[00:16:36] Bobby: I don't know if it's a stark difference. What I will say is that I am impressed at the depth of our team, and just the engineering talent that we have, the product managers, and the team you're leading, Charlie — and also the breadth of what our team does. So when I say that, I mean there are some things in the tech world that we tend to take for granted that we take very seriously: security, privacy, all of these things that are critical to [00:17:00] our customers. Honestly, they're under the hood and you don't think about those. But they're vitally important, and we have folks that are huge safeguards around that. The other thing is just the innovation and the depth of engineering talent that we have now. I mean, Charlie, I don't know the number offhand, but we've had quite a spike in talent hiring over the past year, which also follows a trend that's going on with AI. We're seeing this trend — correct me if you think I'm wrong on this — where we thought AI would be replacing developers. What's actually happening is AI's accelerating development in many cases, because we now have a new tool set. And when you have a new tool set — you can attest to this more than me — you never run out of a list of problems to solve when it comes to building something for your customers. We now have a new tool set, and you have more problems to solve, or you have new opportunities to create with that tool set. So the fact that we've grown — this is just my from-the-sidelines-observing kind of comment — mirrors kind of what's going on in the tech world to begin with: we're able to take a [00:18:00] lot more attempts at problems for our customers.

[00:18:03] Charlie: Yeah, that's been the most fun part of the whole thing for me: we can dream so much bigger —

[00:18:07] Bobby: Yeah.

[00:18:08] Charlie: — than we could before, and really work towards executing some of those things. And to your point, not only have we been scaling the development team rapidly over the course of the last 12, 18 months, but we've also been hiring into much more specialized roles than we ever could before.

[00:18:21] We have people like Prince now on the AI agents and intelligence team, who is an intelligence engineer, right? He is solely focused on the actual intelligence layer and building out the systems for just that kind of work. Whereas in earlier days, we needed more generalists who could kind of do anything that commonsku might throw at them.

[00:18:40] So that's been really exciting. Also, as I've been stepping back in to work with that team a lot more directly the past few months, just getting my hands dirty again has been a lot of fun.

[00:18:48] Bobby: Yeah.

[00:18:48] Charlie: So being able to see firsthand what product means with AI now, and what all these amazing new minds — supercharged by the AI tooling that they all have behind them — can do, has been amazingly [00:19:00] exciting.

[00:19:00] Bobby: Yeah, and behind the scenes, it's funny to see the design team at work, 'cause there's a whole team on design, enterprise, intelligence, as you mentioned — now these specialties that we have in groups. I mean, there are folks like Adam and Daveed and Cassia, who's just phenomenal. If I start naming names, I'm gonna get in trouble, but it's cool to interact with folks and see how they think.

[00:19:18] A great example is just watching the infrastructure being built. When you're driving a fast car, if you're not an engine person, you can take for granted what's going on underneath the hood, and I think that's something I've been able to see and respect.

[00:19:29] Charlie: It's all been really exciting. So of those insights that you've been out getting from our customer base, soaking up all of the context in the industry — what's something you've heard from the customers that you feel has been really valuable to bring to us on the development team, so that we can incorporate it into the next versions of commonsku?

[00:19:46] Bobby: You have been very patient to help me learn how to grow in this role, Charlie. But four things — and this is something you've done professionally for a long time, taking what business problems customers are trying to solve [00:20:00] and bringing that back to the team. I can put it in a framework of four things, and this may be an oversimplification, but here's what I think we're getting to. We're hearing from customers: we want you to build it for us — so presentations and design, size, color, mock-up, descriptions (the descriptor we'll talk about in just a minute), auto-populating things based on using intelligence tools. Build for me, remember for me, watch for me, and price for me.

[00:20:21] So that's the build for me. Remember for me: customers want the platform to know their history, their clients, their preferences. And this isn't just customers — this is also us. We're seeing things and trends that we want to do for the customers and also asking them, "What do you think of this?" Watch for me is a production tracking window that we're looking at. Customers want assistance, agents, whatever you wanna call it. And then price for me. There are a couple of categories, Charlie, you and I obsess over. One of mine is the production side, as well as the price for me. When you're from promo, you're for promo, by promo — and for those that may be new listeners, what you may not know is that Mark and Catherine, who founded commonsku, they [00:21:00] were distributors.

