Building an AI-First Company — Elly Analytics’ Seva Ustinov on AI for Team Productivity

consumer software businesses, consumer AI businesses, healthcare companies, uh home services, they have this thing that website basically generates leads or signups or trials for them and actual conversions happen somewhere else in the payment system in the backend platform. It's delayed in time. So most of the existing tools they don't work with that correctly and this is what what we're solving. So, we're answering the question of what's actually working and what's not true raw revenue and drawers from each channel, campaign, and creative. And we're building AI agents that can help you analyze all of your marketing channels and campaigns and creatives and stuff and over time improve them just by talking to the system and creating rules and doing analysis. >> Hi, my name is Dmitri Bonichi and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI agents podcast brought to

you by Jot Form and featuring our very own CEO and founder Idkin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show. Hello and welcome back to another episode of the AI Agents podcast. In this episode, we have Sea Eust, the founder of Ellie Analytics. How you doing, Seba? >> Doing great. Thanks for having me here. Yeah, it's really uh really nice to finally get down to chat about all things AI and what you're doing. But, you know, so everyone can get a little bit of insight into you, please tell everyone, you know, how you got your start, uh where you've you've come from kind of to get to this point, uh with being the founder of Ellie Analytics. >> Sure. So, uh originally I started I've launched my own digital marketing agency.

That was like 15 20 years ago something like that. Uh and that was a pretty successful company with scale to an eight figure business with 100 plus employees or like building digital sales for our customers uh at a large scale and after that I realized I wanted to build uh products in the same space. I chose it carefully. So uh this is how we launched uh Ellie. I moved to San Francisco raised some venture capital and um so we we've been building marketing data platform uh for performance marketing for software companies and services uh and now we're building AI super agent for performance marketing for consumer software and services. Let me unpack this for you. >> Yeah, please. >> Yeah. So imagine someone spending like 50k per month on digital ads like Google ads, meta ads, Tik Tok influencers, affiliates, uh all those performance channels

and most of the tools today like Google Analytics or even AI tools uh they can help you understand what's working and what's not and improve efficiency. uh if you're an e-commerce business, if you can catch capture conversions on your website uh with a with a pixel and that's it. But AI businesses, healthcare companies, uh, home services, they have this thing that correctly. And this is what what we're solving. So we're answering the question of what's actually working and what's not true row revenue and rowers from each channel campaign and creative. And we're building AI agents that can help you analyze all of your marketing channels and campaigns and creatives and stuff and over time improve them just by talking to the system and creating rules and uh doing analysis. And I'll I'll I'll show that later. Uh but this is where we are right now

where have like uh dozens maybe hundreds clients. Um, we're growing fast. Uh, we're transitioning the whole company to Corser to become an AI first company, not just a first product. And, uh, there's so much exciting stuff going on right now. Uh, so I'm I'm happy to share that. >> Absolutely. No, and I I'm I'm excited to to hear more about uh some of the cool use cases you've been working on. Uh, you know, getting your company to be an AI first company, I guess. Um just you know another question about uh yourself and Ellie Analytics. You know you're moving from being uh originally what your experience was like a services company to to like a AI company. What's that what's that kind of been like for yourself? >> Well uh for me that was I was feeling it like a a completely new life challenge.

So like I've built a large scale agency dividend business. Now I want to build product AI venturebacked uh business uh in San Francisco. So like it's a completely new chapter where I can apply everything I learned from my services business to my own product. >> Um and like that's very exciting and inspiring and I feel like like it it's fun, it's challenging, uh and it's rewarding on the inside. uh I felt like it's much harder to start uh because you have to do so much work up front and you cannot like adjust on the fly like with services businesses like different clients they have slightly different requests and you just like adjust on the fly with product business like whatever you build it's there it's like really um you have to plan long term you have to find like a very small audience because you

