Exit or Die Podcast — AI-First Teams & the Future of Performance Marketing

date: 2026-03-27 type: podcast title: "Exit or Die Podcast — Episode 13" participants: Seva Ustinov (Plurio), Peter Oberlik (ExitUdAI) language: en duration: ~44 min links: spotify: https://lnkd.in/dkqRvDpk youtube: https://lnkd.in/ds8Mdcrc apple: https://lnkd.in/dsFSwy8m notes: - Seva as guest, Peter as host - Topics: dashboards broken, AI agents in perf marketing, speed as bottleneck, AI-first org transformation, shared workspace, fully autonomous marketing endgame - Unique moments: "yolo mode" (gradual trust), 300-person company founder aha moment, "agent" = most overrated AND fundamental buzzword, cost per lead obsession

Transcript

0:08 Hello guys and welcome back to Exit or Die the no-bullshit podcast for founders, investors and the whole startup ecosystem.

0:14 Today I'm joined by Seva, founder and CEO of Plurio, an AI platform designed to turn performance marketing teams into AI-first organization.

0:33 After more than 20 years in performance marketing and building a marketing agency, over 100 employees and over 50 million in revenue, Seva realized something very frustrating. Teams often knew exactly what to do, but execution was very painful and slow.

0:57 That insight led him to build Plurio, an AI agent that automates marketing workflows, detects signals earlier and faster and lets teams act days faster uh which in some cases can change the customer acquisition costs by 40%.

1:19 Today we will talk about agentic workflows, AI-first teams, and why marketing operations might soon be run more by machines than humans. Seva, welcome to Exit or Die.

1:35 Um, hey Peter, thanks for having me here. Pleasure to talk to you.

1:40 Perfect. Seva. Uh, great to have you here as a guest. I would say we directly start into the topic. You spent nearly two decades building a marketing agency with over 100 people and hundreds of projects. What was the moment when you realized that the real bottleneck was not strategy, it was execution speed.

2:09 So actually there were two steps like first after the agency we launched uh marketing data platform. I thought that uh the bottleneck is in uh collecting the right data. So once you have actual full-funnel data and attribution and understand the true performance of each of your channel campaign, creative adset, product, country, everything. With the right data, people will for sure make better decisions and uh know how to manage their ads and user acquisition uh in general. That happened to some extent but then I I I literally saw the next problem and next bottleneck there. Some of our customers they actually spend like 4 hours per day looking at dashboards analyzing what's going on there and changing some something in uh Google ads meta ads and so on. Uh that was literally like half of their job uh to babysit uh hundreds or thousands of ads running at the same time. So I was hoping it become possible someday. Uh but last year agentic AI came uh Cursor Claude Code uh um new models and that was like an aha moment. Okay. Now now we can do I'm not sure how exactly yet. That was an exploration process. Uh but I saw that okay like it's possible to automate at least that part of work completely.

3:44 M yeah you already mentioned the dashboards. So most marketing teams still rely on dashboards, spreadsheets and manual processes. Why do you believe this model is fundamentally broken?

3:59 um professionals building dashboards uh they truly try to make it actionable so I can just take a look see like green red yellow signs and know what to do but that's not how it work in reality in reality you see that this campaign is doing well maybe we should scale it but before that we need to check like what's going on with the adsets uh inside that campaign. What is going on with creatives? Is something started to burn out? Is it an outlier with just from like it's in green because of just one uh sale and simple action converts to like 5 minutes on clicking through your dashboard and to double check things uh to like to make to make sure with agents agents can do all of that for you and like double check things and say like I am sure but this is the right action. Also there's the second part of this dashboard usage like yes there are a lot of regular tasks like optimize budget uh reallocate something somewhere that's regular things uh and we can talk about automation of that but uh then there is like an explorative creative part of the process like if you oversee hundreds of thousands millions of dollars in ad spend you'll will have some intuition about like which kind of targetings work better in general or in specific cases which funnels or like bring in higher quality lead types or customers with higher LTV something along those lines and say okay like can I check the data like behind behind this idea what's actually going on there and with dashboards some things you can do like click click filter, but only things that were specifically built in and designed by somebody who who built the dashboards. With agents, you can ask it like, "Hey, I I I want you to check like this idea, what's actually going on across like different types of regional targetings." Uh the agent will do that. Uh so I see such a relief in the eyes uh of heads of performance marketing when they realize they don't need to put a task. They don't need to wait for a week for somebody to build a custom report for them. They can just like go with the flow with their agent, talk to it for like 10 minutes, 30 minutes, 3 hours to explore everything they wanted to explore by code but couldn't.