[00:21:01] I was a distributor. You bring these things in with you to the platform that are very nuanced within our industry, and one of the nuances within our industry is pricing. An example of that: let's say it's Fairware, who's focused on sustainability and transparency. They might have a different pricing structure set up that affects the way they go to market. Same thing with a distributor who focuses on collegiate licensing. There's a whole layer of collegiate licensing markup and margin that they have to consider in addition to the product itself. So there are pricing configurations. One of the biggest takeaways that I have is that, among many of our customers, there are similarities, but there are as many differences in terms of their business model and who they serve. So if we can figure out how to create this brain around the soft data of what they know about their clients and their verticals and their industries, along with the hard data that we already have such a great handle on — in terms of PromoStandards and tracking and shipping and all that — we're looking at a really cool future where we get to spend less [00:22:00] time in the industry (I'm speaking as a distributor) on the ops side of things, and more time on strategy, client experience, and all of those things.

[00:22:06] It's a really cool frontier, and I think we're looking at one of the brightest futures we've had in the industry in a long time.

[00:22:11] Charlie: Yeah, and I know we'll get deeper into that when we talk roadmap in a few minutes, in what might become your longest skucast ever. But I do find it really interesting that you bring up context there, because in our recent company-wide hackathon, I think we both led teams, and both of our projects were centered around how we can provide agents with meaningful context to do their jobs well.

[00:22:29] Bobby: Yeah.

[00:22:30] Charlie: Yours very much centered on that client brain concept; mine much more on the commonsku secret sauce — how can we build a lot of our kind of knowledge and industry expertise into agent instructions to help them do better. So I'm really excited to help build both those things out and turn them into something real for the customers.

[00:22:46] Bobby: For sure.

[00:22:46] Charlie: I'll ask you one more thing before I hand you the microphone and let you throw the questions back at me. So, you've spent how long in this role before you made the jump to the AI thing?

[00:22:55] Bobby: Several — let's put it that way.

[00:22:56] Charlie: I won't pry. A big thing I loved [00:23:00] seeing from you specifically, even before you made this jump, was starting to use AI tools to automate big pieces of what your previous day-to-day had been.

[00:23:08] What does it feel like to disrupt your own work and be able to kind of totally pivot from the thing you were doing into something entirely new, having functionally replaced large parts of what you did before?

[00:23:17] Bobby: Yeah. As I mentioned at the beginning, it's both unsettling and exciting — and I think more exciting than unsettling, but it happens every now and then. You know, in content marketing, we saw the wave coming before it hit, and so you had that experience of, "Oh, no, is AI going to replace me?" And then you went through the experience of, "Oh, no, this is a huge help to us."

[00:23:34] So just even on the podcast: you can take an eight-hour production window on a podcast and collapse that down to two to three hours, thanks to the AI tools that are baked into podcasting. One of the most fun ones is we have two very popular newsletters: the AI Promo Brief, that looks at AI trends through the lens of promo, and then the Backpack, that Ritz, our production manager, manages, oversees, and creates — that one's on trends in merch that are happening. And we've been [00:24:00] able to now take agents and build those for us. It would take me maybe six to eight hours to build an issue, because you're going in hand by hand, researching what relevant articles are important for our audience, selecting those. Well, now with an agent, you can train it to know what's important, to select the right tools, and basically shoulder 80 to 90 percent of the work. So for the AI Promo Brief, for example, I have four agents that work on my behalf for each issue. One is a researcher, one is a selector, one is an editor, and one is a writer. Now, humans take the last 10 to 20 percent and finish it, because one of the most important things for me is I'm always looking at this through the lens of some of our most creative distributorships, and always asking the question: will this be interesting to Winston at Creative Boulevard?

[00:24:45] Will this be interesting to John Vo? That's my lens when we get it to that last mile that we take. But to be able to marshal agents on your behalf for a research project that took that amount of time is pretty phenomenal. And Charlie, just for the skucast: just last [00:25:00] week, I pointed Claude at a project that I've been thinking about for a couple years. We've had over 350 episodes of the skucast. We have had 20 to 30 events that we've done, countless webinars. When you think about that, I have data sets of questions going back several years, and probably upwards of four to five hundred interviews. And a lot of them were along the same themes: growth, growing a team, sales, leadership. And so finally, I was able to point Claude at my database and say, "You know what? I have another interview coming up on the topic of growth. I know I've created more questions on this before." And Claude was able to create an entire catalog of the questions I've asked, and compiled over 3,000 questions.

