From Puppies To Programmatic To AI Agents

0:12Welcome to Ad Exchanger Talks, the podcast devoted to examining the issues and trends in advertising and marketing technology that matter most to you. Today's episode is sponsored by Basis, the leading intelligent operating system for autonomous advertising. Its enterprise AI solution transforms campaign briefs into strategies and media plans that integrate directly into omnichannel activation. You can learn more at basis.com.

0:50I'm Alison Schiff, editor-in-chief of Ad Exchanger. And this week, I'm talking with Tanya Mamica, COO of Kova.ai, about a problem that programmatic has never quite solved. The ad transaction is automated, but media buyers still spend hours checking campaigns, reconciling numbers, and chasing down issues across platforms. Kova is building AI agents to handle that work within the tools that media buyers already use. We'll talk about how agents can earn trust, where people need to stay in control, and who's responsible when an AI agent gets it wrong. But the benefits are real. One agency and a pilot with Kova reported saving roughly 500 hours a month.

1:30It's not just about saving time, though. It's about catching mistakes before they go live, doing checks that might otherwise get skipped, and getting clients' answers faster. But before we get started, I'd like to take a moment to tell you about our upcoming Top Women in Media and Ad Tech event. Join us on the evening of November 3rd in New York City as we celebrate the leaders, innovators, and builders moving advertising and media technology forward, including Tanya. She's one of our honorees. The awards gala will take place at the Lighthouse at Chelsea Piers, honoring an extraordinary group of women across AI, commerce media, privacy, programmatic, measurement, and CTV.

2:08I'd love to see you there, and I've got a promo code for you. Use the code TWTALKS for a discount on your ticket when you register. Hey, Tanya, welcome to the podcast. Thank you. Excited to be here. So rather than asking you to share one thing about yourself that not a lot of other people already know, I'm going to steal your thunder and start off with your background because it's an interesting path. So you got your Bachelor of Science degree from the University of Melbourne in animal physiology, and then you worked with animals for years, like maybe over a decade. You were a business manager at an animal hospital in Victoria, Australia.

2:47You were director of the Australian Veterinary Business Association. You co-founded a startup in 2015, which was like a marketplace for vet management consulting, connecting vet practice owners with admin experts around the world. And then you made a turn. 2018, you co-founded Martin and you were the CEO too. And Martin is something I think a lot of our listeners would recognize. It's an AI-powered media buying and measurement platform with a focus on business outcomes, incrementality, optimization. Martin was acquired by Pubmatic in 2022. It became the backbone of Activate, which is Pubmatic's sorta SPO solution that goes direct to buyers.

3:30AdExchanger covered that whole journey. My question, I do have one, is how did you go from puppies to programmatic? And I am a little proud of myself for that alliteration, by the way. What happened? I mean, isn't it obvious? So many people, I feel like, just land in ad tech through, you know, various crazy paths. But yeah, I feel like mine was a little bit of an interesting personal story as well. It's pretty funny. So my now husband and co-founder, James, and I were living in Australia. And we decided that before we settled down and started a family, we wanted to travel together. So we both quit our jobs and took our savings and we got work visas for Canada.

4:16So we plan to travel from Patagonia up toward Canada and just see how long the savings could last us. And then when we ran out, we could either go to Canada to work or come back to Australia. And so this part won't come as a surprise. Our savings ran out in San Francisco of all places. No doubt. We maxed out our credit cards and we were staying in this horrible hotel in Tenderloin. It literally had bedbugs. It was awful. We'd look at the Super 8 across the road and dream of staying there. So we knew it was time to start working. James is an AI engineer, so he got a job at Tesla and we moved out to Reno, Nevada, which is where the Tesla giga factory is.

5:03So I didn't have a work visa and I started kind of tinkering on my own side project, building a website for veterinary business consulting, as you said. After about 12 months, I applied to some startup accelerators and we got a call from Ocean, which is a faith based accelerator in Cincinnati. We're not religious and we had no idea where Cincinnati was, but like could not. A lot of Americans don't know either, Tanya, don't worry. Sorry to all the people from Ohio. Whoops. Right. But we said yes. And we agreed to move to Cincinnati for a 12-week program. So as soon as we got there, they were like, listen, we really like you guys, but your business idea sucks.