6:43 Yeah. So the workflow is way smoother than before. You save a lot of time and you already mentioned uh AI tools and um what is the difference between AI tools and AI-first teams.

7:03 This is a huge topic. Um AI tools first of all like AI tools are evolving. Everybody knows ChatGPT just chat to talk to. Then there was like an era of uh projects like advanced use of AI meant that you created a lot of projects uploaded like context and instruction to it and you have separate project and separate maybe thread for each of your questions. So the agent has the context there. then became an era of uh Cursor or Cursor-like tools where you store and reuse your uh context for across time and build on top of that. So you ask uh your agent to to hey go use our project context, business context, product description, marketing funnels and now like create a new landing page for me, landing page structure or something and over time you build more and more context and uh the better outcomes you can get from uh from the AI uh and you can still track like what's exactly is going on, what is doing, what's changing, what produced the next error is uh like generally speaking right it's like Claude Code style uh work where you don't look at what it's doing don't look at uh even at the files and structure it's creating you just like say you just ask your agent like hey go do this thing and by the way like you know where where the context is start and it will like go for for for minutes and uh make it over

8:50 Yeah. So what do you think? So what actually changes inside a company when AI agents start handling the operational workload?

9:01 Um so we've been ourselves as an organization on this path for over a year now. I mean like actively retooling and changing processes and team structure and everything. And step one, we used to have a role of like sales assistant, somebody who collects data to prepare for sales call, updates, things after after calls and so on. Uh that role is gone. Now like everything except actual sales calls is uh done with uh AI for inside Cursor inside Claude Code or now like we play with OpenClaw a little bit. Step two again with examples from the sales team. Uh we used to have two account executives me my co-founder and two account executives. one of them uh decided to leave at some point and like do we actually need another salesperson like if offers preparation uh meeting follow-ups research of everything like we've built a system where every chat every email every call transcript every document related to a customer is stored inside our like company like second brain. So any kind of preparation takes like one prompt and 90 seconds. So like customer lead qualification like whatever you can imagine repetitive work except for actual calls. So now instead of having multiple people in same roles and uh assistants we have fewer people in the same role and also their work transitions changes first a little bit then a lot into people managing those agents and and like it sounds like a fancy word like managing agents. What exactly is it does it mean? It's really simple. uh you just like ask your Cursor Claude Code Plurio like the agent we build for marketers to do something it does maybe with some errors or you don't like the result you ask it like hey next time take into account this and this it's it produces the result you want yeah congratulations you are managing an agent you're improving it over time just by using it and giving a feedback uh that's it so so now all roles slightly transition like more and more into uh this flow where instead of creating a task to another person in the company delegating to someone they just ask an agent and if something goes wrong they update the instructions. I believe this is how things will play out for most roles in most organizations over the next uh couple of years. less and less assistants, less and less repetitive people in the same role and more and more agents workflows and steering of those into the uh right direction. But that's just layer one.

12:11 Exactly. Yeah. Yeah. Yeah. I mean uh it's the the most important thing or or the the best thing is to to is is the the the execution part, right? because you you will be faster, you save more time, you save money, uh teams will be smaller than before. So I think that's that's the the the core of AI agents, right? For me like I'm a AI B2B SaaS founder. For me there's it's a huge question where this ends like is how the role of B2B software companies changes what do they actually provide that in-house teams with Claude Code can do on their own and I think it will gradually move to much more powerful and sophisticated tools. So yes, you can vibe code like a analytical dashboard for for your ads data. Uh yes uh you can uh run your Claude Code on your uh analytical data to answer some questions. But let's say you want to run agents that build ML models that create automated rules to start scaling and descaling creatives at the right moment of time with the highest expected uh value from those actions. Yeah. Okay. Uh can I vibe code the simple version of that? Probably yes. Can I do it accurately at hundred thousands or millions of dollars of ad spend? Probably not. Next example is taking more responsibility. Like previously the whole s industry said like hey this is a tool pay $20 per month per seat. Uh use it be happy. It's your responsibility what you do. We just make sure it works and doesn't break uh too often. Like that's not a value anymore. If if people need tools they just like ask their Claude Codes to build something uh for them to some extent. Of course there are like complex systems and so on. Uh but what I want to highlight the real path to generate value for your customers now is taking responsibility for end to end workflows. So it's not like we're providing a tool that allows you to explore your ads performance. No, we are providing you tools, dashboards, workflows and agents and oversight of those agents that actually manage bidding and budgets and ad optimization better than you would with Claude Code and that's a new kind of value proposition. uh so we as an as a product can actually take a whole workflow from from uh your organization and be responsible and and improve it and that's not something SaaS founders are ready to do I think.