[00:25:38] This took minutes. And it compiled over 3,000 questions, so that the next time I have an interview on a particular topic, I can go pull from that. So those kinds of heavy research topics — we've been able to see the collapse of that in content. This gets into the argument that AI is freeing you up to be even more creative. There's this sort of debate going on where it's almost pitting AI against humans; I see the opposite. [00:26:00] I see that AI is actually creating an opportunity for you to be even more ingenious with the product that you end up creating.

[00:26:07] Charlie: Yeah, and I've been seeing a lot of the same things on my teams as we start to bring on agents to take over some parts of the role for product managers. I know we'll get into what that means in a few minutes, but one of my favorite agents that we have running is a new agent that processes every customer idea that commonsku customers submit through the Got an Idea form.

[00:26:26] We get dozens of these a day. This was previously something we solved with a very large meeting, where we had all the product people and all the customer success managers and designers and data people all sitting in a room, and we would look at every ticket. Each one would get maybe five seconds of all of us looking at it, going, "Yeah, that seems like a cool idea."

[00:26:44] And then it would kind of be passed off to the PM to consider amongst all the other potential things. The new agent that we have doing that pre-pass is amazing, because it digs into the ticket, it digs into the customer who's asking for it, and compares against the analytics of what they've done on the platform.

[00:26:59] It'll assign it to [00:27:00] my PM and then also group it with all of the tickets around similar themes, so that we can see, "Oh, this is the 12th thing we've got around this filter on the repeat order report — maybe that should get a little more attention and float a little higher up the list." Or, "Maybe this is part of this broader theme around email automation that Dieter, our automation workflow PM, should consider as he gets to that broader feature set."

[00:27:24] That's one where basically the core human role is still there. It's still up to my product managers to really deeply consider those ideas, think about how valuable each of them is for our customers against how much effort that thing will take to build, go talk to the people who have requested it, and really get to the core of what the issue is.

[00:27:40] But that very first manual touchpoint, which had previously consumed a lot of my team's time given the volume of amazing feedback we get from our customers, has been completely automated away — which leaves time for my team to go much deeper on the actual heavy-lifting work of doing the deep consideration of what the ideas actually mean.

[00:27:56] Bobby: Yeah. And Charlie, you're being too humble. You authored that whole experience. I remember seeing [00:28:00] it — I keep thinking, by the way, and telling you this live on the mic, I need to get a video capture of that, because it's such a cool process. I think, at the scale at which commonsku has grown — there are thousands of customers now — the feedback is significant in terms of things that we can do, opportunities.

[00:28:13] And so to see you marshal that has been pretty cool. So speaking of your role, Charlie, let's talk about that for a minute. For folks listening that may have just tuned in who aren't engineers, they're not in software, they're not in SaaS: explain your job. What does a VP of Product do?

[00:28:25] Charlie: Yeah. So the VP of Product Management — and all product management — our role in an organization is to determine what should be built. We have our amazing development team who do the actual building, and our team's role is to make sure that we are going out and listening to the customer base.

[00:28:40] We're considering the needs of our customers and commonsku as an organization, and we are putting the highest-priority-possible things in front of our developers to make sure that we're always building the most important things for the company and our customers at any given time. Day-to-day for my team, that means lots of getting out into the field and talking to our customers, and seeing what distributors and suppliers and [00:29:00] decorators are actually dealing with on a day-to-day basis.

[00:29:02] Lots of data work in the analytics around what's happening on the platform, how people are using it, where things are working, where things are not working, where maybe there's some friction. And then also bringing in those signals from things like the Got an Idea form to add another layer of insight in terms of how we prioritize those things that should be built.

[00:29:21] And then we work very closely with our designers. So Adam runs our user experience design team, with currently Musa and Daveed — and a new role opening up soon, if anyone listening happens to be in the market for a new user experience design role — to come up with the flow and function of those features.