5:53You need to find something else to build. on, by the way, Demo Day, which is, you know, where you present your pitch, your startup to a room of, I don't know, like 500 investors or something is in 10 weeks. So we basically had to start from scratch. It could no longer be a veterinary business consulting startup. And we somehow landed on building a chatbot for SMBs to help them with their marketing. So we presented that at Demo Day and the feedback that we got from investors this time was, we like you. Great idea so that was progress but you need to be selling to media agencies and brands not smbs to which we replied more or less what's a media agency

6:40luckily one of our advisors was the ceo of one so he took us in and we worked out of their offices with the goal of identifying a problem that the agency has that kind of some version of our tech that we had built up until that point could solve. So it was there that we came across the wonderful world of programmatic. And so what started as a kind of one travel adventure turned into, what is it now, eight or nine years in Antec. And that startup eventually became Martin, as you said, the DSP that was acquired by Pubmatic. Was Martin someone you knew? Oh, my God. No. Well, this is because it was like the whole startup pivot thing.

7:24So in this kind of crazy process of, okay, we need to completely pivot our startup, our investors are not liking this at all. But we do need to present something in 10 weeks. We decided that we needed to rename the product as well. And chatbots at the time had, you know, human sounding names. So we got out a book of baby names and we just started going through it. That's amazing. So Martin was your baby. Martin was our baby. We kind of said, okay, if you like squint, it sounds a little bit like marketing intelligence. So maybe let's go with Martin. Yeah. There is an A and an I in Martin also. Yeah.

8:04Yeah. There you go. So you were the, once it all worked out and Martin was a thing, you were acquired by Pubmatic, like I said, and then you were the VP of Activate at Pubmatic for around three and a half years. And then you and your co-founders, all people that you knew from Martin and Pubmatic, including Kyle Dozman, the former CRO of Pubmatic. You co-founded Kova.ai, Kova, the startup that's building AI agents for media buyers. And we're going to get into what that means and what that looks like in a little bit, because I know just saying building AI agents for media buyers is a little reductive.

8:42It's kind of a shorthand. But what's the origin story for that one? Like what were you seeing and experiencing with sellers and buyers because it's activate at Pubmatic that made you guys say like, huh, like we think it's a good time to strike out on our own and there's a need for this. Yeah. So we when we finished up at Pubmatic, we kind of had a three year contract. My so James and I kind of briefly explored pursuing other interests in completely different fields. It was a very brief pursuit. I think after less than a month of thinking about what else we'd like to do, we came to the conclusion that the opportunity and challenge that AI and ad tech presents right now is too big to ignore and that, you know, based on the experience and skill sets we developed over the past eight years, it would be silly to jump ship now.

9:35The opportunity or insight as we saw it and continue to see it is that AI is being built into the platforms, right? So Mente is building AI and getting really good at optimizing Mente, and the trade desk is really good at optimizing the trade desk and so on. But the majority of the work of a media buyer lives between the platforms. So running reports across different platforms, pivoting Excel spreadsheets, checking discrepancies, reconciling, troubleshooting, a lot of explaining and coordinating. And that was not being worked on. And I think aside from that, I've always said that one of my favorite things about this industry is the people.

10:19So we, this time, like you said, are really lucky to be joined by two co-founders, people who've worked with in the past. So that's also really fun as well. What I think is interesting about you guys is that the pitch isn't like AI is going to reinvent media buying. It's more like here's AI for tasks that nobody really wants to do. Although I want to put a pin in that statement because some of those tasks are important to train a brain so that you even know what you're doing and can think strategically. But yeah, I mean, cross-platform pacing checks and looking at the discrepancies between a DSP and the ad server, doing general campaign QA, trying to figure out why a line item isn't delivering.

11:08I mean, there are better uses of people's time. But then again, like the promise of programmatic was that a lot of this stuff would be automated. So like what happened there? Why didn't it get automated? Like why was it so messed up when the promise was so big? Yeah, I don't think that it was messed up. I think that the automation, I think when we think about programmatic, happens within the transaction itself, right? So real-time bidding absolutely automated that aspect of media buying and selling. And I think AdCP does that in the agentic world. What we chose to focus on and what COVID does is completely different.

11:52And like you said, it's the unsexy stuff that lives between the platforms, but that is absolutely necessary. Yeah, it's a good question of why that part wasn't automated. I don't know, you'll have to ask someone who's been in Antic for longer than I have when it was being designed. It makes sense to me, though, because you can't think of everything at the beginning. And so there was this big idea, let's automate the transaction. And all the ancillary stuff just fell by the wayside. People picks it up. They already knew how to use spreadsheets. And so all of that manual labor just cropped up naturally around it.