15:16 Yeah. Yeah. I mean let's talk directly more about Plurio. So you mentioned already uh a few things. Is that exactly what Plurio does instead of traditional marketing analytic tools can or is there more on on that?

15:35 So what we provide right now is like let's say you're head of media buying team performance marketing team uh and you manage uh maybe an agency a few in-house employees and a million dollars in ad spend. What we provide uh is first it's a chat with access to your data platform and to your business context that acts almost like a junior analyst and you can ask questions, get results, charts, reports, everything on the fly. It's not 100% accurate, but it's fast and it covers like 90% of your requests. Uh so that's just a value that unlocks your like possibilities truly understand what's going on like in your uh ads and sales funnels and performance and everything. Second part workflows. So instead of looking at dashboards and making changes yourself uh you ask an agent like hey do weekly budget optimization inside meta. It does a work for like 10 minutes, 20 minutes, 30 minutes like if it's a huge account and gives you like a prioritized list what exactly needs to be done and why with all those checks and double checks and you just got to click approve or change instructions. So instead of looking at dashboards for 4 hours per day, you can look at this thing for like 30 minutes once uh and that's it. From there we say like okay if you like this workflow and you like its recommendations let's ask our agent to create an automation that will do that uh those changes automatically and you'll still can like confirm the each each action and once you're tired of that because it just works you can put in full auto mode. So, and it it it actually turned out to be really important to make that transition gradual because people also need to learn and adapt how the system works to be able to to to rely on it. So, like it's just a chat play with it. It's just a workflow. It's it recommends you something like give you a feedback and know like it's a simple no not that simple but it's a understandable automation and this is like explanation for for each decision. If you agree, approve. If you disagree, give give feedback. And then like after a few weeks uh of moving through that learning process, people can actually understand what's going on and trust it and say like, "Okay, yolo mode. Uh I like all their actions. Uh I don't want to oversee them anymore." So this is where we are today. Like today with our product we can gradually automate the whole ad management part like budgets, bids, stopping, relaunching things like the most boring and time-consuming part and then we're moving to like next uh modules which are even more exciting.

18:48 Yeah. Uh I mean we we already talked about uh the speed, right? So you you mentioned that the AI agent is is faster and you can acting even hours earlier and that means also that can change the customer acquisition cost dramatically right. So Oh yes absolutely. Why is speed such a decisive factor in performance marketing today?

19:18 performance marketing especially in Meta, Tik Tok and like those discovery platforms uh it's actually a huge auctioned marketplace where you compete with thousands of advertisers including other categories for the same people and their like screen time. So even today uh like only 5% of your uh creatives actually will become winners. So only 5% of your creatives will actually make you money and make most and most of your money will come from from just a few of them and then on top of that each creative has its own lifetime like uh you launch it the platforms learns uh on it uh then you start to scale it uh it generates value for you then it's uh like people starts to get tired of it creative fatigue kicks in uh the performance gradually moves uh down, but then you can relaunch it for separate audience or later or somewhere like for for for another country, another product. And that happens at the scale of thousands of creatives at the same time. And they change like you have new metrics every day, sometimes every every hour. And we work specifically uh with the segment of companies that spend money on digital ads and generate leads, signups, trials. What that means is that there is a delay like you show ads today, you have leads today, you'll have sales in a week, you'll you'll know LTV uh in a month. So what that introduces is like okay like this creative actually like the performance is slightly going down but maybe those leads will convert to paying customers we should keep it like that's the logic inside people managing this thing uh manually because they always have to wait and wait and wait to see the actual performance while agents uh they can build forecasts they can uh check what's the expected value of this action now comp uh compared to the one year of previous data like h how would that play out previously on your on your account? Uh so while people wait they spend money on ads inefficiently when they wait to scale they they they lose new leads paying customers and revenue. uh when they optimize for like intermediate metrics instead of LTV forecasts they attract wrong customers that that will churn soon all of that like each of that thing can add like a few% of efficiency here there but like it accumulates into huge numbers uh like 20 30 40% of customer acquisition cost uh overpaid or revenue left on the table Like that's real and that's only the beginning. Like the more and more people companies uh will use AI and automation to take actions more decisively and more accurately. They will just outbid you in those auctions and you'll and you won't have an access to the audience at all like in a year or two.

