Product Led Growth Leaders — Say Goodbye to Guesswork: How AI Is Redefining Ad Optimization

---

date: 2026-01-27 format: media type: podcast language: EN participants: - Seva Ustinov (Plurio / Elly Analytics) - Thomas Watkins (Host, Product Led Growth Leaders) topic: AI Agent for Performance Marketing status: published owner: Seva source_refs: - Product Led Growth Leaders podcast, Episode 177 - Original transcript source + Apple Music published version published_links: - https://podcasts.apple.com/us/podcast/177-say-goodbye-to-guesswork-how-ai-is-redefining-ad/id1724028481?i=1000754870281 - https://open.spotify.com/episode/7iTNAti1VbQpDQnP6KkmHj notes: - Discussion of Plurio product, performance marketing, and PLG approach - Also covers Seva's Cursor setup for company operations - Website transition: ellyanalytics.com → plurio.ai - Recorded January 27, 2026, published March 2026 - Product descriptions reflect earlier stage — need to be updated for current product state - L-Analytics / ellyanalytics.com mentions = old branding, ignore - Personal stories, opinions, identity, and storytelling are evergreen and accurate - After-the-cut conversation includes valuable Cursor/shared workspace discussion

---

Podcast Details

Title: Product Led Growth Leaders, Episode 177: Say Goodbye to Guesswork — How AI is Redefining Ad Optimization Host: Thomas Watkins (3Leaf) Guest: Seva Ustinov — Founder & CEO, Plurio / Elly Analytics Date: January 27, 2026 Language: English

Published Episode Links

• Apple Podcasts: https://podcasts.apple.com/us/podcast/177-say-goodbye-to-guesswork-how-ai-is-redefining-ad/id1724028481?i=1000754870281
• Spotify: https://open.spotify.com/episode/7iTNAti1VbQpDQnP6KkmHj

---

Raw Transcript

Thomas Watkins: There, we'll just roll right into it. All good.

Seva Ustinov: I think I'm good, like, I'm… Yeah.

Thomas Watkins: I think so, yeah, I think so. Okay, okay, great. Okay. Welcome to Product-Led Growth Leaders. We've got a great guest for the show today, Seva Ustinov. He has a background as a software engineer in leading startups, and today he is the founder and CEO of Plurio, which is an AI agent for performance Marketing teams. Seva, welcome to the show.

Seva Ustinov: Thanks for having me here!

Thomas Watkins: Yeah, absolutely. So let's start off, like I always like to, in understanding the problem. Help us understand, before we get to the solution that you're building, what are the problems and challenges that people are dealing with?

Seva Ustinov: I'll tell this as a story. So, originally, I was building my own marketing agency and scaled it to 100 employees, and we were responsible for revenue growth from digital channels. And then, me and my co-founder, we decided to launch software product, and we thought, like, oh, okay, what's the core of that? It's data, attribution, like, to run ads efficiently, you need the right data. So we launched that. And then we saw our clients, marketers, spend… like, 4 hours a day looking at those numbers in our dashboards, and making decisions, like, what should I do about this creative, or that campaign? Should I scale it? Should I stop it? Should I relaunch it? Should I change something there? So we decided to build an AI agent that automates that, freeing up Many, many hours of, Media buyers and performance marketers, while, improving efficiency by better decisions, faster decisions, Managing, like, large-scale ad accounts.

Thomas Watkins: Right, I mean, anytime people are doing something hugely important like that, like making sure that the company stays alive for having… from having enough revenue coming in, you've got to make sure the decisions are sound. Can you give us an idea of some of the different types of metrics that people would have to stare at for so long to try to make sense of?

Seva Ustinov: That's a great question. So, we specifically focus on segment of consumer software and services, so what they see in their ad accounts is, like. What's their ad spend, CPM, CTR, cost per click, and cost per leads? That's the only thing they see there. What they actually want to see is how many Qualified leads. Actual sales and revenue, maybe trials and conversion to paid subscribers. Each ad's generated, so they want to optimize for… revenue and LTV, not just for leads or sign-ups or trials. And, like, this is what our agents actually do, like, collect the data and help optimize for that.

Thomas Watkins: Okay, so before we get to the agents, it sounds like you've done a few things even before the technology, because if I'm a decision maker, and I have to look at data. if… you're kind of making the case it's the wrong data a lot of the times, right? Like, it's metrics that aren't as important as they could be. Why is it those metrics are used? Are they just easier to calculate, or are the people who are putting these interfaces together, are they missing something?