12:32And then once you do something a certain way, it's really hard to change your behavior. It becomes entrenched. And I also think that once something is an annoying task and one person has to do it, they want to haze other people and make those other people also have to do it. Like junior media buyers shouldn't have to or shouldn't get away with not having to do all of these things that I had to do when I was coming up. Yeah, absolutely. And I think also the, you know, the AI tech in the past two years or so has gotten so much better, obviously. So I think some of the stuff just couldn't be automated at the time.

13:09let's talk about how kova works like we'll do the long hand instead of the shorthand so i'm a media buyer at an independent agency or or wherever um at a brand how do i use kova what am i handing off what am i still doing myself how does my day change after my first pot seven pots of coffee, whatever. Yeah, great question. So maybe a good way to think about it is exactly how we onboard clients. So the first thing that a media buyer does when they are logging into Koba for the first time is connect their platforms. So we connect at the moment to about 50 or so tools that media buyers use. So that's DSPs.

13:56And we talk obviously a lot about programmatic. That's where our background is, but COVID does connect across search and social as well, verification and measurement partners, and then also productivity tools like OneDrive, Slack, Teams, email, Asana, monday.com, and so on. And also databases like Snowflake and BigQuery. So as a media buyer logging in for the first time, you would just connect the platforms that you use. And our goal is really to make that list exhaustive so that from a media buyer's perspective, everything, every tool that you touch on a daily, weekly, whatever basis to do your job is available and connected in Cova.

14:37So once the agent is connected, it fits into how the team already works. And this is something that was important to us and has kind of formed the thesis of everything that we built. There's no new dashboard and no process or overhaul. So we're not building any new ad systems. We're not building a bidder or a data platform or a curation platform or ad CP or anything like that. Our agents simply help the team work more efficiently across and within the tools that they're already using. So the agent, the media buyer has now connected their platforms. And the first thing that they usually we see them do or advise them to do is just to start off with some really simple, we call them daily or weekly briefings.

15:24So they are kind of non-customizable reports that get delivered every day, either to their email or Slack or Teams or whatever it is with, you know, showing what has happened in the platforms over the past 24 hours. The cool thing is that they can start to interact with it. And because it is a bit of a shift, right, from how media buyers typically work. So it's, I think AI in general has a bit of a cold start problem, right? Like you get into an AI platform and you can do anything, but like, what exactly can you do? So these daily briefings just show you a little bit of what the agent can do, and you can start to interact with it and ask questions and eventually have it take action.

16:06So now you're connected, you have your daily briefings, and then once you're a little bit more comfortable, you can start to schedule some tasks or delegate more complex workflows to Kofa. So that's where that delegation or automation part of the role comes in. And we've seen over time, and it probably takes, I'd say, three to four weeks minimum until media buyers get really comfortable delegating complex tasks, right? So where they're stringing things across or creating agents across different platforms, different spreadsheets, and maybe even their email in order to be able to kind of delegate something to the agent.

16:53Is there anything media buyers are weirdly precious about that they just don't want to give up that you've found now that you've been out there working with buyers for a while what don't they want to give up uh well i can tell you what they definitely want to give up and that's um that's you know inserting third-party pixels into 150 creatives at 9 p.m on a friday it's so fun though so it's funny i always think about that when you know people say well don't junior media buyers need you know some training like yeah they need trainings But like you do that once, you don't need to do that for two years and it comes up often.

17:39So that's the kind of stuff that they are very happy to give up. I think to answer your question more directly, in terms of things that they're a little bit more hesitant, it's about building trust with the agent, right? So that's why I think some of those briefings and getting reporting and checking it against the platform numbers is important because it builds trust. Once they build that trust, they start making changes and using COVID to actually execute in the platforms. I think the bigger that that change is, the more risky it is, right? Like whether it's setting up a campaign with a large budget or ensuring that maybe scheduling a task that your ads are paused at, you know, 11.59 p.m.

18:25on a Friday and actually feeling like you need to go back in the platform to check it because the, I don't know, client's offer runs out that weekend or whatever it may be. So I think just the more riskier the task gets, until they've built some level of trust, they are not ready to hand it off, which I think is particularly understandable as, you know, someone who's trafficked campaigns before I wouldn't either. It does raise an important question though, which is who's accountable when an AI agent makes a mistake because it will. People make mistakes constantly. They're wrong decimals, a missed flight date, the wrong creative goes live.