22:42 Yeah. I mean uh one good example from from my side is the the warm- up phase of an of an ad, right? So you have to wait uh the warm up is going for the ad number one for example slower than the other one for ad number two, right? And then you have to check manually. Okay, maybe what's what was the reason for that? And agents are faster, right? because they can analyze that and can maybe work on that and so yeah that's that's that's very interesting and yeah currently I love that specific example uh because uh like that was one of the first cases we wanted to automate completely uh and I started running agents on the data and saying like okay if the new creative is this in this warm-up phase didn't produce any leads should we at what point should we stop it uh at $20 of ad spend no we killed too many winners future at $50. Still no, we'll still kill some winners and even just one winner will will outpay the whole ad spend for tens tens of them. At at 100, we still have a few winners even even there. Uh so what actually worked in this case is to ask agent to run ML models with multiple parameters to understand the combination of ad spend and uh CPC and CPM at different levels of the ad spend that will definitely say like okay like the chances of this creative actually becoming a winner is like near zero. But those learning phases like uh they are crazy and you expect people to like make make that decisions on the fly like no it's not possible.

24:29 Yeah. Yeah. I mean that's the best point. Um we are hearing more about agentic workflows, autonomous AI systems. Do you think are we moving toward a world where AI agents coordinate entire marketing operations?

24:49 I think the end game here is full fully autonomous performance marketing teams. I mean like completely we yes we start with like the datadriven part where you have numbers decisions automation but then we move to creatives analysis like we can actually uh analyze like we're building this now uh where we'll analyze your creatives your competitor's creatives by hundreds of parameters uh figure out like the true drivers of what exactly works better in terms of catching attention and converting to high value high LTV customers and start generating like next best creative uh formula. Uh so now you have this part automated and that's like another third of the work just to like explore things and uh come up with the right ideas. Uh creative generation is moving fast as well. Same thing we'll do with uh all the like running ads playbook uh like what is the structure of our account? What kind of targetings do we use? Uh what's um how do we split between like experimental campaigns and evergreen campaigns? All of that can be uh analyzed by swarms of agents every night to extract insights that could improve your playbook. And then that playbook could be applied to your accounts first to control uh first to suggest then to control then to implement. So we're moving to a world where the whole cycle of performance marketing uh is uh closed and automated and then each part of that uh system is improved autonomously over time with more and more data from more and more customers. So at some point we'll end in the world where AI agents that are built and improved by companies with access to a lot of data will be so much better than uh what you can uh come up with on your own at least in performance marketing side like I don't have insights about like the brand marketing viral marketing but like in my area of expertise in performance marketing 99% could could be automated. I don't think it's it will happen in one year. Maybe two three like okay between one and five is a reasonable expectation.

27:26 Yeah. And that's also very fast. And now we we know that AI agents can do nearly everything. So what about the skills on the marketer side? So what skills should marketers focus on if AI is handling the execution layer?

