---

About Justin : LinkedIn

FULL TRANSCRIPT BELOW

Shamanth: I’m excited to welcome Justin Gerrard to Intelligent Artifice. Justin, welcome to the show.

Justin: It’s awesome to be here, Shamanth. Thank you for having me.

Shamanth: Yeah, I’m excited to have you. You came very, very highly recommended, and I’m also excited to have you because you’ve been on two very, very different sides of the picture that we’re going to talk about.

Shamanth: You’ve been in category-defining companies like Twitch and Discord in-house, and now you are very much running solo advising early stage startups, early to mid-stage startups. And so you’ve worked in a pre-AI world, you know, managing large teams and a post-AI world using AI instead of large teams.

Shamanth: So take me back to some of the work you’ve done with Twitch and Discord.

Justin: Yeah.

Shamanth: Just set some context. Tell us how the teams were structured around you doing the same kid of work that you do now with AI.

Justin: Yeah, absolutely. I guess first I’ll start with Twitch, and it’s interesting because now I’m more on the fractional CMO side, which has been more encompassing, whereas a lot of the roles I was doing in the past were more kind of traditional growth, performance marketing-oriented.

Justin: At Twitch, when I was at the company, I was brought in to build out the performance marketing function. The team had never run paid ads. They wanted to stress test whether this was going to be a meaningful lever for the business. They had always grown through organic channels, word of mouth, through their creator ecosystem, but now they wanted to try out paid ads.

Justin: So initially it was just me setting up the buys with an agency. Then I ultimately built up a team, about four media buyers across paid social, retargeting, programmatic, across mobile and web, all the known channels. We had another agency helping us do creative at scale.

Justin: We had a data science team supporting us with analytics, as well as a team working on fine-tuning our LTV model, and then ultimately a PMM team to support on larger campaigns. And I reported into the head of growth and media, who reported to the CMO. It was a pretty traditional structure in terms of what I’d seen in growth teams.

Justin: At Discord, I was brought in via an acquisition of a startup where I was leading marketing. It’s called Backyard. It was basically embedding live gaming experiences into a video chat. I would say we had pandemic market fit, maybe not product market fit, but pandemic market fit, and it did really well during that time.

Justin: When the company was acquired by Discord, I was brought in to do ASO, run their ASO and optimizations in the App and Play Store, as well as creative ad optimization. Similarly structured, I was reporting into a head of growth. It was a team of four, a mix of media buyers and me on the ASO creative strategy side, reporting through the CMO.

Justin: Differently, we did all of our analytics in-house, and we leaned on an internal creative team to make all of our assets as opposed to an agency. I guess what’s different now, thinking about where I am on the GTM side and being an embedded CMO. It’s hard to directly correlate my experience, right?

Justin: One’s more of a solopreneur setup, one’s a larger team. But what I’m doing now, when I think about measuring campaign performance, when I think about pulling out insights and new creative ideas and hooks and ways to differently position the brand, putting together any type of new content, that used to take weeks of back-and-forth and making sure that all the teams were aligned and bringing in the designers and making tweaks, etc.

Justin: And a lot of that work now, using these AI platforms, can be done in a few hours, as opposed to a few weeks, particularly on the data analysis, research, and fine-tuning new concepts piece. And what that’s unlocked is a tremendous amount of productivity and speed that didn’t exist before, and it seems like that’s the new paradigm that we’re in, whether you’re working for yourself like me or even new teams and setups in existing mature enterprises.

Shamanth: Yeah. And do you notice that mature enterprises have been able to make the shift to this new way of working very easily?

Justin: What I found. So most of the companies that I’m working with now are earlier stage, but I’m obviously tapped into my friends who are at larger companies. I think it is very…

Justin: I don’t think there’s one model that companies are using. I think some companies are like, “Hey, we’re really leaning into this AI thing. Everybody’s going to be building skills, and we’re going to have these markdown files of all the marketing elements. Everyone’s going to be using that.”

Justin: And others are like, “Let’s use it to speed up the process of creating certain assets,” but it hasn’t bled throughout the entire org. So I think it really depends on the leadership and how quickly they’ve wanted to adapt to the tools.