Seva Ustinov: It's actually about, like, the… real life… complexity of technology. So when you run e-commerce. You just put a pixel on your website, and it tracks all purchases, because they happen almost the same time the ad click happened, and you can just merge those things together. Google ads do it, meta ads do it. And, third-party services. They're pretty… there are a lot of good solutions there. But, when you think about… Consumer software, fintech, AI services, on the web, like, the customer flow goes like this. They see an ad, they click, they look at the website or quiz or something, they sign up. And that's what can… that's a lead, that's what could be tracked by a pixel. And then actual conversion happens like, 7 days later, 2 weeks later, sometimes multiple months later with LTV and stuff. And the data is stored somewhere in Stripe, in backend, in the CRM system. So, this is why it's hard to collect that data by ad platforms, and this is why you need to first centralize this data, and then take those delays into account when you make, like, automation of the ads and make decisions about the ads. So this is why people stare at these dashboards and think, okay. Do I think that these leads will actually convert to sales, and should I scale these ads now? Or should I wait, get the data, lose some time there, because I don't know actual revenue from that ads? And that happens at, like, they run Hundreds or thousands of ads at the same time. And this is where it gets tricky.

Thomas Watkins: Now, going into the, about to go into your solution, but are you coming from this from a stance of. you kind of know what those folks need. People who are on revenue generation teams, you know what they need to see, what they need to know, and you're helping give them that better intelligence so that they can operate better. Is that the foundation of how you're… Delivering your solution, or is it you're just… Building the technology to do it better, and cheaper or easier.

Seva Ustinov: I think it's kind of both. So we started as a data platform, so this is where we collect the data and give that for people to make decisions. What we noticed is that, like, larger companies, they have in-house team, they have some kind of analytics and dashboards and attribution already. So with AI Agent, we can connect to whatever data you have, to our data platform, to your internal data platform, and… Like, you essentially see a chatbot, and you can ask questions like, hey, help me analyze all of this and suggest best actions to improve efficiency.

Thomas Watkins: Okay, okay, so now, getting into the solution itself, it is an agent, so is it kind of an LLM type of a experience? Okay, and then, as a business owner, I have this agent from Plurio, and I, effectively, it becomes almost my chat GPT of sales decisions, and it's just, like. It's specifically good at Helping me with my sales and revenue decisions and product decisions.

Seva Ustinov: It's specifically good with, performance marketing decisions.

Thomas Watkins: What is performance marketing, for those of us?

Seva Ustinov: Google ads, meta ads, TikTok ads, Reddit, influencers, affiliates, so channels, ad channels where you can put money and generate leads and sales in return. So, not brand marketing, not outreach. Add channels that are scalable and that actually drives revenue.

Thomas Watkins: Gotcha. Okay, so, so looking for that sort of behavioral stuff of… and getting those numbers and statistics and metrics up. And that actual action… actionable stuff, right? What people are doing with regard to… the sales pipeline. It's specifically…

Seva Ustinov: Helps you understand performance of your hundreds and thousands of ads, ad sets, campaigns.

Thomas Watkins: The ads, gotcha, gotcha, gotcha, gotcha, gotcha. So it's still, it's still about the ads, gotcha. So this would be the same as if you're doing a campaign, and you're saying, I'm running this On these platforms, it's either an image, a video, or a text, and you set it up, and you set all these parameters around it, and then it's got some kind of a performance that, it comes back with and says, okay, this is how many people opened it, this is how people… how long people stayed on it, this is how many of them led to leads, and so on. This is… so that's… that's basically what it is. And then this is taking all of that as intelligence That you, as the performance marketing person, can go to one place, even though you're using lots of different tools, and operate it from there?

Seva Ustinov: Yes, so it centralizes the data and makes it available to your agent. So imagine you can ask things like. Hey, I launched… 100 new creatives last week. which of those should I stop, which of those should I scale, and what's the right moment to do that? Like, previously, people had to, like, guess, like, okay, maybe at $50 spend… we… we can, like, make that decision, stop or scale. But the agent can analyze historical data. and come, okay, like, this is the exact set of conditions that indicates that this ad should be scaled faster, and give you more revenue, and this should be stopped, because there's no chance they will actually generate high-quality customers. Same thing with scaling, same thing with analyzing like, different… landing pages, like, should we drive traffic here or there? With days of the week, with different targetings, with different offers, like, what kind of ads works better for us, not just in terms of number of leads, but actually in terms of number of revenue and LTV. And, like, When you have… Hundreds or thousands of ads running at the same time, Having an agent that can do that large-scale analysis and then automate those actions, you can actually tell the agent, so, like, hey, I like your logic, Let's do… this set of actions under these conditions automatically every day, and I'll just confirm them. So it's analytics… Insights and suggestions, and full automation of that large-scale ad management.