[00:29:36] So to look at how they work in context, what the implementation might look like, and what some of the different considerations of some different ways a flow could work might be. And then we work really closely with our developers as they build those things — to test them, to ship them to our customers, and then to measure the impact of those things that we've built and delivered, so that we can learn from them and plan the next thing.

[00:29:56] So I call it the best job in the world. I know I'm a [00:30:00] little bit biased there, but to me, it's this amazing role that perfectly balances being out there with customers, solving real business problems, while getting to sink your teeth into some deep, meaty technical issues.

[00:30:10] Bobby: What I'm laughing at is, if you're tuning in and you've never heard Charlie before: yes, Charlie is this optimistic all the time. Like, all the time. He has no rollercoaster like I do living inside me. Okay. Well, let's talk a little bit about AI adoption in general, because you and I grapple with a lot of customer feedback and adoption.

[00:30:28] And one of the interesting parts is trying to map AI solutions toward a network that has all kinds of experiences with AI. So we talked about the adoption curve with AI inside the platform. As we build AI, you and I have talked about the analogy of a co-pilot stage — where humans are completely in control — to eventually a pilot stage.

[00:30:45] Is that analogy correct? Can you share more about the vision for what we think of in terms of — and if you don't mind covering this part too — folks may not know what this means, but I think most will: talk about the whole headless option versus building [00:31:00] things in the platform, if you don't mind.

[00:31:01] Charlie: Yeah. So I sort of see the evolution of AI features in commonsku itself as kind of a ladder that we're all climbing together — from where we were three years ago, where there was a little bit of machine learning in things like our recommendation algorithms, but pretty basic from an AI perspective, to where we're going, which is really a fully agentic future, where there are lots of amazing independent agents working on projects for distributors, delivering a whole bunch of value to them.

[00:31:25] The steps along the way there: so the first things we built in the platform go back to 2025, which feels like forever ago now — a series of dedicated AI tools. These are things that use large language models, diffusion models for image generation, and basically replace one step of the core workflow.

[00:31:45] The ones that stand out to me there are our Description Rewriter, which we've shipped, which takes the description from a supplier — which tends to be more targeted at a distributor audience and at a search engine audience than at someone who would actually be receiving a presentation and deciding on which [00:32:00] promo products they may wanna use for their own business — and rewrites it into something that's targeted at the specific customer, and more broadly, just at the human consumer who's not an expert in promo.

[00:32:09] We've also released our image mock-up tool, which takes details about — currently — a Connected Plus product and all of the details we get from the Connected Plus suppliers on how those products work, what the available decoration methods are, how those decoration methods work, plus a customer logo, and can create a very nice, accurate mock-up of what that thing looks like in the wild. And the third piece, which is related to that mock-up tool, still in this sort of tool space, is an art configuration agent, which looks at that product, looks at the instructions the user has given about the project and how things might be decorated in it, and starts to do some of that heavy manual configuration work for the user.

[00:32:52] It's going in and setting decoration locations and methods. It's doing the logo descriptions for you. It's providing information both for the [00:33:00] mock-up tool but also for the potential future decorator around, you know, is this single-color embroidery or multicolor embroidery, and how should it work?

[00:33:07] And those are the pieces that really start to excite me, because of the around 38 million clicks that happen on commonsku every month now, about half of them are in that product configuration world across presentations, estimates, sales orders, and shops. So we just have this huge amount of opportunity to go in there and build these tools that help our users do their jobs.

[00:33:27] So that's the sort of phase one, where we build discrete tools that are currently live in open beta on the platform for our Advanced and Enterprise tier users. And we've been seeing awesome adoption across the board. Over 10,000 uses each now, each of which is saving our customers minutes of time.

[00:33:44] So just the impact of the first tools pass has been pretty incredible.

[00:33:47] Bobby: We get this question a lot from folks that have really dug in with AI: "Hey, which direction are you guys going? Are you building a headless solution, like what Salesforce did, or are you building in-app? What's your take on that?" And I'll be the one to say it for you:

[00:33:59] The roadmap is [00:34:00] always subject to change. Is that right? We're wrestling with these things as we speak, but I like asking them live.