cannot like solve everything for everyone. Uh uh everything takes time to build but on the bright side everything is compounding. So you have one client then four clients then 10 clients uh like close to 100% retention rate. So it's like growing over time. you have more resources, you have more systems in place, connectors, um, attribution modules, new reports like applying to our own product. And, uh, at some point I started feeling that like we've built something really meaningful. There's still no default option in our space uh, to solve like reporting, attribution, and automation. and we have a strong chance to become one. >> And would you say that and the reason I asked that and and great uh great response there very very cool stuff. Would you say that your services business really led you to having this idea, right? Because it doesn't seem to

be in, you know, kind of the same. Yeah, of course. Yeah. >> Like first of all, like there are two two sides of that. First of all, I've been managing digital ads at a large different scales and levels for like basically my whole life. >> So, I know what is what's what's there, what's missing, what's relevant, what's important. >> Uh, and the second thing is like our services businesses was closer to a consulting site than to just a conveyor belt. Uh so we're we've done a really deep dive into each business and like that basically like living through a hund lives with different businesses inside. So now with all of that experience and knowledge and I can apply that to my own product business. >> Yeah, I think that's that's really that's it's kind of key here, right? like a lot of people um you

know start these AI businesses for different reasons but you you've experienced what exists in digital ads right there's uh looker studio there's the built-in ads platforms analytics and none of them are really I think most of them leave something to be desired you know for for for for insights right you have to find the insights from the dashboard in front of you where um and that that kind of puts you in a in a very non-time time advantageous spot as a digital marketer. >> Yeah, especially like think about software businesses. You have this cost per trial from Facebook, from Google, from Tik Tok, from influencers, but you don't really know true customer acquisition cost for each channel or like the influence of different channels to the mix or like what's bringing the highest LTV customers, the best customers. And like often times >> the cheapest

channels actually bring the worst clients but you don't if you don't have the data like you cannot actually adjust to that. >> Yeah. The answers to the questions of yeah like you said like what is the real LTV I'm getting um is very difficult. I've attempted to set up these types of uh setups in Google Ads and it's kind of complicated to get m you have to have like multiple tools working together in order for you to get even a reasonable answer to the LTV question and a lot of times I don't even feel like it's very accurate. Um you'll have to be like like I remember I used uh mix panel um before and have you heard of that? >> Of course. Yeah, use, you know, like the there's you have to use mix panels and G4 and then pipe that into Google ads

again. It it just doesn't feel like a very good system and I think that's, you know, most people find solutions through their struggles and it seems like you yourself have have uh done that here, right? >> Right. >> Yeah. >> Yeah. Awesome. Well, I I really appreciate that insight and and what I'd love to see a little bit more is you mentioned uh you know, bringing your your company to be um more of an AI first uh company itself and and you know, we talked about it a little bit before the call and I think it'd be really cool. We don't do this as often on the show, but you know, for those listening on audio, feel free to head over to the video portion because I'd love to kind of see some of those specific use cases of uh inside of cursor cuz you

know, cursor, cloud code, all these are considered coding tools right now, but they're really more than that. They're really definitely uh they're more than that. And you showed some some cool examples. So, uh feel free to share your screen and we'll jump right into it. So let me start with like the most the basic idea first and then we'll build up more sophisticated cases uh from there. Let's say hey Corser can you analyze a new competitor called uh split metrics and add it to uh my competitor analysis folder using our workflow. >> Yep. So the idea is that with chat GPT I can like ask questions. I can ask it to Google something. I can copy some overview and descriptions of what I'm working on. Uh and that's it. Then Chpt produces the answer. I can copy things back. Uh but it's like always this

uh back and forth. And this is not what AI really promised us. AI promised us that it will can do all the work that you don't have to do anything you don't like. You just ask things and they happen. Right. >> Yeah. Right. Yeah. >> Uh and most people think that it's it might happen long long time in the future for but in reality we're really close uh to that vision already. It's just for some reason uh all these systems they are branded as a developer tools but in reality we can use them for everything for text for slides for documents for presentations for um like uh prototypes posts uh scenarios anything and like this is an example here. So I asked uh Corser to do competitor analysis and it started googling features, pricing pages, founders, their background, user feedback, G2 income terror reviews, their