Thomas Watkins: So it has some autonomous behavior that it can do. It's not just that you're… so I think that's… goes into my next question, is if you can help the audience understand what's the difference between using, Plurio versus just pulling up your ChatGPT account or, your Claude account and just throwing it in there? What do you have that's special above and beyond that basic kind of LLM stuff? And I think you mentioned one agent. I don't know, I haven't played around with all of the agentic features of the popular LLMs yet. In terms of how autonomous it can be. But, what are some of those differences?

Seva Ustinov: So, first of all, Our agent has access to your data. and understands how to query it and makes it accurately. It can literally generate SQL queries to the data warehouse and pull the right data. Second, it understands sales and marketing stuff in general, like, what, like, that you should analyze? performance in cohort view, not in cross view. It should take into account delayed conversions and make forecasting, and so on. but more importantly, it understands your business specifically. So we… when we set up our agent, we do an onboarding flow, where we… collect the right business context about your sales funnels, your channels, your ad strategy, your product offerings, your typical conversion timelines, and LTV and stuff. So when you ask a question, like, okay, what should I do about these ads? It already knows all those details, while other agents, they will just give you some generic replies that might be… that even might be wrong, because they don't take into account things I mentioned. So this is where the… Key difference from generic solution comes in.

Thomas Watkins: Gotcha. And so when you say, has access to your data, as a new user of Plurio, what do I do? Do I go to, Plurio, and I sign up for a subscription, and, then, then is there some kind of a setup process where I get access to the data, or is it just over time, you're saying, hey. here's this link to this Google folder, and I want you to see the stuff there, or here's a link to this data table, like an Airtable or something like that, and I want you to pull data from there. How does this linking process go? Is it, like, onboarded? Is it over time? Is it… how does it come together?

Seva Ustinov: Right now, it works like this, so we have a dedicated analyst and customer success manager that have, like, a couple of calls with you to ask a lot of questions, and then they, like. Process it and put as a business context to your agent, and you can actually see it on the platform, ask your agent questions about it, or edit it. And over… and same thing with the… connection to the data. Like, we help it set up, make sure it's accurate, make sure Agent understands all the specifics of your setup channels for sales funnel metrics, and so on. After that, you can just ask your agent, like, the moment you notice, like, okay, by the way, we launched a new product, and this is… these are campaigns about that product, this is how you identify it, please, like, remember that. and it will save it into the project context, and every time you, like, start a new chat, it will already know what's new about your business. So we help set up it then first, but then you can do it on your own, just by talking to your agent.

Thomas Watkins: From a PLG perspective, I'm always thinking about the continuum where on one end, you have something that's totally self-serve, and you know, people log into the website, and they… they click, and it's totally just them and the machine, versus on the other end of the continuum, people help along the way, and it's consultants in the background, and how much… where are you on that continuum? Are you guys all… almost all software, except for some of that setup stuff, the consulting that you offer?

Seva Ustinov: I think we are moving in the direction of the self-service. Right now, we still need those consultants for the whole onboarding process, but over time, we add more things to the interface so people can actually try it on their own. Because it's actually, like, a pretty streamlined process that could be done with a human, or with the AI agent. It may just ask you a few questions, or ask you for a link or something. So, it makes… sense to do… both. My personal fav… I'm a big fan of High-touch sales and high-touch onboarding at first, knowing your customers, understanding their true needs, giving them direct feedback and solutions they ask for, and then automating it into a self-service flow.

Thomas Watkins: Yeah, excellent. And, you know, so, what, give people the website here is, Elly Analytics?

Seva Ustinov: Right now it's ellyanalytics.com, we're in the process of migration to Plurio, I think by the time this…

Thomas Watkins: Let's… let's… let's make sure… let's make sure that we… everyone hears the… what it's going to be, because by the time people hear this, I think you'll probably be set up. Plurio.

Seva Ustinov: Plurio.ai.

Thomas Watkins: Yes, Plurio.ai, P-L-U-R-I-O dot AI. Ladies and gentlemen, we're, speaking to, Seva Ustinov. And so, what are some of the things that, people have to look forward to in your roadmap?

Seva Ustinov: Oh, that's the cool stuff. So, right now…

Thomas Watkins: Only the stuff you feel comfortable talking about, of course.

Seva Ustinov: I've got, yeah, it will be on the website soon as well. So… Imagine the full cycle of ad management. You generate hypotheses, like, what could work, what could, bring you much revenue? What kind of ads? What kind of… Creatives, hooks, offers, landing pages, Pricing, everything, all the variables. Marketers and product managers tweak in order to find the right, That's like…

Thomas Watkins: Like, the levers to pull?

