Our guest today is Lomit Patel, Chief Growth and Marketing Officer at Trying Your Best (TYB). With over 20 years of experience, Lomit has helped category-defining companies scale revenue efficiently across consumer and emerging platforms like Roku and Tynker. At TYB, Lomit focuses on community-led growth, building customer participation loops, and creating systems that compound trust, revenue, and retention over time. Today, Lomit joins us to share how AI can be used to build smarter, leaner growth systems, enabling rapid testing, automation, and creative ideation while keeping human oversight at the center of scaling and long-term success.

---

About Lomit : LinkedIn | Blog
Connect with TYB: Webpage | LinkedIn

ABOUT ROCKETSHIP HQ: Website | LinkedIn | Newsletter | Youtube | Podcast Website

FULL TRANSCRIPT BELOW
Shamanth: I’m excited to welcome Lomit Patel to Intelligent Artifice. Lomit, welcome to the show.

Lomit: Excited to be here, Shamanth.

Shamanth: I’m excited to have you, Lomit, because as I was saying just before we hit record, I’ve been following a lot of your work from the time you wrote the book. But even before that, when you wrote the book, I didn’t really know or understand AI. I just registered this as something cool and interesting, but I didn’t quite comprehend what it was. And I think now when we are talking, I truly comprehend exactly what you were so prescient in foreseeing more than five years ago when you wrote your book.

And for those who are wondering, the book was called Lean AI. Let’s start with that. At that point in time, five years ago, AI was pre-LLM. What problem were you trying to solve when you wrote this book? What inspired the book when AI wasn’t… I think that AI as a phrase may have been somewhat cool, but it certainly wasn’t mainstream. What problem were you trying to solve? How did people react to the idea back then?

Lomit: Yeah, so it’s funny how quickly AI has moved in the last five years. But going back to writing the book, Lean AI, when I was working at a company called IMVU, the biggest challenge in the use case we had there was that user acquisition was a huge lever for us to scale and grow the business. Managing huge budgets with millions of dollars a month, there’s a lot of complexity around different variables and really being able to figure out how do you allocate budgets across different platforms? How do you optimize bids? How do you optimize creatives?

Outside of trying to hire a much bigger team, we wanted to figure out: is there a leaner way to get smarter around leveraging the data and trying to automate some of the workflows and tasks and processes to manage that campaign? That was the initial seed for looking into AI for us.

I would say the benefit of working at IMVU—people may not remember, but The Lean Startup, which was a book that Eric Ries wrote, he was a co-founder at IMVU as well. And so the philosophy of the company was all around test and iterate. And so the question with Lean AI was: How can we do that at scale so that we can test, learn, and iterate at scale on all of our paid user acquisition? So that we could be able to surface and make decisions a lot quicker. That was the main use case.

Where AI really came into it was that we really used it for Machine Learning. Because we had a lot of data, and it was really about taking that data and trying to optimize who were the right audiences. Because in those days it was all around lookalike audiences. So the question was trying to figure out the user segments and identifying who are our best user segments, how much were we willing to pay for those user segments, and to predict out what the lifetime value was on those segments. So then we could automate a lot of the processes around how we were buying media across Facebook, Google, and as we continued to add 15 to 20 more channels, it became a lot easier to manage that workflow.

Shamanth: Interesting. I have so many jumping-off points for follow-ups from what you said. But from what you’re saying really, AI was a way for you guys to test, learn, iterate, and do it fast and rapidly and really embrace automation as much as you can. Talk to me about what you mean when you say ML, and how that’s different from what we know today as AI.

Lomit: Yeah, so I would say ML, machine learning, is one component of AI. The way we used it was to look at a lot of data and try to identify patterns around that data. Like we wanted to get, for example, an outcome to like what were the correlations and causations that were driving our highest lifetime value? And through machine learning, it was able to process that data to identify for us what were the key variables that were leading to high lifetime value.