Shamanth: Yeah. And that’s also been my experience from the conversations I’ve had, that in larger organizations, there’s much more variance in terms of how much adoption there has been.

Shamanth: And there’s even been variance in terms of how they define AI adoption as well.

Justin: Right.

Shamanth: But with smaller companies, as you just pointed out, where you’re working pretty much solo, you’re able to have disproportionate impact, which is crazy compared to what was just possible even just a few years ago.

Justin: Absolutely. I would not be able to do the type of work that I’m doing now without access to these tools, and it’s not to say that I’m just able to speed through everything. It’s just that the initial 60% of the research and data analysis elements that would have had to be done before is almost instantaneous now. So all the other elements of the work, you can focus that much more on the strategic work, and I think that is where it’s really impactful.

Shamanth: Yeah. In preparation for this call, when we spoke, you said Claude has become not a tool, but an operating system.

Shamanth: Talk to me about what that means and specifically how that manifests in how your typical Monday morning looks from the moment you open your laptop.

Justin: Yeah. So when I am able to open my laptop, it’s definitely right after dropping the kids off. So the kid craziness always

Shamanth: Mm-hmm

Justin: always happens first. But yeah, it really is an operating system.

Justin: So I work very heavily in Claude Desktop. I am not technical, so you’re not going to see me really in the terminal doing a bunch of fancy stuff yet. Hopefully down the line I’ll get there. And I know in our last conversation we talked about how you’re really Claude-pilled and maybe doing some deeper work.

Justin: But I found that working in those folders that have the deep context on all the different companies that I’m working with is incredibly helpful to keep me organized and focused on where I can make an impact. So when I open up my laptop, I’m immediately going into Claude.

Justin: I am looking at the meetings for the week and deliverables. I have MCP connectors for my note-taking app, as well as Google Drive and the different docs that I use to curate GTMs and all of my notes. I’m feeding that directly into Notion. This gives me a list of to-dos.

Justin: And I’m using Claude to go back and forth, understand my key insights, and stress test that I truly know what the key priorities are to work on there. And in addition to the strategy work,

Justin: I’m also helping with a fair amount of hands-on execution for some of the startups. So that can include putting together a lifecycle drip campaign for a new product release or updating positioning and messaging docs, doing an overview of creative assets. And where I’ve found Claude really helpful is, here’s all the context.

Justin: Help me get to about 60 to 70% on the actual documents themselves, and then I can do the fine-tuning, that last 20 to 30%. And what this has really done is, because ultimately at the end of the day, you get the output from these tools, and then you have to use your wisdom and experience to say, “Will this actually resonate with customers?”

Justin: Like, does this make sense? Maybe the formatting, maybe the value proposition is in there, but it’s not quite right, you know? There’s that little bit of dissonance. So even that being said, the value per hour that I can provide is much higher because I’m starting the day with these tools, it has all the context, helps me get to 60 to 70%, and then I can push it forward.

Justin: And so again, it allows me to do more of the strategic work, but also be way more hands-on and helpful with the execution piece as well if I need to step in there.

Shamanth: That definitely resonates because I’ve had a number of conversations, and one through line I’ve noticed is that a lot of very experienced practitioners don’t let AI do 100% of the work.

Shamanth: In fact, in my view, if you let AI do a hundred percent of the work, that is a red flag. Yeah. Maybe for very simple tasks you could allow that to happen. In fact, when we hire, we just went through a bunch of hiring and a very immediate disqualification was when it was very clear that someone had used AI over ninety to ninety-five percent with very little editing.

Shamanth: Yeah. So I can see that. And I think what’s also underappreciated is that it’s not easy to step back and have that critical pushback layer where you’re using your own inputs and insights on top of the sixty to seventy percent that the AI has provided.

Justin: Yeah.

Shamanth: That’s not easy, so good on you. And just to double-click on some of that, could you pick an example of a GTM plan you built or a project you worked on for any startup that you worked with?