Seva Ustinov: Yes, so, like, they're, like, right now they're doing that manually, they're launching, like. A dozen, or maybe a hundred of new creatives each week, and see how that performs. They… what we have right now is automated analysis of what's already run, automating scaling and downscaling of the ads you launched manually. And what we are going to do… and that's already very valuable. We see customers use it and love it, and it saves them a ton of time, and so on. What we're going to do next is we want to… run… hundreds of agents. Every night… every night to analyze hundreds of different segments and perspectives to find the right insights from the data you have. Like, what kind of settings actually work better for you? So you don't have to ask it. It's constantly searching for those insights, and every morning will bring you, like, hey, you probably should stop advertising in that region, because there were no actual revenue from there. You probably should focus on video creatives more, because they're… 23% better than image creatives, except those few cases. Like, insights like that, every night from large-scale accounts, like, that is gold.

Thomas Watkins: It is. Because it's so hard. Yeah.

Seva Ustinov: The next step would be to… out of those insights to build a self-updating knowledge base, like a perfect playbook, how to run ads for your specific case, product, sales funnel, regions, and stuff. And check that, like, people who run ads actually follow that playbook automatically. and then analyze competitors' creatives in the same way, like, what's trending, what's scaling fast, what's working there. And from all of that, generating new hypotheses, what to… launch next, like, next hundred… like, next dozen of creatives to try, and sales funnels to try. We want… we probably won't do actual ads generation, because there are so many products already out there, but we're giving, like, a formula of the next best creative for you to generate with humans or AI services. And then we'll manage the rest of the cycle.

Thomas Watkins: Oh, that's really, really exciting stuff. And this is kind of the brave new world of AI-empowered tools, and it's just opening up whole new worlds of what's even possible that were just… would have been totally impossible just 5 years ago.

Seva Ustinov: Exactly. Yeah, I'm extremely excited about this.

Thomas Watkins: Yeah, really good stuff. Well, ladies and gentlemen, hopefully we all can go and check out Plurio. You said Plurio.io or AI?

Seva Ustinov: AI.

Thomas Watkins: Okay, Plurio.ai. Plurio, as in plural. Plurio.ai. Everyone, go and check that out. And, yeah, exciting stuff. Well, thank you so much for being with us today.

Seva Ustinov: Thank you, pleasure having this conversation.

Thomas Watkins: Okay, that's the cut. Great! What'd you think? I thought that was really good.

Seva Ustinov: Yeah, I love it. I like it. At first, like, I had to… find the right words, how to explain our domain and segment and stuff, but after that, it's, I love it.

Thomas Watkins: I always, in people who are multilingual, to be able to talk in depth about business in multiple languages is, that's amazing to me. But by the way, where are you located?

Seva Ustinov: San Francisco.

Thomas Watkins: Okay, where are you from, then?

Seva Ustinov: originally from Russia, then I lived for 6 years in Bali, then a few years.

Thomas Watkins: Lived all over.

Seva Ustinov: Nomads, like.

Thomas Watkins: I've lived all over the place. Yeah. I'm so jealous of digital nomads, because that's kind of the life to live, and if you can do distributed work, you might as well travel and do stuff. So…

Seva Ustinov: That's fun. That's really fun for a few years.

Thomas Watkins: But then it gets old.

Seva Ustinov: Then you get, like. The fun part is that everything is new each time and generates so much emotions and dopamine.

Thomas Watkins: But…

Seva Ustinov: After, like, a year or two or three, it's just, okay, whatever, new hotel, new country, what's… what's the.

Thomas Watkins: Oh, interesting.

Seva Ustinov: And second, like, Close human connections. It's so hard to maintain them not in person. So, like, at some point, I decided, like, no, I want to settle, I want to be close to my best friends. And that's more important than just seeing, like, country number 41.

Thomas Watkins: Right. That's a cool problem to have, but I guess it's… you're past… you're over it now. You're past that. Cool, alright, well listen, the, the episode will come out probably in either late February, possibly early March, or something like that. So yeah, I think by the time you have your website out, people will see the new website. And then before we jump, I don't know if you know what we do. Did you… did you check out what we do for, at 3Leaf? It's okay if you didn't. You're… you're a busy dude. Let me, Let me actually… let me share.

Seva Ustinov: I'm happy to hear it.