For example, that would help us identify what was the right messaging on the paid ads that were driving the high-end users. And then in the product, it helped us identify what was the ideal user journey around onboarding in the first seven days that we had to personalize that most users were going through, that led to a higher lifetime value. And so getting those insights enabled us to start personalizing the ads and the users that we were trying to target. And then on the product, it enabled us to personalize the whole onboarding experience to get people to stay more engaged, which helped with the retention and ultimately helped us with monetization.

Shamanth: Interesting. So if I understand correctly, you guys were able to zero in on the messaging or the visual elements that were most correlated with high LTV users and that allowed you to personalize the user journey. Did I understand that correctly?

Lomit: You did. And I would say the big difference then versus now is in those days what we were able to do was to create a template framework for ads that we’re running through like Facebook and Google and other platforms. And we were able to create an API feed to change the headlines, the copy, and the call to actions and the images. The difference is we couldn’t do that with videos in those days, but now you can even adjust the types of videos. So that’s the big difference. So a lot of our creatives in those days was more around static images. We had videos, but with videos, it was more around just putting a lot of videos out there and just getting signals on what might resonate.

Shamanth: Yeah, certainly. And with the LLMs, what was your first impression of LLMs and did it feel like something completely different from the ML-driven automation and AI that you guys had been pursuing many years prior to that?

Lomit: Yeah, my first impression with LLMs—and I think it’s true even today—is that it really simplified how to interact with AI. Because before when we were doing all of this AI work, I still had to go through a product team. I had to go through an engineering and a data team to put things into place, and so there were dependencies, like things weren’t happening quickly.

Now with LLMs, you’ve got, for the most part, a pretty friendly user interface that you can prompt. You can ask questions and you can get responses pretty quickly. You can ideate pretty quickly to get to outcomes a lot faster. And so that was my big “wow” in terms of it removed a layer of what I had to go through before to get to responses and execute on ideas.

The other big difference is, before when we were doing AI when Lean AI was written, we were just really doing it based on our data set. With LLMs, you’ve got fast access to a lot of data so you’re not kind of limited in terms of how you can get insights. Like you can put things into LLMs and ask for insights, not only for how it relates to you, but how could it relate to other competitors and other verticals, and you can get a lot more of a broader view in terms of how you compare to others as well.

Shamanth: Yeah. And if you only have your data set, that’s limited. And there’s also a limited number of things you can do with that. And obviously, you can do much more with an LLM that’s trained on the world’s stuff. What’s an example where you personally replaced something that used to take a whole team and you replaced that with AI?

Lomit: I think one thing as marketers that takes time is reporting and analysis. And one of the things we’ve been able to do through leveraging LLMs is be able to have LLMs pretty much generate a lot of the executive summaries around campaigns and recommendations. And it can do it literally in like minutes versus that was taking us like weeks to do.

And the upside of that is, it enables us to still keep our teams pretty lean and have people focused less on trying to come up with the analysis, but more about how do we take that analysis and the insights to continue to build and grow different campaigns. And then also ideate around different creative ideas too. Before it was a lot of brainstorming, but you were limited to the people that you were brainstorming with. Now you can brainstorm with the people that you have as well as having a whole army of people to brainstorm with, to validate and verify those different ideas and help you stack rank those ideas in terms of impact as well.

Shamanth: Yeah. Certainly. How are you seeing that changing today in 2026?

Lomit: In terms of the creative side of it?

Shamanth: Yes.

Lomit: Yeah. I think the great thing in 2026 is that LLMs not only help you in terms of how to talk about something in terms of copy, but it also helps you visualize visual identities around creatives as well. It gives you ideas in terms of coming up with different scripts for videos. It helps you come up with 10 different images that could resonate based on those scripts and it also helps you test different hooks and headlines.

But I think the best thing for me and my team is, it just helps us communicate a lot easier with our design team in terms of what we want. Because before, we would try to put things in, but now with LLMs it can literally create that creative brief for us and it can help come up with the briefs that are a lot easier for the design team to understand. And so again, it helps the design team start off at a much higher cliff to getting the work done than having to go through two, three rounds of iterations.