[00:34:05] Charlie: Yeah. So as we move away from these little discrete tools — and we have more of those to ship too; we need to do price configuration, we need to do customer management, there's a handful more on that side — the next step is to start building agents that can chain these things together and do meaningful work for users in commonsku.

[00:34:21] And the question we always get on that front is: is this a tool that you're going to be building into the platform, or are you going to be building the connectors to tools like Claude Cowork and GPT Work and Cursor, so that our customers can build their own agents to interact with commonsku?

[00:34:38] And my answer is yes. This is a path where both are extremely viable. To me, they have their own somewhat separate use cases where we think each can shine. So on the internal front, the amazing agent that we're building into commonsku itself is our good buddy and mascot, skubot, who will be awakening in the platform shortly and starting to do meaningful work for our customers.

[00:34:59] So in [00:35:00] Q3, our internal agent will be handling kind of two fronts. We'll be rolling out an opportunity agent flavor of skubot, which will be able to dig into the reports and help commonsku admins and reps find new opportunities in the business. We're thinking here really of that farming side — looking at the book of business, looking at previous orders, trying to figure out who might be due for a reorder or some cross-selling.

[00:35:21] And to your point, Bobby, there's been a lot of very hardcore work on things like the security front to make sure that a rep can only access the same data through skubot that they can access themselves in the platform. So a lot of the nice pieces of building our own agents internally, directly within the app itself, is we can build in those controls for our customers, along with, like we talked about, providing it with really meaningful context about what things mean in commonsku, how that specific distributor tends to operate, what sorts of products might resonate with that specific client based on their own history.

[00:35:51] The other big exciting skubot functionality coming in Q3 that's under active development right now is presentation building. That's where it will be able [00:36:00] to look at the details of a project and start to help build out the actual structure of that presentation for those customers.

[00:36:08] So pulling products, reasoning about them to think which ones may or may not work, and then critically — this is where we get into that co-pilot phase — working with the human expert rep who's already in the platform doing this day-to-day, to help them put together an awesome presentation faster and better than they could by themselves.

[00:36:24] Bobby: Yeah.

[00:36:24] Charlie: You know, thinking through things from different angles, making sure that the lead times are appropriate — kind of all the core promo knowledge that you're always on here preaching all the time — making sure that our agents are tied in with that.

[00:36:36] Bobby: Yeah, it's gorgeous, by the way. We saw this presentation this last week, of where it's at currently, and one of the things that I loved is it's built in to where you can either let the agent do things on your behalf, or you can intervene, and I thought that was a cool way for a distributor to get the best of both worlds.

[00:36:52] Charlie: Yeah, being able to sort of pop back and forth between classic commonsku and new skubot-driven commonsku, I think, will be a really great experience. It'll definitely be [00:37:00] a learning curve —

[00:37:00] Bobby: Yeah.

[00:37:01] Charlie: — for a lot of the customers. It's definitely a learning curve for us, even as we're trying to build it.

[00:37:05] Bobby: Yeah.

[00:37:05] Charlie: But we think the potential value here of that co-pilot model is extreme.

[00:37:10] Especially for, let's say, more junior reps joining a team, to help them through some of those paces. But even for super senior people to move more quickly — to be able to spend less time on the configuration of the actual products and just let them focus more on the true value of what they're offering to their customers: that white-glove service that our distributors love to provide to the customer base, and really making sure that the projects are as successful as they can possibly be.

[00:37:35] Bobby: I love, by the way, that skubot was benched as a player a few years ago and buried behind all of our other branding, and now he's brought back as an A-team player on the first string. So it's cool to see skubot making a rebound.

[00:37:47] Charlie: Storming back in. He has this adorable little winking, thinking animation thing — and I don't wanna hold up this feature a day longer than we need to because of how valuable it is, but we are doing the thinking eyes as part of the release.

[00:37:58] Bobby: I love it. I just have a [00:38:00] question about why you think things like the Description Rewriter were such a hit. Suddenly it was a big hit, and I got one review from the Description Rewriter. And Charlie covered it, but in case we need to cover it again, because we covered a lot there: the Description Rewriter is this tool that allows you to take data — [00:38:16] let's say nylon data that's on a bag, that's really important to specs for what the bag is made of — but customers don't care. They want it to be cool. They want it to affect the impact they're making in their branding. And so distributors are having to go in and hand-rewrite these descriptions from supplier data that's built into ESP and SAGE and things like that.