Shamanth: Walk me through the exact plan you built. You don’t have to share specifics if you’re not comfortable. But walk me through the plan. What were the first inputs you fed in? What were some of the early outputs the model gave back? And what was the point when you said, “Okay, this is about as far as the model’s going to take me. This is where I’m going to take over.”

Justin: Yeah, for sure. And I think this is the key piece for anyone who’s doing that strategic layer work, particularly putting together GTM plans, like having a thoughtful process about that. So one of the startups that I’m working with right now is doing a localized relaunch in one of their core markets to try to drive increased user density, and I was basically putting together a GTM plan that combined online and offline elements to help make this successful. And so in terms of working directly with Claude, Claude’s skills are amazing in terms of helping to set up the formatting for these types of plans.

Justin: So within any GTM plan, you have core messaging, positioning, demographics, audience sizing, key KPIs, sequencing, risks, and outcomes that you want to drive. And the way that this data is presented, particularly when I’m working directly with startups where I’m presenting what the plan should look like, the formatting and clarity of message is really important.

Justin: So I think Claude is really good at helping me develop this GTM skill where I’m putting in these inputs and it’s giving me that first initial formatting of, “Hey, here are all the things that we need to think about within the context of building out this plan.”

Justin: And we can get that first draft of the doc out, and this took a lot of time in the past. So it’s nice to have that. Where my brain kind of takes over next is actually in thinking through the tactics that will work, the channels and different approaches on the strategy, based on what I’ve seen in the past.

Justin: So for this particular GTM, I knew an out-of-home component would probably be important and helpful in providing that awareness and consideration layer, along with doing things that are more down-funnel. As well as working with local press and influencers as that on-the-ground piece and establishing a formal campus ambassador program.

Justin: I also knew that before doing the on-the-ground work, we could use a digital layer to do seeding through social and UGC content, where we’re able to put out a bunch of different concepts to do the initial messaging and positioning work before we’re actually on the ground, to make sure that when we are there, it’s landing in the right way.

Justin: And before we step on the ground to do the IRL component, where I brought Claude back in was the research and competitor analysis piece. So what local groups should we be working with? Who might we want to partner with? Pulling the initial list of potential influencers, getting a sense of the competitors that are in-market and who have launched in the past.

Justin: And so this gave me, putting all this together, I put in the initial insights, I got that output, I layered it in with the competitor research data, and then cleaning up my version of those plans so that it all formats and looks nice. And then after that, I put my brain back on again, which is like, okay, these are all the inputs. This is what I inputted. This is the final doc. Does this all still make sense? Adding that keen eye to each of the different components, and it may have also added some things you just want to catch that may not make sense.

Justin: So essentially what it’s done is supercharge the creation of the doc, which again can take a lot of time. That time I can then spend on more strategic work, and on the competitor analysis, to make sure that I’m bringing the most successful tactics to the market. But after that, you’re almost constantly keeping your brain on throughout the process, because you can get so enamored with how quickly these tools can work that ultimately the output has to result in impact for the company.

Justin: Yeah. So those two things have to go hand in hand.

Shamanth: Yeah. That all makes sense, and a couple of things you said really resonated. Number one was that you actually go back and forth. It’s not a linear process. It’s not like Claude hands it over to you and you perfect it, right? You’re still going back and forth.

Shamanth: And the areas where you are intervening the most is where you’re basically saying, “This doesn’t really make sense,” or you said the outdoor advertising really makes a lot of sense. But Claude, with its understanding, just doesn’t know that. But you, with your pattern recognition across real-world companies, do know that.

Shamanth: And so you’re able to intervene and do something that’s better than the average of the internet that Claude is providing.

Justin: 100%. That is actually one of the best ways of crystallizing that thought that I’ve heard. You’re able to do better than the average of the internet, right?

Justin: And that’s ultimately where the arbitrage exists. That’s where the human-in-the-loop piece still exists, which is based on our wisdom, based on not wanting to do the median of what everyone else is doing, but also having an appreciation of what that is to help

Justin: set that initial context. That’s super important.

Shamanth: Yeah. And I would also argue that for a lot of early-stage startups and high-tech startups, you cannot have the median of the internet knowledge impact your strategy because the median of the internet probably doesn’t know about high tech, probably doesn’t know about the leading edge of marketing and growth and everything.