Shamanth: For sure. And that’s very similar to what we’ve noticed as well, because earlier a lot of raw assets would have to be generated by the designers. And right now the strategists generate a lot of the building blocks, put a lot of the building blocks together. The designer has to do something that’s just the last 20%.

And interestingly, I wondered at some point in time if designers are gonna be out of jobs after this. What I’ve noticed is that on our team, we’ve actually grown the number of designers massively. We’ve grown the output. We’ve dramatically increased the output of each designer. I think the whole team’s throughput has gone up massively. I don’t think the designers are gonna be out of jobs because the designers are now able to do what they wouldn’t be able to do in the past, and so they’re able to do more. But I’m just curious what your perspective has been in terms of whether you see AI impacting what designers do in terms of taking the jobs or how do you see that working out?

Lomit: Yeah, to be honest, I agree with you there. That designers are ultimately… AI can do things, but it can’t do it all, if that makes sense. And I think there’s a bit of a backlash now. There’s been certain companies that have really just used AI to generate ads and put those out there and they’re being called out for that.

And then I think the key thing and why designers are really needed is ultimately, designers are the ones that will take whatever AI is coming up with and still try to make it look authentic with the human connection piece to it that people can resonate with. And I personally feel that one of the key roles of designers is it’s more than just being a designer—it’s being a creative strategist now.

Because we work with a lot of Gen Z brands here at TYB and the big reason why a lot of them love working with their super fans through TYB is ’cause the fans, when they create the content, it’s a lot more authentic. And obviously, with AI, they can create that content and they can publish it a lot quicker. But you still need human oversight to really ensure that the content is resonating and saying the right message. There’s templates for everything now and frameworks, but the designer ultimately is like the QA person before you push out a code base or a feature release. The designer is pretty much the one that’s putting the eyes and ensuring that it’s aligned with the brand.

Shamanth: Yeah. And I am also noticing that the strategist needs to be much more hands-on today. As I said, a lot of the assets are now generated by the strategists on our team. And the strategist needs to have an eye for what is good output, as much as the designer needs to.

To switch gears a bit, you talked to me about your workflow in preparation for this call. Something I found very interesting in what you said is you use different AI tools and models and you basically hit them against each other. Can you walk us through that process please?

Lomit: Yeah. One of the things that I found—and it was probably more by chance—I used to use ChatGPT a lot as an example. But clearly, Gemini has gotten better as well and so is Perplexity. By default, I would always use ChatGPT, but what I started doing in the last year or so is when I, for example, wanna ideate, come up with a SWOT analysis on a business product feature or something that we’re looking to launch or coming up with content ideas. I generally have whatever ChatGPT comes up with and then I will use the same sort of prompts and put it into Perplexity and Gemini as well to see what they all come up with.

And then I treat it like a debate stage. And so once they come up with the recommendations, I tell ChatGPT: “So what do you think of these ones coming from Gemini and Perplexity?” And I do vice versa. And it’s interesting that, for the most part, the good and bad thing is I think the LLMs are built to please people, if that makes sense. They want to make people happy and they would tell you what you want to hear. And by pitting them against each other, what I found is they ultimately encourage debate within the LLMs and by getting them to grade each other’s recommendations, I’ve ended up coming up with something where the final version after two, three layers of debating is so much better than what I initially got from all three of them.

Shamanth: Oh yeah, a hundred percent. And this is something that I and my team have had to learn the hard way, but the first output from LLMs are terrible generally. They’re average. They’re certainly better than they used to be six months ago, but they’re generally not very good. I’ve actually used something very similar with coding. Basically, when I’m building stuff with code, I ask one LLM or even like one instance of Claude to critique another instance of Claude’s code. And again, I have them verify each other’s work that helps catch bugs.

But I do have to go back and forth with them 4, 5, 6, 10 times before we have a clear consensus. Something else I also find effective is basically asking them to score each other’s outputs and just say, “Hey, here’s a rubric. How likely is it on a one to 10 scale or a one to a hundred scale to be bug-free?” And I think with all of a lot of this, my experience is it’s also important to give them some sort of rubric for evaluation because I’ve found that otherwise, they would just make up stuff anyway.