own competitor comparisons and like some extra more extra researches and in the end it created a file. Where is it? Uh it's not here. Hey cursor, where is the file? Oh, okay. I see. Uh not split matrix, split matrix. Okay. There is there are some issues. Uh >> yeah, you know, it's uh it's uh it's hard for me to even understand that sometimes the the the audio audio, you know, audio syntax is never going to be perfect. >> Mhm. Um yeah. So like the idea is that on the left we have all these files like in this case it's my personal finance, it's my healthcare information, uh some home projects, learning projects here and most importantly like my my uh company files like company info, company story, business model, this competitors folder, financials, investor updates, uh product vision, current product overview uh and so on.

So basically I moved my notion to files uh here and now once they uh once I moved them here I can work with them. I can create new files. I can ask it to read other files. Uh yeah it's actually found the right company and now uh creating this file for us. And uh there are so many use cases I want to show you. Uh this one like it will take a little bit of time to to finish. Uh the nice thing that happens here is that I have rules how I want my competitors to be analyzed. >> Yeah. basically like last time for the first time it did it in the wrong way and I have okay like do real user reviews uh check all these uh review platforms check Reddit check glass door >> uh check user sentiment check uh funding grounds uh

announcements press releases and everything. So next time I ask it to do to do this task, it will follow these rules and if I don't like something, I can just ask it to update these rules. And this is how this experience becomes truly personal. >> So over time, >> my cursor understands me better and better and when I ask something, it does exactly what it wants, what I want. So, it's like it's similar to having like a personal assistant that truly learns over time what I'm expecting from >> competitor landscape overview updated as well. Uh yeah, and these are all the files, all the info. I'll take a closer look later, but basically I do this every time I I find a new interesting company in our space. Uh, and it took like like you saw it. It's just one prompt, one correction, a few

minutes of work. Uh, and it's done. Imagine. >> Yeah. It's not much work at all. >> Yeah. And imagine like with chat GPT, I would have to write like this whole long prompt, maybe store it somewhere so it knows what I'm actually looking for. Uh, so this is a case number one. Let's just stay here from for for a minute if you want to like discuss it. >> Yeah. So I mean this this is the the type of stuff that I I find interesting about this workflow first and foremost like when you're working in this and you have all these different uh MD these markdown files, right? Um, what I find intriguing about this rather than working in a LLM itself is that this is like a bunch of files you can reference for later and and and then go back and edit them much

better. Is is this one of the reasons you you feel like using this is is a little bit more effective >> workflow-wise or >> Yes, this is the most important one because it's it's it works with all my files. I don't have to copy anything back and forth. it can search it uh search my files but like at some point you'll have more and more of them. Uh like there could be prompts like this uh like go find a company story and product overview and vision files somewhere in that company folder and um come up with three ideas. uh what how to get a good PR uh about uh what we're building. >> Sure. >> Uh so like I don't have to I I know it's somewhere in my files there are this company story files and product overview and so on. Uh I don't

have to even think about them like explicitly. I don't have to copy files. No, I don't I can drag files one by one here so it gets the right but uh if your real assistant in real life can can find relevant files then the agent can find uh relevant files. So it read all of those it also found an executive summary file. Um and now based on all of that information it can it it came up with some different story angles and story lines. how can I can get a good PR uh for my business. So with chat GPT that would require to give it explicit context about everything I want to ask and here I just talk to it and it has all the context it needs. >> Nice. Now I'm reading the sorry interesting. So, how much time a day do you feel

like you're spending in this? >> Probably on average between one and three hours per day. So, I as a founder I have like a ton of calls uh with my team, with clients, with partners, with investors. Um, but like 99% of my like work on my personal tasks nowadays happen in Corser. I hire people uh and the flow of hiring people using Corser. I work with present on presentations and talks and vision and strategy using Corser. Uh it's uh I don't even know how to describe it. I can show you like more examples of how how that works in real life. But uh over the last three to four months, 99% of my personal use cases switched to Corser. The only thing >> What was the what was the like catalyst that made you realize this was even a thing for for this like type