19:03Agents are far from perfect too. I see that when I interact with Claude and ask for, hey, can you briefly summarize this story that I wrote so I can plop this one-liner into WordPress because that is one of my tedious tasks and the line that comes back is just what? Although that's a little more, I don't know, probabilistic than a wrong decimal point. I know you guys place guardrails outside the agent. How does that work exactly? And yeah, who is accountable when an agent screws up? Yeah. Well, first, I love that you said that humans make mistakes too, because I feel like this is very similar to self-driving cars.

19:46Like we hear about the few mistakes that are made, but we don't actually see or measure the thousands of times that the self-driving car has avoided crashes far more effectively than a human driver. But you're right. So we think about it in three layers. So the first is guardrails. And as you said, the important thing is that they sit outside the agent. So the agent shouldn't be able to raise its own budget limit or approve its own change. So those limits are set by people and the agent works inside them. We call them playbooks and they exist at kind of different levels. So whether they're at campaign level, platform, brand agency level, but they are essentially the rules of how you want the agent to work and also specifics about the agency or that brand client how they want to run the media.

20:43So the second is matching autonomy to risk. So pausing a line item that isn't delivering is cheap and easy to undo. Moving a big chunk of budget is not. So the thing that is reversible and low risk can happen with less friction and that the things that are expensive or hard to undo go to a person for approval, which can just happen with one click in Slack or Teams or whatever. So the third is the audit trail. So, you know, something that is very readable, not a log file, but a plain language record of what changed when and why. And one of the jobs that our clients set up is a weekly record of every change in the account so that when a client asks what changed, the answer already exists.

21:34so you know to answer the question of who is accountable I think same as with humans the agency is always accountable to the client and agents are a tool the agency uses the same way that it uses a DSP or any other platform one of our clients actually described this really well they said that Cova gave everyone in their portfolio an insurance policy because it reduces mistakes and increases consistency. And I think that's a great way to think about it. Agents can make you safer, not riskier, if they're set up properly. I think the fear and also just the question, who is accountable, comes from the fact that there's this assumption that everything will be handed over to an AI agent.

22:24And if that happens, then what if, you know, an agent has control over spend and then it all goes out the window? Every dollar is spent in an hour and it should have been, you know, spent over the course of a week. But people do do that, too. That has happened. Absolutely. Yeah. I mean, and I also think we're a long way from that. I don't know. Maybe technically not a long way. Right. But in terms of adoption and even the necessity to do that, I think we're a long away from that. So to give you a bit of an idea of kind of what we're seeing is every right action the CROVE agent takes needs to be approved.

23:07And we were planning to keep that in place for a long time. We've had some clients start to ask us to remove the necessity for that, as I mentioned, with some low risk actions. and we haven't even done it yet. So I think this world of like end-to-end, the agent is responsible for the budget and it's just, you know, optimizing and shifting and I don't think we're there. But to your point, yes, humans do that too and many times have I set up campaigns in a DSP that, you know, spend by 1 a.m. rather than pace evenly throughout the day. It's an oh shit moment. It happens. All right. Well, we're going to take a quick break.

23:55And when we're back, we're going to keep talking about AI. But of course. So stay with us.

24:09I'm Sarah Sluice, Editorial Director of AdExchanger. And with me today is Michael Olson, EVP of Client Development at Basis, which is supporting our podcast this month. We're going to talk about something that doesn't get enough attention, paying invoices. What is the cost of slow financial reconciliation and who is the most impacted? Hi, Sarah.

24:34Tanja Mimica: Thanks for having me. The cost is time, cash flow, and trust. When financial reconciliation takes too long, payments get held up and finance teams, whether they're in agencies or ad brands, are just chasing down numbers. People are spending a ridiculous amount of time figuring out why two systems do not match. Agencies probably feel it the most because they're stuck in the middle carrying the float. They have the client asking questions on one side and media partners on the other. The longer it takes, the more time everyone spends fixing the past instead of focusing on performance and health and growth of the business.

25:13So you outlined a very frustrating problem. I would not want to be that person chasing down invoices. Is this primarily a technology problem, a workflow problem, or a people problem?