Shamanth: You know, I would argue some of the models have gotten more sophisticated. They are much more knowledgeable than say a year ago, but I don’t think they match the domain knowledge of someone like you that has spent a decade in the trenches. I don’t think we are yet there. Go ahead.

Justin: Yeah. Yeah, you’re kind of always doing the math, like you’re doing some of the pattern matching, but then you’re like, “Well, when I’ve done this in the past with a similar company, this is what I also ran into as a challenge. So based on where this company is and their budgets, this may not work,” right?

Justin: Like

Shamanth: Yeah. Yeah, yeah.

Justin: So you’re constantly pulling in all of that. And I know this isn’t exactly on topic, maybe kind of interesting, which is around how, if you think about even more junior operators doing GTM work, they’re basically just operating off of, “Hey, what am I able to pull from doing some Perplexity research or ChatGPT or whatever?”

Justin: And it’s like, okay, this seems to have worked for this startup. Well, there are a lot of nuances in a lot of those GTM plans that aren’t necessarily talked about in the TechCrunch articles and in the Reddit threads, and they make it seem like everything was unicorns and rainbows when it happened, but you may do that same exact tactic and miss a lot of the actual challenges that occur.

Justin: Yeah. And so that makes me think it’s going to be very curious to see how that’s navigated in the future with people who are over-relying on some of these tools.

Shamanth: Yeah, yeah. And I think you can already see that it’s easy to spot people who are over-relying on these tools.

Shamanth: I think it’s becoming easier to spot, at least for somebody that’s experienced. I think you can catch some of this easily.

Justin: Yeah.

Shamanth: Just to switch gears a bit, I know you’ve worked a lot on performance marketing teams, on creative strategy in particular, and that’s an area that’s been changed quite dramatically by AI.

Shamanth: And I know we were chatting in preparation for this call about how it’s relatively easy to produce 1,000-plus pieces of automated UGC on a monthly basis. Talk to me about how a brand might be able to do that if that’s what they need to do.

Justin: Sure. So I think there are kind of two ways of doing that, right?

Justin: Just to take a step back. The reason why companies are doing this is because everyone has access to software now, so the moat is really within distribution and how quickly you can get content out there, and also whether you can cut through the noise. So I think there are kind of two ways to go about it.

Justin: One, if you are enterprising enough and have some creative and technical background, you can use Claude and pull in some of these image and video generation models to create your own content at scale, pulling in your brand guidelines and documents and doing all that work and then posting out to your own channels. You can do that. What I’ve found is that if you don’t necessarily have that experience, that can be kind of tough to build on your own, although you can do it.

Justin: So what I’ve seen more companies do is work with partners that can essentially help you create content at scale. And one that I’m aware of will allow you to create 1,000 TikTok-style carousel posts a month, so about 250 a week, where it’s ingesting that content, pulling in hooks based on competitors scraped from the Meta Ads Library or other content out there, helping you post all these formats into a queue of approved content, and then you can parse through that, select what you want to go live, and post it across actual accounts.

Justin: It’s pretty fascinating that you have access to that. I think the one piece is, one, you should definitely be testing all these tools, right? I come from a growth background, so does Shamanth. Try everything, see what works for you within reason.

Justin: But you also have to know what it will drive relative to the results that you’re trying to get after. So some of these platforms are great. You’re getting 500 to 1,000 views per post, which if you’re doing 1,000 per month, that’s 500,000 to a million views. That’s awesome. But is it actually driving down-funnel results, right?

Justin: Like, is it driving installs that are leading to registrations as opposed to just views? So I think it’s really important that as you’re using these incredibly productive tools, you’re actually measuring the down-funnel impact.

Justin: That’s where I see some companies miss when they’re turning some of these systems on.

Shamanth: Yeah. And I also notice, and I would be curious to hear your perspective on this, I also notice that a lot of companies try to make these tools autonomous, and in my experience, that’s a trap.

Justin: Huge trap. Huge trap.