of work, right? Because mo like I said earlier, most people look at this and they think it's uh for coding, you know? Um like uh I noticed that every single person have its own uh use case that like triggers something inside them and they like get it and and switch to cursor or clo code or like similar tools. For me personally, that was um when I wanted to create a coherent vision of the future of the company. Uh so we have this data platform as a foundation. We have this AI agents uh as a vision for the future and we're launching them right now. And um I needed to describe to to to create like a coherent story around that. Um and I could do that using a whiteboard or mirror boards. I use them. That's actually like the only tool outside of course I

use daily. Uh but that would take a lot of time and I cannot work with that as fluently as uh with AI tools. I could I tried to do that west chat GPT but I figure out that I have to retell again and again uh the necessary context and it's always forgets it and then I have to create a new chart and uh like I I cannot control it in the right way. So like I decided okay I spend a few evenings just to collect all that context like company info company story um so that I don't have to retell it every time I want to ask a new question. Yeah. Right. >> And once I had that uh here in files, uh it's just it's it just became like so obvious and so easy and I I cannot even imagine like how would I

do it differently? Why would I talk to AI that doesn't have all of this information? Why would I do it manually when I can like brainstorm on the fly and restructure things just by talking to a cursor? Yeah. And you know, I I I think that leads to another question. Um, and obviously I'd like to see another use case after after this question, but how much time Well, it's a two-parter. Sorry. You said that you're bringing people on that you want to like do this and this to be like how they're operating in work. What percentage of work because you said 99%. That's kind of why I first asked that question, right? For you, you're like spending all your time. What percentage of work do you feel like people are in your company doing this type of workflow now? What percent and then where do

you want it to and and how are you getting I'm guessing you have a goal. >> No, like so I I'm just I'm trying to gauge adoption. you don't have to have like a a hard step, but like where are you kind of at in the process of getting it to be the case? And what is your ideal goal for like your workforce to to be doing this like all the time or like that's what I'm getting at, I guess. >> Um, so different roles have like different >> Yeah. Mhm. >> preferences. No, I can't. I'll I'll better stick to this folder personal super agent. >> Um, so for example, it was relatively easy to sell this idea to developers uh once they figure out their own use cases. So for them that was like writing and like uh auto tests like quality assurance and

updating documentation because previously that was a very tedious and boring tasks that nobody wanted to do and now they could do it with just like one prompt. Uh and like they found that use cases that use case they loved it and like they started experimenting with uh others. Now, I think our best de developer uh writes pretty much all the code with AI with multiple windows using cursor and cloud code and spends something like a $1,000 per month on tokens, but the productivity is extremely high. Others, they could use it here and there. So, it's like um uh gradual uh transition. Um and and that's fine like the the hard part is to find your first use case for develop for project managers because like the data platform we have integration for each client connecting to all the data sources it takes some time. So

we have this small integration projects for uh each client and we have analysts and project managers who does that. For them, the tipping point was when they realized that they don't need to spend an hour per day updating the product docs and follow-up emails. Uh let me show you an example. Uh so go check the latest um meeting transcript in the transcripts folder in the projects folder and update um client beauty one status card. So in this ca in this case the transcript is already downloaded. In reality, it would connect to Fireflies, get the latest transcript and update everything on its own. Um, so it read this this transcript here. Uh, and now it will update this card. Uh so imagine you're a project manager >> and the only thing you need to update your like project status file is to ask your cursor to

read the latest transcript and do this work for you. Yeah. based on the whole transcript this thing uh like wrote the updates like the attribution promo code issue fixed uh uh something else happened uh technical readiness everything is in green and so on so like uh project managers can have their own rules here I believe yeah uh what they want what's the format what's the structure uh what they want it to follow uh to analyze to highlight and so on. And uh all these updates happens basically on their own like this is what boat quote unquote the bolt in the project managers. Um our operations team like we do a lot of hiring nowadays. Uh so there is a lot of recruiting to have it here. Mhm. >> Yeah. So basically they like those people they never ever saw code or programmed anything on their