[00:38:32] So when it populates in commonsku, it's from a manufacturer's POV. And so with a click — like, not multiple clicks, with a click — it just turns that description right around. And by the way, the writer in me loves this, because we always like to hate on AI and writing. It's brilliant because it's fast.

[00:38:46] My theory is it's so insanely successful because we know promo, and if you can nail speed and simplicity, you're gonna win the hearts of buyers. Why do you think that resonated so much with customers?

[00:38:56] Charlie: Yeah, I think a big piece of it, to your point there, is it's just there, right? It's right [00:39:00] beside the description field. It's a little button. You click it, the description gets better —

[00:39:03] Bobby: Yeah.

[00:39:03] Charlie: — which has a meaningful impact on the presentation or shop or sales order that that's being built into.

[00:39:10] And to me, those are the kind of delight features that I love when we can roll out to the customer base.

[00:39:14] Bobby: Yeah.

[00:39:14] Charlie: We can't swear on the skucast, right? So we can't say exactly what Norma said in response to the thing, but —

[00:39:20] Bobby: But I love it. FYI: Description Rewriter.

[00:39:22] Charlie: Yeah.

[00:39:22] Bobby: Yeah. For Q3, what's on the horizon just beyond AI? You've been really good to remind us, too, that we have multiple tools at our disposal: automation, AI. And for those that don't know, two of our three strategic initiatives this year have been around AI and automation. But what else is on the horizon for Q3?

[00:39:37] Charlie: Yeah. We have a very big docket of things shipping this summer, across all of our teams. So on the AI side, we'll have that skubot presentation builder rolling out in beta with all those context pieces built in. On our workflow team, the very first version of our workflow automation engine is going live next week, where Dieter — my PM whose job it is to just reduce extra [00:40:00] clicks on the platform by all means necessary — is shipping his first version, which is designed to keep projects moving along.

[00:40:06] So it basically monitors the state of all open forms and creates tasks for reps to go follow up. Dieter has a very bold vision of the future there, where agents will also be grabbing some of those tasks as needed and pushing those forward.

[00:40:16] Bobby: Absolutely brilliant, by the way. Just the work —

[00:40:18] Charlie: He's —

[00:40:19] Bobby: — with AI too is just brilliant.

[00:40:20] Charlie: So excited to work with Dieter.

[00:40:21] Like I said, all the amazing new people we've been bringing on have really supercharged what we can deliver. By the way, anyone who's interested in that beta: there's a thread on the community forum where Dieter's looking for people who wanna try out some of these automation features and help inform those future versions of it.

[00:40:34] As we have more of these agents working on our customers' behalf, as we have more of these automations running, we'll be rolling out a new notification system just to let people know how things are happening in real time, which will, again, we think, help pace projects better.

[00:40:47] The average end-to-end commonsku project that goes from presentation all the way through to bills takes about 41 days on average. So we think we have plenty of opportunity, just through building these tools, to help our distributors move [00:41:00] faster — to accelerate their business, accelerate their receivables, all kinds of very useful business flows that we think we'll be able to speed up there.

[00:41:06] Another super exciting one that we'll have going live in the next couple of weeks is our print-on-demand integration with our launch partner, Fulfill Engine, which to me leans into a lot of these same themes around taking work that's very complicated and automating a lot of it away. So if a commonsku user sets up a new print-on-demand shop in commonsku, it's almost set-and-forget: once these products are pre-approved and pre-configured, there's no processing required.

[00:41:33] Users check out in the shops, pay for things on credit cards, and those orders automatically flow into all of the great standard commonsku stuff, where they get grouped into orders. They're all associated with the specific reps and accounts for commissions and billing purposes. And again, it takes what was previously a very big, heavy manual lift and just automates away all of that heavy manual work.

[00:41:53] And a lot of those learnings that we've been picking up as we work with Jason and the amazing team at Fulfill Engine — we're starting to think, "How can we [00:42:00] incorporate some of that mindset just into the broader promo workflows?" Right? Like, how can we make sure we're doing approvals early and often, so that things don't get stuck in proof approval — which can then cause shipping to get messed up on the project and eat your entire margin, kind of thing.