Thomas Watkins: Yeah, yeah, yeah, just real quick before we drop. So we do 3Leaf. We do, product design, largely for, startups and anyone working on, kind of, new ventures. And, my computer's, like, reacting really slow right now. Oh, I hate when this happens. I think I have just too many things running. Okay, there we go. So, we do UX, like, the full scope of UX, anything from, you know, like, on fractional team basis, anything from research to the core information architecture to the, graphic design. And essentially, we… fill in the gap between the, vision and how things actually come out on the product. So, when folks are trying to become, like, the clear category leader, better than the rest, like, way better, natural, smooth UX, like. you know, like, Slack and Zoom and the way they are, and kind of, become their sort of team doing and crunching out the work for that. So, anyway, that's just a little bit about us. Wanted to throw that out there. you being in San Francisco.

Seva Ustinov: This is your portfolio, and yeah, I like this clean style, visual accents, And, and.

Thomas Watkins: Like, even the good behavior, so it's like the good… it looking right, but, like, even when people click through and they do stuff, that stuff's easy to use. Like, sometimes I talk to startup founders who, they're like, we have lots of features, but people don't find them. Or, you know, people start using our product, but they get, like, halfway, and then they, like, drop off and stuff like that, so… our whole thing is to make sure the whole process is super smooth, super easy, and just feels, like, really fluid. So, we use, like, our kind of design psychology experience of doing that. So, anyway, that's a little bit about us.

Seva Ustinov: Nice. Love it. I actually have, started building in December, like, my personal CRM in Cursor.

Thomas Watkins: Oh, okay.

Seva Ustinov: throw…

Thomas Watkins: Oh, Cursor! Are you liking Cursor?

Seva Ustinov: Oh, yeah, like, I basically live there. The latest thing… so we have a shared workspace for the whole company, so, like, we see, Chats, messages, meeting transcripts. product docs, project docs, code, and it's accessible for everyone on the team. Like, okay, what's going on?

Thomas Watkins: You just, you built it in cursor.

Seva Ustinov: Yes, it's basically like a set of Git repositories. I just had to onboard non-technical people there, but it works fine. So I can ask, like, hey, this client asks… how we're different from those guys. Go check competitor analysis, go read all their meeting transcripts so you know what they're asking for, go check our knowledge base, and if you don't find something there, go check the code, and then read my… Communication style, and give me an email draft.

Thomas Watkins: Oh…

Seva Ustinov: And it's, like, 2 minutes working, and it's done.

Thomas Watkins: And it's better than ChatGPT, because you're not just sitting there back and forth, like, explaining the situation every time and doing…

Seva Ustinov: With ChatGPT, I would have to, like. Manually drop all the relevant docs.

Thomas Watkins: Yes.

Seva Ustinov: And even that would be hard, because, like, how, like… How do you do that when everything is different places? Now, everything is here in MD files, so… and with cursor rules, on each, level, so when I, ask it, like, go get those things, it knows where to search for them, and when I ask, like, go… check our competitor analysis, it already knows from rules that it should check the strategy folder, the competitor analysis folder, specific, Structure there, and… it's… technically, you can replicate it in ChatGPT, but each query will take, like, 20 minutes, and it's better to do it manually, probably.

Thomas Watkins: Right, right. So many… we're in the future. We are living in the future. Good stuff. Awesome, Seva. Well, anyway, good, talking to you, and, yeah, stay in touch on LinkedIn or whatever, and, and yeah. So, have a good one.

Seva Ustinov: Happy to connect, really enjoyed this. Yeah, let's stay in touch.

Thomas Watkins: Likewise. You too. Have a good one. Bye.

---

Key Quotes

"I was building my own marketing agency and scaled it to 100 employees... we saw our clients, marketers, spend like 4 hours a day looking at those numbers in our dashboards."

"Performance marketing: Google ads, meta ads, TikTok ads, Reddit, influencers, affiliates - channels where you can put money and generate leads and sales in return."

"Our agent has access to your data and understands how to query it accurately. It can literally generate SQL queries to the data warehouse and pull the right data."

"I'm a big fan of high-touch sales and high-touch onboarding at first, knowing your customers, understanding their true needs, giving them direct feedback and solutions they ask for, and then automating it into a self-service flow."

"We have a shared workspace for the whole company - chats, messages, meeting transcripts, product docs, project docs, code - accessible for everyone on the team."

---

Key Topics Covered

1. The Problem: Marketers spend 4+ hours/day looking at dashboards, making decisions about hundreds of ads
2. Why It's Hard: Consumer software has delayed conversions (7 days - months), data scattered across Stripe/backend/CRM
3. The Solution: AI agent that centralizes data, understands your business context, and automates ad management
4. Differentiation from ChatGPT: Connected to your data, understands marketing specifics, knows your business context
5. Roadmap: Hundreds of agents running overnight for insights, self-updating playbooks, competitor analysis, hypothesis generation
6. Cursor Setup: Shared workspace for whole company, everything in MD files with cursor rules