own. But it took like an evening or two to connect to our >> uh ATS application tracking system to get all the updates and information about candidates into files uh here in Corser and to set up uh workflows that gives like a second opinion and recommendations about each candidate looking for gaps in their like uh their CV and interview and our requirements. Uh so like we're not delegating decisions to AI yet >> but we're using it as a second opinion and uh like once you do that for the first time you can never go back and also that expands capabilities. So for for example our project management team leads previously they had to go through each um meeting uh on their own. Now they can just analyze transcripts. >> Yeah. So you ask like how much people on the team actually use it. >> Correct.

>> Uh so I think like some people use it just maybe once or twice a day just to like as a substitution for CH GPT. Uh but all the managers and leads uh like team leads uh senior stuff. Uh basically the more responsibility and tasks you have the more you use uh corser because it becomes like a huge leverage for any kind of repetitive tasks even intellectual even uh uh like managing like 30 integration projects at the same time with dedicated project managers. It still requires like so much effort and attention. uh but uh AI can help with that a lot and especially AI in cursor or clot code uh because they can follow your own workflows and you can guide them and you can automate lots of repetitive things. >> Yeah. Yeah. And it and it kind of works across departments too, which I

think is really curious or not curious, sorry, really exciting. You know, like um whether you're in marketing and uh sales and and and engineering, we were just talking about it beforehand. There's there's so many different um areas that you can you can really enhance your day-to-day productivity with this. >> Absolutely. Yes. Uh I I forgot to mention sales and marketing, but uh those guys use it. they were the first to to try it out and and and to adapt actually. >> Yeah. I feel like that's a that might be a common thing in in businesses in general is like sales and marketing usually is just trying to like see how they can do more, right? Because like sometimes with marketing is just like doing more marketing doing more sales, right? And will give you results. Um whereas like the quality of of things um uh

in some other aspects of the business like code and and whatnot, obviously this is very beneficial for it, right? But the quality of the code still needs to be there and and whatnot. So >> Mhm. Uh let me show you a few tips that might be when you use it to use things like that. >> Um so first of all uh I always create this corser rule files courseer rules uh and I describe everything I would want my assistant to know. uh the structure of my files and what leaves where. Um that's number one. Sometimes I just add like a brief description of everything so it knows what I'm talking about. Uh then very important part is like anti-h hallucination rules. Uh because when you work with cloet especially with clotssonet uh it tends to like write things that are not based in the ground

truth uh because it's very proactive and it starts creating files and you just you ask it for one thing and it creates 10 more things. Uh so I have special rules to >> yeah basically treat everything we we write here together uh as as a specs for the creation of the company. So like the code is a source of truth for developers and all these documents like strategy and vision and everything uh they should be treated not just as text but as a building blocks for the for the company and uh every statement should be could be tracked to its sources uh marked as canonical if it's like a source of truth here or a reference like where did we get this or for example this file with company competitor research uh is uh referenced from the file with competitors overview uh that is referenced

from executive summary file and all of those references they have this mark. So next time AI wants to change something it can find all the places that are like dependent on this. um I ask it to reuse all the phrasings exactly because it tends to like change words slightly. It changes the meaning slightly. But when we are talking about like for example ideal customer profile definition, we want it to be exactly accurate not almost accurate. uh and if it needs adaptation uh I asked it to mark different statements as adapted based on the original source from like referenced referenced file. So like this 10 lines they change the performance uh of the agent completely. it stops doing things on its own that uh only like uh that you don't want to and um start uh doing exactly what you asked it to do. Uh yeah,