Shamanth: Talk to me about what you’re seeing. Yeah.

Justin: I just feel like, again kind of to the earlier point, if you’re turning on these systems, even if they’re as finely tuned on content that’s working, and competitors out there, if you just kind of let it run on its own, the output from what I’ve seen has never quite been at the level that’s actually going to drive consumers to take an action.

Justin: I think a lot of times when I see that volume, the pieces of content that perform well, particularly in terms of UGC, have an emotional resonance that’s so clear in the content and that pulls the potential customer or user in, and a lot of times this AI content can’t quite do that if you’re automating it.

Justin: Now, it can get you seventy to eighty percent there. There’s a bit more on the curation piece that’s required, and I just… Again, I’m not trying to say that it won’t exist, but right now as it stands, the companies that just turn it on and let it go are like, “Hey, I’m creating all this content,” but it’s not driving any results, and I think that’s why.

Justin: It’s because the full automation piece isn’t quite there yet.

Shamanth: Yeah. And in many cases, some of the people I see don’t even know what’s in the content, what’s in the video. And that’s a scary place to be because they have no idea why it’s not working.

Justin: Yep.

Shamanth: And

Justin: And brand safety or anything else, right? Like these, oh,

Shamanth: yes. You know

Justin: Yeah, like you put out content out there. I was working with a client that had some performance media that was working really well, and I was digging into it, I was like, “Hey, this asset is working well, but you know it has about 18 curses in this content.” So is that actually what you want to optimize for? They said, “Oh my God, we didn’t even realize that. It was just kind of working well when we turned this tool on.”

Shamanth: Yeah. Yeah.

Justin: Yeah.

Shamanth: Yeah. Indeed. To switch gears and talk about something else that stuck out to me from our last conversation, you said that in the last year or so when you worked with AI to build GTM plans, AI has not given you a single genuinely net-new idea, and that observation really stuck with me.

Shamanth: Talk to me about what you meant and what specifically your experience has been.

Justin: Yeah. So I think maybe an analogy that could be helpful here is that AI is a really fast librarian, but it’s not an author. Like it has read every piece of text that’s ever existed. It’s able to get to an insight extremely quickly, which is very valuable, but it doesn’t necessarily help you find the next insight that doesn’t yet exist.

Justin: Like, the librarian isn’t writing the next book. And so if you’re using Claude to help you with a GTM plan, and it’s helping you come up with a bunch of ideas based on things that have been done in the past, that’s helpful, but it doesn’t necessarily have the context on specific cultural moments or untapped audience behavior or something that’s emerging based on new trends or a new brand that’s coming to market, for example.

Justin: So I think the undocumented insights are really where human operators still have the leverage, particularly in the creative space. And in terms of where we are right now, and again these models are moving so fast that I can’t have the hubris to say that this won’t change in a year or whenever, what I’m seeing right now is that it’s more like a better search than genuine invention. But I still think it’s incredibly valuable. Just knowing that difference I think is important when you’re actually trying to move the needle on these campaigns.

Shamanth: Yeah. Yeah. Certainly. Right? It’s like a very smart research assistant, very valuable, not quite an inventor yet. Right. Yeah. We’ll talk about this maybe when AGI comes, we’ll

Justin: We’ll have to do a follow-up when that happens and be like, “Hey Justin, now what do you say?”

Shamanth: Let’s talk tomorrow. Yeah, tomorrow. All right, let’s do

Justin: it. Yeah, tomorrow, right? Forty-four hours and we’ll have a different

Shamanth: conversation. Twenty-four hours, yeah, yeah.

Shamanth: Yeah, and you also mentioned that the big differentiator with AI-assisted work isn’t the model, it’s the humans in the loop, which you also mentioned earlier in this conversation.

Shamanth: Can you share an example where that last twenty to thirty percent mattered the most, and where the human in the loop was the big differentiator?

Justin: Yeah. It sounds trite. Everybody makes this whole kind of taste argument, “Hey, taste is kind of the last mile.” I do still kind of agree with it. Even though everybody talks about it ad nauseam, I do kind of agree with that piece. So I was working with an ed-tech company to help them refine their paid ad strategy.