27:47 I think the first like milestone is to excel at uh managing agents uh because like it's like new computer literacy, internet literacy like like at some 20 years ago people learned how to Google in the right way then how to use ChatGPT in the right way and now it's time to learn how to maintain your own agents so that they like solve things for and and and like again fancy words really simple thing uh just give your agent the feedback and improve instructions gradually that's it that's the whole like agent uh training uh thing um that's step number one step number two is like a meta uh skill of learning with AI like I remember times where like to truly grasp uh like a new type of skill or new area or new something it took like half a year to read books to Google and uh read papers uh to maybe talk to other people about it to do your own experiments if you can with code if you can like at least like with spreadsheets or something uh to play with it a little bit. It's more like uh learning why you when you use it, right? So that's that's the exact I I just want to highlight that like I I believe it's extremely important. Uh yes, you can just uh like wipe work to the result like just like don't make mistakes continue until it's ready. Uh but still at least in the next few years maybe more uh if you want to do a great not not just good job but great job you need to understand the key concepts the criteria of success the choices between different classes like how this uh work can can be done. Yes, agents will do that, but like you'll still you're you're still the ultimate responsible person for for what's going on there. It's like a moving goalpost. Things that could be done just one-shotted with uh generic agents like it's not uh value anymore. Anybody can do it. Uh so where can you actually add value in your work? It's an understanding next layer of the technology of work like workflow efficiency. So you need to learn fast and you'll need to learn fast with AI using AI. And this is why I started like 10 years ago it took like a few months to really grasp a new thing if you spend a few hours at least a few hours per week on it. uh now it should be done like in hours and days by asking the right questions by asking AI to do experiments and MVPs and playgrounds for you. Uh and then like say okay I explored this options how to how to automate this uh this is the right one it's working uh yeah let's move to the next thing.

31:03 Yeah. Right. I mean every company uh now says it's AI powered, right? So when someone is using chat GPT in in in the team then companies say hey we are AI powered. So from your perspective what's the biggest mistake uh teams make when adopting AI?

31:35 I think the biggest mistake is to think that if you use ChatGPT like that's enough uh that was enough year and a half ago year ago like this wave of adoption uh is about making AI systems AI agents actually automate actually like take some of regular workflows it's a meeting preparation it's a lead qualification It's competitor analysis, it's creative analysis, it's uh reading your dashboards for you. So if you are not automating at least simple repetitive uh parts of the work uh then like uh then you're like behind the the goalpost. You can do that with Claude. It's actually like powerful enough but really but has really simple interface. You can do that with Cursor or Claude Code. Okay, they're like scarier because they were designed for developers. Uh but they're actually like generic agents. Uh in marketing you can like in performance marketing you can do that with us with Plurio. We'll give you interface workflows data all of that help you make it work and uh teach you how to improve it over time and like deliver new modules over time. But like the biggest take here is moving from chatting and like using as an advanced Google to actually automate a big chunk of work. That's today. Uh tomorrow I can like give you a glimpse into what what's coming tomorrow.

33:08 Very very interesting insights. But what do you think? How do you move a company from experimentation to truly becoming AI native?

33:27 I've made that transition like a year ago. At that time, we were like a 30 person organization. It started with uh a few things. Uh first me as a founder, I started to play with AIS a lot like at least a few hours a day. Lovable Replit Cursor vibe coding some things and then move doing like actual work uh like non non-developer stuff as well any kind of knowledge work creating a pitch deck collecting the context about your company and so on so if founder or C-level doesn't do that there them themselves it will be really hard to change the organization and this is why so many of my friends stay up till 3:00 a.m. playing with OpenClaw or something. It's not the most productive uh way of spending time, but it's it's needed to truly understand uh like what what is it and how to how to apply it across your organization. Step two, uh we've done like two times a week on meetings with the leadership team and with uh the whole team. uh we spent like 15 minutes sharing uh our use cases like what's the new thing we were able to do with this tool or that tool uh with agentic. Third we've created the thing called uh shared workspace nowadays I think it's called like company's second brain uh so everybody I mean absolutely everybody in the organization has access to the same file file structure across all departments. It's like a Notion or like a Google Drive but uh with files on your computer synced between computers and what and what it does is anybody can run an agent Cursor Claude like whatever any kind of agent uh to and and this agent will have access to data and instructions and everything from everybody on the team. I saw a lot of organizations trying to to implement AI within one department. Okay, this one department got advanced. Everybody else is still in the stone age. So like truly truly truly strongly recommend uh to do some kind of shared workspace second a single point of entry to your AI stuff across the whole org the key was to convert each key employee into like having its own aha moment like okay I tried uh to structure 100 uh resumes for for for this position and they I did it for me with my criteria and we iterated a little bit and now like I'm happy like okay this person is converted uh they will try uh on other things with marketers like okay I tried to tune this uh optimization workflow and it actually started giving me like uh the right recommendations now I see how it saves a ton of boring and time-consuming onboarding work and frees up for front line. So once everyone has that aha moment and use the shared workspace and leadership actually understand with like firsthand how that how that works and play with it like every every day. uh then it starts to compound because everybody in our organization contributes to this like discovers new workflows, new automations, new way of doing things uh and like it becomes like really like a a slope to go faster and faster and faster.