another block like anti hallucination rule uh is basically the same. They just repeat it in different words to to make it more clear what I wanted to do. >> Uh >> very smart. And there are more things here uh and I I I share it as an example so people don't have to start from scratch. Uh we won't go through all of them this time but anybody can uh can read them can adapt them. And the basic workflow here is that I just work with cursor on my daily tasks and when I see that it uh deviates from the expected road I ask it here like I ask it here to just update this these rules. So uh it's not like a some kind of a special process to set up your rules that takes a week and it doesn't work. No, I just work

with it and gradually improve rules just by asking it to follow specific workflow next time and that's it. >> Yeah, I think that's, you know, guide rails are really important with AI. Um, without them, they can kind of go off the rails for a lack of a better phrase. Um, and kind of go and do a lot of work, but maybe necessarily might not be the type of work you want, then actually cause more confusion later. So, that those are really important. >> Exactly. Let me show you how to start with this whole workflow because it might seem uh like it's a oh it's a developer tool. You'll have >> right have someone to install it for you but in reality it's extremely easy. >> Uh so step one we go to corser.com. We download the corser. We >> click next next next. Uh, and

it's done. Um, then we go to Whisper Flow. Uh, we download it. We go through next, next, next, next next, next next, bind the key. I I prefer like a function key. Um, so now anywhere on my computer I can uh give me alternatives to whisper flow. just using as this uh as an example uh that anywhere on my computer I can press a button function talk to it and uh it will >> I've seen you've even made your uh your wait did you make your your search bar perplexity >> uh yes and actually like my main browser is comet but there are not that many use cases yet uh >> yeah that's cool though I didn't know about that that's that's a really I mean honestly like that's what a lot of people would prefer anyways to because Google I think >> pretty good.

Yes. >> Yeah. I think Google's great obviously. Um but this I I I know a lot of people love Perplexity for obvious reasons. So this is cool. >> Mhm. >> I like this. >> Oh, I personally use mock whisper because it converts speech to text locally without sending it to servers. >> But whisperflow is the best default option. It works for Mac. It works for Windows. Uh it works great and uh it's pretty secure. I just my personal preference is to you to to do it locally. >> Nice. That's really cool. >> And then um so I've created this uh repository basically a collection of files uh to for people so people could start not from scratch but like from some from some template. >> Mhm. Uh so when you open the cursor for the first time you see a window like this open project

clone repo connect something you don't need to know anything you just press open project uh create a new folder folder three 4 and open um and here's your cursor zero files zero files open. Just just it uh and I say, >> "Hey, download this uh locally to my to my folder and don't do it as a subfolder like place it right here uh and guide me through next steps." >> And I give it a link to this template. You see, I noticed that even the word like GitHub or repository scares people a little bit. So like the nice thing about AI agents and and uh Corser that you don't have to know how that works. Uh you know that like okay that's a template it could be downloaded. I just ask uh Corser agent to do that for me. It does its magic. uh like

downloading files, doing something here. Um I >> Yeah, it does it without you needing to know how to do it, right? You say do do it and then it does it. Yeah, >> you still like need to know what you want to achieve like I want this template on my computer, but everything else could be done like without technical details. >> Uh yeah. So here are talks all all things I showed you u some basic scripts readme files cursor rules uh so like everything I show you today you could explore it on your own uh you could ask corser to to explain what's going on here and how to use this examples yeah oh oh one more extremely important thing Um, at some point you'll probably want to wipe code something. >> Yeah, for sure. >> Like a connection to Gmail or to Calendar or

to Notion or to Fireflyy's transcripts. >> Yeah, whatever it is. Mhm. >> Yeah. Let me show you like one life hack how to do that. I'll start a new chat and say I want to connect to Fireflyy's API and get uh latest uh transcript and I want to do it repeatedly uh like uh do a research on how to do that based on the like uh specs or best practices. or something. But most importantly, um, look for real user reviews and testimonials >> and suggest a way that actually works for people. >> And like that's a prompt uh that gives you a 99% working solution. like it's not just something that could work. It's not just something googled from the internet. It's actual solutions that work for real users. And uh this way you don't have to to spend like hours or evenings uh figuring