Justin: I helped them source a really excellent performance marketer who came in fractionally to support. And the challenge was that we just didn’t have the creative execution at scale. They had a designer who could put some stuff together, but we weren’t able to get the type of volume that was required to do all the testing and bring in different concepts.

Justin: And so we brought in one of those companies. I won’t name the exact platform, but a platform where you could create tremendous volumes of AI-generated content at a relatively low cost. Interestingly enough, this team also had some humans who would help kind of smooth the edges on that last mile.

Justin: And the ultimate output wasn’t actually bad. The ideas were clean, the hooks made sense, there was clearly thought put into it, and the visuals generally made sense. But the key piece that was missing was the humans in the final execution. So as much as some companies try to say now that consumers can’t necessarily tell when it’s AI, I just found that that’s not quite there yet.

Justin: So the approach and everything about the content, when you’re working with something, that last piece that had the actual human selling the content, and when you’re working with an education company, that actual tie-in with the person using the product, that human component is really important.

Justin: Yeah. And selling products in particular that require a bit more understanding of how they work, where the value proposition isn’t really immediate. It’s not like using an AI tutoring tool where you just do a TikTok and then you’re right into the product and can use it.

Justin: So that was kind of the component. Like the ideas were great, but the fact that we didn’t have, for this particular type of content, humans in there, if you had just let the campaign run, it would not have performed well. And we actually did A/B test it. We said, “Hey, let’s take these concepts with AI creators and take these same concepts with actual human creators and see what works better,” and we actually weren’t getting any conversions with the AI creators.

Justin: So I think that’s the piece, recognizing good ideas, recognizing good formats, etc., but understanding, having the user empathy based on running a bunch of campaigns, to know what is actually going to move them to the next step of their customer journey and conversion.

Shamanth: Yeah. Interesting. Yes, human in the loop is what makes such a huge difference, and I’m also realizing that after being very enamored with going all in on AI for a long time, you know. And I mean, I love AI, I’m just starting to realize how critical the human in the loop is, and also how hard it is to take that step back and use your own judgment and critical insights, because it’s just so easy to let the model run wild. And Justin, we could, yeah. Go ahead please, yeah.

Justin: Oh, I was just going to say, yeah, just maybe as a closing point, like you have to sometimes snap yourself out of it

Shamanth: Yeah

Justin: of how impressive these things are. Because they can make you super productive, but they can also make you lose the human magic.

Shamanth: Yes.

Justin: But it’s hard, and I have to work really hard as these models continue to get better to snap myself out of it and say, “There’s that little itching feeling in the back of my mind.”

Justin: Yeah. Is this right? Okay, it actually doesn’t feel right, and I’m glad I’m still able to catch that. If I’m just entirely open to doing everything, I’m going to lose that, and that’s losing the secret sauce

Shamanth: Yeah.

Justin: that I’ve built up over all this time.

Shamanth: Yeah, yeah. Yeah. I was chatting with somebody that said this is the first time creating something gives you the dopamine rush.

Shamanth: Because traditionally, dopamine rushes are when you do mindless things and you just waste your time, and now you can actually do something useful, something very creative, and there is a dopamine rush. That is a wonderful thing, but that’s also something to be very careful about, and I think that’s what I’m hearing you say and echo as well.

Justin: Absolutely.

Shamanth: Justin, I know we are coming close to time, and we could keep going, but I want to be respectful of your schedule. This is perhaps a good place for us to start to wrap up. But before we do that, can you tell folks how they can find out more about you and everything you do?

Justin: Absolutely. Check out my website. It’s justinrgerrard.com It has some background on how I support startups, and I’m pretty active on LinkedIn. I’m posting a few times a week. So if any of this is interesting, all my musings are on there, and you can also DM me on there as well if you’d love to chat or connect about anything that was discussed on this pod.

Shamanth: Excellent. We will link to all of that in the show notes. And for right now, we will let you carry on with your day. Thank you so much for your time. Thank you for being a guest on Intelligent Artificers.

Justin: Thank you. It was a lot of fun.

---