37:19 Yeah. I mean that's definitely easier in in in in smaller teams in smaller companies when we look to enterprises then it's definitely something where the C-level um has to move forward and take everyone with them to really go into the future right because when you lose someone in the whole organization then you have I don't know you have a leak Yeah. Right. Because then this one employee is in the stone age, right? So, yeah.

37:51 Yeah. I have a friend, he's a founder, co-founder of like 300 person company, still not that big. And he's like more in the strategic role than in operations. And he was wondering like, hey, do I really need to do this vibe coding thing? Do I really need to play with it myself? As long as I understand like what's going on, sure, that's enough. Uh he was like talking about it for for a few weeks and then one day he like tried it himself and then he prioritize resources a little bit better in the like some some uh um task list that was actually used by half of the organization. uh it's just like old style and hardcoded and he's like wow his expectations changed his understanding of what's possible changed his take like he was not even in operational role and he changed his mind like okay I need to be in this myself because that changes everything about my organization and this is the same thing I see with our customers with uh marketing teams typically we have like some champion who's excited about this thing and then we see a team of uh media buyers, marketers, they're like like yeah another AI tool uh selling us the future. But like you need to you need leadership to to to not to believe to try it themselves to not to take it from others' words uh and then to like help everyone in the or to get through its own aha moment uh with uh agentic this is how you do transitions.

39:41 Yeah. Yeah. Perfect. Perfect. I mean that's the the perfect explanation for that. Um yeah Seva now we are coming to maybe the fastest part of this uh podcast the exit or die rapid fire. I'm not sure if we are faster than AI agents. Um but it's a quick fire round. Five questions from my side and uh one sentence answers from your side please. Are you ready?

40:12 Yes.

40:13 Perfect. My number one question is um what do you think is the most overrated AI buzz word uh right now?

40:23 Oh, that's uh the word agent for sure. Uh because uh it it it makes it sound fancy when in reality it's it's a relatively simple thing to to understand and to to to play with it, but it's truly it's truly fundamental and it's so overhyped at the same time.

40:44 Yeah. Then my second question is um one marketing metric uh founders obsess over too much cost per lead sometimes cost per customer. I don't know why is that but in the western world people still ask their marketing teams to focus on uh on cost per lead and uh think that like other teams will handle like the conversion to paying customers and so on. That's not how that that's not how that works. Uh you need marketing to optimize for highest uh for for ROAS and highest LTV customers. Otherwise, you're attracting the wrong audience.

41:26 Mhm. Very interesting. One thing AI will replace in marketing first.

41:33 Ad management will will be automated first. We're doing it right now like that's happening today.

41:41 Then one thing AI will never replace

41:41 taste, creativity, uh deep understanding of things.

41:49 Mhm. Yeah. There are humans uh needed in in that kind of spaces. And yeah, the last one would be if you were starting another company tomorrow, what would you build?

42:03 I mean, we just made a pivot a few months ago. I really like what I'm building today. uh like we want to make in the end we want to make fully autonomous system that does best performance marketing in the world.

42:19 Yeah. Perfect. Perfect. Yeah we are coming nearly to an end of of the podcast. Um so far very interesting uh insights a real really really really great journey uh what you what what you have and what you had in the past. If you could give one advice to founders trying to build AI-first organizations what would it be?

42:47 Go try and do it yourself. Uh everything else will will follow. If you're a solo founder, like you have no other choice, but you if you're a founder of an existing company, you have a temptation to delegate this thing. Uh that's not how it works.

43:05 Yeah. Yeah. Perfect. Yeah. Seva, that's it for this episode of Exit or Die. Um again big thanks to you Seva for showing us how marketing teams might evolve when AI becomes the operating system. Do you want to tell our audience to the end? Um anything what you want to mention? Maybe just highlight that yeah we just rebranded to Plurio we raised uh another 3.5 million in uh seed round uh we are out of the beta phase so like we're ready to automate your performance marketing with AI so I'd be happy to talk to you personally if you manage like hundreds of thousands or millions of dollars in ad spend we have cool stuff for

44:04 Yeah, perfect. Yeah, and my last message is the real shift is not just smarter tools, it's faster decisions. That's it with this episode. See you soon and thank you Seva for joining me.

44:15 Thank you. Bye.