[00:42:14] Bobby: Yeah.

[00:42:14] Charlie: So lots of super exciting stuff happening on that front. Slightly later in the year, on the intelligence front, once we have the skubot presentation builder out and building amazing presentations in the wild, the goal from there on is to go way deeper down the workflow. So very quickly move on from building presentations to shops, to estimates, to sales orders — including all the nuances involved in all of the pricing and configuration across those forms.

[00:42:36] And then, to your point, Bobby, as we start to look from that co-pilot model towards what a more autonomous agent on commonsku looks like, that's where we get into things like agents monitoring projects to catch problems in advance. So seeing when a purchase order is trending towards shipping late — if it's going to miss the in-hands date — getting ahead of those sorts of things.

[00:42:56] Or, you know, the simpler case: if we get shipping information from a supplier and [00:43:00] everything checks out, potentially being able to send the shipping notification to the customer without any human involvement needed at all. Obviously, all totally configurable by the distributor, with lots of controls over what happens when and how.

[00:43:11] But the goal we've set for ourselves as the product team for the year is we'd like to eliminate close to 80% of the manual effort that happens across the platform. And while we do it, we want to see commonsku projects become more successful and faster. So every stage taking less time, less human effort, for projects with higher margins and fewer problems.

[00:43:29] And we have amazing visibility into all of that data across the platform, so that's what all of our development efforts are targeted at.

[00:43:35] Bobby: Charlie, when I was a distributor, one of the things that I had to do — and our team had to do — is we had to set a threshold for minimum orders. If any order was above, let's say, $5,000 — I think that was our threshold at the time — any order above $5,000, it had this extra TLC layer. And to know that you can point an agent at those kinds of scales of numbers and have it babysit those orders for you is phenomenal. One thing — and I don't know if I have a question here, but an observation I've had as I've been talking with customers, and I'm also working with our [00:44:00] team more: the evolution of the industry has gone from solo salespeople working with clients on a one-on-one basis to a team-based selling model.

[00:44:09] We've been saying this for years, and it is the iteration that the industry has followed. When you're working with a client — let's say they're in the $100,000 to $250,000 category — the $100,000 category, depending on the average sales volume, you can manage on a one-to-one relationship. But once it gets beyond $250,000 in annual sales with one client — half a million, a million — you're talking about team selling. And so commonsku has been built on this integrated workflow model, which I can't emphasize enough as I've been watching our engineers at work and talking with our customers. It's interesting to note that bigger customers have come to commonsku through the years because of that integrated model, because it allows teams to work better together. I'm now seeing AI as a critical layer on top of that. So you have team selling driving a lot of the innovation happening with distributors, combined with our integrated workflow that combines all these. This is a very complex and messy industry, as we all [00:45:00] know. But to have an integrated workflow that is able to do those handoffs automatically is huge. And then to see AI come in and be able to add a new tool set to all that is pretty phenomenal. When people talk with you, or suggest or hint at the SaaS apocalypse, for example, and the state of AI and SaaS and software in general, what are your thoughts?

[00:45:19] Charlie: I know you said this earlier in the call, but "I'm very excited" is the short version. I see a really bright future for software as a general field. And I think for vertical-specific software more than basically anything else.

[00:45:32] Bobby: Hmm.

[00:45:32] Charlie: So when we think about SaaS — and commonsku is a great example of this — every business has always theoretically had the ability to go out and build their own software.

[00:45:41] Any distributor could have gone out and built their own promo order management system with a whole bunch of integrations into a whole bunch of different systems. That is what commonsku was in its early days. To the Mark video that we recently shared on LinkedIn — it was Roman. It was the Rightsleeve order management system.

[00:45:56] What multi-tenant SaaS like commonsku does is it basically lets [00:46:00] a large pool of — typically — competitors, you know, players in an industry, all pool their resources into one organization, and that organization builds the single tool that solves all of their needs, and gets all the combined expertise of those customers, the combined resources of those customers, the combined data of those customers to create an offering that's greater than what any one of them could create themselves.