out like different roads. You can just choose the right road from the start. >> Absolutely. Yeah, >> it works with chpt and everything as well, but uh when you want to create a new connection or some code or something or open source project, download it. Uh you want >> Yeah. want it here. Um what do you think about this whole idea of transitioning from chat GPT? >> Well, I I think it's a lot better for a couple different reasons. Um like uh we mentioned at the beginning uh there's the documentation and all in there uh whether it be the the cursor rules and guide rails and you're also able to like organize everything in there a lot better. Chat GBT for all intents and purposes has like those different um uh you know GPTs you can add in there and it has some reference knowledge

you can keep in there but it doesn't it doesn't quite have this level of uh memory you know working memory is kind of is kind of what's really key here and that and that that brings me to kind of uh where I'd say my last question would be for you as we do we are hitting the uh top of the hour here so I'll just ask um where do you feel like this is going to impact workplace productivity um on the masses right like this is something you're doing that I think is really advanced as really cool for your own company as well when do you think this is going to hit most people that they should start working like this or will it ever happen or is the market just long on this >> uh I believe it's already happening uh when you think

about like why combinator companies I think 90% of them use workflows like this for everything. Uh companies I know like maybe a few% of them actually know this and start adopting it. They hire like this time they hire like real AI officers that help them drive transition and uh implement more of AI workflows into their companies. But um what's already happening is that companies are some of them already stopped hiring noni native employees junior senior team leads managers uh it's like if you have three candidates and one of them already experienced this kind of workflows and knows how it works and passed the learning curve and and bring their own use cases to the company uh will definitely have a much higher priority over those with like relevant experience but who still stuck in the CHP era. >> Absolutely. No, I think that makes a

lot of sense. So, with that being said, uh as we wind things down, what would be the one uh thing you'd like to shout out? You know, obviously for people to go and check Ali Analytics out is what I would recommend, but is there any final thoughts or shoutouts you'd want to give uh to close the episode out? Let me show you like a glimpse of the future where things with uh AI are going and what we're building here. Uh so basically imagine the way you work with um files and docs and everything here in Corser. Uh you can work with your own uh ads and analytics. So we're basically building a corser for performance marketers from consumer software and services companies >> and you can talk to it in a way like hey can you show me the metrics you have access to and

build a simple report with seven key metrics for the last seven days. It's in alpha version. It's uh so like same the same way it works in corser with simple files and scripts in a similar way it works here with the whole data warehouse and data. >> I see it's doing an SQL query. >> Yeah, it's doing SQL queries. Uh it can um have access to all the tables and connections and descriptions. uh you can have like six sequential queries and everything. It will just build this report we want in a minute. Uh and this like layer one. Layer two is when we ask it to create rules that will for example stop burnout creatives the moment it shows the signs of burnout. So we don't have to spend extra money on non-effective ads or scale top performing ads. >> And that's what we're launching

today. And the next step that will be built like in the next couple of months I believe uh is like actual iterations and changes on the ads itself and keywords and creatives and add assets and everything. So you can uh send a prompt um hey analyze everything we've done so far with our ads. Analyze changes from last week. analyze trends uh from u Google ads and Tik Tok ads from their libraries from of our competitors and suggest changes for the next iterations of uh tests uh and all of that could be done automatically okay with oversight of performance marketers. It's always there in the driving seat. Uh but it basically 10xes the powers of marketers. So they can launch instead of launching like 20 tests per week, they can launch 100 or 200s and find what's really working and scale it and do it efficiently.

And this is where things are going. >> Well, that's amazing. I think uh that's going to be awesome for everyone to check out. Um so everyone make sure to go to uh leanalytics.com, check out everything they're doing there. Um, we really appreciate having you on the show, Seah. >> Thanks for having me. Happy to share all this knowledge and use cases. I'm so excited about AI and where it's going. >> Thank you so much. Well, have a wonderful rest of your day. And to everyone listening, please leave us a like, comment, and we'll see you in the next one. Peace. Bye-bye.