Lomit: Yeah. And in terms of the group—and you’re right I do something similar in terms of giving them to score it either out of a scale of one to 10 or a grade from A to F, whatever. But the rubric is: if my, for example, CEO or my board is looking at this, from their evaluation, from their perspective, how do you think they would be grading this?

Shamanth: Speaking again, speaking of somewhat high-stake stuff, something you used LLMs for was due diligence in earlier roles that you worked in. Talk to me about what that looked like in practice and what the AI was able to do that humans would have found it hard to do.

Lomit: Yeah, I’ve been involved in a couple of different startups and, at least since LLMs have come out, I’ve been through a due diligence process where we were looking to sell the business. Generally with due diligence, you get a whole series of questions that get thrown at you. The way LLMs have really helped me is to really get better context and understanding of what those questions are and in terms of how to go and find those answers a lot quicker.

Part of it is either going to the internal team or other parts of it is where to get the data and give it to LLM, so then they can help me respond back in a way that is concise and investor-friendly. For the most part, it saved a lot of time because before, going back to internal teams that are already operating a business while you’re going through due diligence, it tends to be a bit of a distraction in terms of the time. And so coming up with something where you can go through a couple of revisions and use AI to maybe do 60% of the work, 70%, then giving it to the teams to just validate and verify that their responses are right, really enabled us to sell a lot quicker and faster in the due diligence process when we had a number of different investors that were potentially looking at us at that time. And I think that also helped accelerate the process of getting the business sold as well.

Shamanth: Yeah. And today, 2026, there’s just like a ton of tools, ton of models. How do you evaluate and decide for your personal use which ones are worth adopting?

Lomit: I would say yeah, sometimes there’s just too many, right? And so it really comes down to what’s your core use cases in terms of how you would wanna use AI.

And for me, on the marketing side, generally it’s around content creation. It’s around improving workflows. And it’s around what we can automate on the workflow side and what AI can help us do that. And then it’s around how can we get faster insights from the different data that we have and extracting that and coming up with different ideas and running different experiments.

Generally, ChatGPT can do a lot of that for us, I would say. And Gemini as well, I tend to go to those two for different reasons. And then there’s certain like platforms we use. As Facebook is pretty solid in terms of the AI, in terms of running campaigns, it’s really about feeding the creatives into that.

Paid ads, I feel creative is a huge lever. And so what can we do on the internal side, leveraging AI just to get the team to be more efficient around getting those outputs? And then personally I like to dabble with some new stuff here and there. But generally, what I find is that there still tends to be a little bit more hype around what AI can do for people. I always feel like trying to play around with it personally first before you try to bring it into a business and use it on business.

Shamanth: Yeah, I know we certainly live in interesting times and I always learn a ton when I speak to you. And this is perhaps a good place for us to start to wrap up. Amazing to hear from you and also just hear your perspective of how AI was AI before it was AI, if you will. And I think that’s been fascinating to hear. And this is perhaps a good place to wrap. But before we do that, could you tell folks how they can find out more about you and everything you do?

Lomit: Yeah, the best way—I love to hear from people—I write a lot of content, so you can either go to my blog, which is lomitpatel.com, or you follow me on LinkedIn. I put a lot of content out there as well.

And what I would say is the big thing now, it’s less about trying to prove the value of AI, but it’s really about getting comfortable with how to leverage AI. And the people that are gonna leverage AI and get comfortable with it are the ones that are gonna be able to stand out because there’s more and more companies now that want people on the teams to be AI literate. And that doesn’t mean you need to be building on AI, but you need to be able to strategize and leverage it better.

Shamanth: Certainly. And certainly, we’ll link to your site and LinkedIn in the show notes. And for now, this is a good place to wrap. Thank you again for being on the show, Lomit. We’re excited to put this out into the world.

Lomit: Thanks for having me, Shamanth.