[00:46:20] And as we push that thousand-distributor mark, and have all of our very broad base of suppliers as well, to me, AI is just another layer on top of it, where we can now build these amazing agents that draw on that experience from across our customer base. We can pool the resources of all these distributors to hire these, frankly, amazingly impressive people that I'm honored to show up and work with every day — whether it's the data scientists or the staff-level engineers we've been able to bring on.

[00:46:46] Even just the new technical leadership we've been able to bring on to run this very big development team that can produce all these amazing things. To me, there's just a level there where we can operate at a scale now that, frankly, we couldn't two years ago. And I think the things we'll be able [00:47:00] to deliver using a SaaS model — that looks a lot like what a SaaS model always looked like — will just be turned up to 11. Like, the absolute best version of those things.

[00:47:08] When people talk about the SaaS apocalypse — it's, can any of these companies go build their own things now? Sure. And you know, I encourage anyone listening to go build their own tools. It's something that's very, very doable with vibe coding now. You can pop into Claude Code or Claude Cowork and build yourself some little apps or some little artifacts, and they're fantastic.

[00:47:25] We use those internally all the time. To me, what it does for us as a professional software development organization is really just level up the goals. We have that whole new layer of competition, and we need to make sure that what commonsku evolves into is way greater than the sum of its parts.

[00:47:41] Bobby: Yep. And I wanna add to that. The industry has this real bent toward new and shiny, and part of my job is to look at what's going on in the landscape and competitive climate, and there are these point solutions that solve a problem. But what that kind of can create — as someone who had a very Frankenstein tech stack as a distributor (we did print and promo, and it was impressive, [00:48:00] but also very difficult to manage) — there are a lot of point solutions that can take your eye off the ball in terms of your priority of running a business. So my encouragement to distributors is to keep those priorities strong and solid. There are going to be point solutions. Remember the integrated workflow. I know it sounds like I'm flying the colors, but it's huge, because we have interesting studies coming up around what we've been able to see teams scale, and their growth within a five-year experience with commonsku.

[00:48:22] It's astounding. I've already had a peek at the numbers. It's absolutely amazing. Anyways, I'm getting off task. We have a few minutes here, Charlie. I'm gonna ask you one more question. What are some of the coolest ways you've used AI just personally?

[00:48:31] Charlie: Personally, I'm planning a honeymoon, so it's fantastic for figuring out the details of which wineries you should visit in Croatia, if anyone's so inclined. Professionally, I've set up a few things that are extremely helpful. So one is my weekly planner, which starts every Monday morning with a review of all the development that happened the previous week, all of the tasks we have in the planning phases.

[00:48:51] It lets me know details about which customer meetings are happening, what we've been hearing from different channels. It literally pulls in all that context and gives me a full [00:49:00] overview of all the data on the platform and what trends we've been seeing in terms of the behavior of our customers and their customers, to potentially look for any alarms — which has been fantastic.

[00:49:09] I've also set up a handful of dedicated agents for things that are a little more specific to my role and my team. So I have one that reviews requirements. When we're working on a new feature — like the Description Rewriter, or like the custom URLs for shops that our shops team is working on right now — making sure that we're going through and thinking through all the security considerations and all of the potential hiccups that could come up, and all of just the bits and pieces to make sure that we've ticked all the boxes and that we're going to put our best foot forward in terms of what that feature can be.

[00:49:38] And then I've also been doing some very cool work recently on data modeling, which I'm not quite ready to share broadly yet. But to your point around the potential future commonsku index — just ways where we can use the aggregate data on the platform to provide amazing insights, both to our customers and potentially also to agents working in the system, around what the common distribution of [00:50:00] sizes is on a hoodie order for a middle school in Texas, kind of thing.

[00:50:03] Just going really deep into all of the accumulated data and knowledge that we have in commonsku to try and really suss out unique applications that otherwise might just not be possible.

[00:50:13] Bobby: Charlie, thanks for joining us. This has been fun. I can't believe we spent an hour on this, but I'm so glad we did. And for those that want more information, be sure you tune into our newsletters. We'll drop some links in the show notes, but we have more about AI coming up in the future.

[00:50:27] Stay tuned, and we'll be back.

[00:50:29] Charlie: Yeah. Thanks for having me, Bobby. It's been fun.