25:26Tanja Mimica: It's all of the above, but I'd lean towards mostly a workflow problem. The industry definitely doesn't have a shortage of technology or data. We have a shortage of systems that actually work together. And when they don't, people end up filling in all the gaps manually. They end up downloading spreadsheets, comparing reports, sending emails, chasing down discrepancies. It's a mess. Technology is obviously part of that answer, but another piece of software by itself just simply doesn't fix it. It's imperative for technology to actualize omni-channel data monthly and push to reconciliation systems.

26:06Tanja Mimica: As an industry, we have to connect the tech stacks interoperably, end to end, and then allow the technology to do the heavy lifting to align the data between the contracts, media plans, media transactions, and the payments. Okay, so it sounds like technology and workflow can make this just a bit easier. What does automated reconciliation look like? A lot of automation talk right now. And if I'm thinking about the future agentic state, ideally, reconciliation just happens. AI is constantly looking at delivery, spend, invoices, payments, and then flagging issues as they happen. Instead of somebody finding them three weeks later, and in a lot of cases, it should be, the AI should be able to fix those issues without a person ever touching them with the correct harnesses in place.

26:59Tanja Mimica: people should only have to get involved when something unusual happens, which is the human oversight. That gets away from this huge month-end exercise that costs a lot of time and resources and much closer to knowing where you stand financially at any given time. This is the ideal state basis is going towards for our connection with Prisma by MediaOcean, a platform used by numerous agencies for financial reconciliation. Okay, so it sounds like by connecting those two different platforms, we can and adding agentic AI, we can make it a lot easier and people won't have to chase down invoices all the time like they used to.

27:37So thank you so much for sharing with us what that would look like. And thanks to Basis for supporting our podcast.

27:43Tanja Mimica: Thanks for having us, Sarah.

27:52All right, we're back. And I think congratulations are in order because you are one of Ad Exchanger's top women in media and ad tech honorees in the AI Media Innovator category. So congrats. It's an event that we host in November, celebrating the leaders shaping the industry, shaping the future of the industry. But I want to look back for a second. You started, Martin, like we were talking about during the first half in 2018. And now you're on your second ad tech company. I don't think I'm going out on a limb to say that ad tech has historically been a pretty male-dominated domain, especially at the top.

28:34What would you say has changed for women in leadership in this industry over that time, like almost a decade? And like what hasn't? What still needs to change? Yeah, I could talk about this for a long time because but I won't. But I don't know. I think a part of me, a part of me thinks that it doesn't matter, you know, whether you're a man or a woman. I think you need to do the job and you need to do it well. and you will be recognized for that. Now, I think that's, you know, coming from maybe some privileges that I have enjoyed that I know others haven't and, you know, maybe a mindset. But I think a lot has changed in the fact that we're talking a lot more and supporting younger women entering the workplace.

29:29I think in terms of what hasn't changed, I think that's a harder discussion and I think a lot of it rests not necessarily within, you know, the leadership of an ad tech company, but at the federal government level. And that part in terms of, you know, and maybe someone who's, you know, as someone who's not coming from the US can talk about that. I think in terms of paid parental leave, America is behind. I think in terms of, you know, child care, and obviously this is all relating to, you know, women who choose to have children, but I think it is about 75 % of women in the workforce. So I think those are some of the changes.

30:21And like I said, I can talk about this for a long time, but I won't. But I think there are some of the changes that have to happen for women to be truly supported and on an even footing when they do choose to have children to allow them to remain in the workforce. Because as you mentioned, that's where we see the drop off women in leadership. That just, it is more difficult to compete with men when you also have different expectations and no paid maternity leave and and no affordable childcare and all the rest. So some progress and some TBDs, but... Yes. Well, I will see you, by the way, in November at the gala.

31:05So that'll be fun. We haven't met in person before. And back to Kova. Yes. I don't know. I'm going to say it way cleanly from that, but I do agree. what would you say like a new agent has to prove before people trust it, right? I mean, is it just that it's not messing up or is there like some other unlock so that people do trust that the job is going to get done? Yeah, I think the trusting the job is going to get done is something that we've seen users be able to validate fairly easily, right? Because I think any task, if you just think about what media buyers do in the platforms, that they have a very specific task outside of platforms that they want to achieve.

32:01So let's say, you know, you mentioned discrepancies between, let's say, a DSB and a third party ad server. The task is well defined and it's easy to see whether it has been, whether the agent has performed it accurately or not, right? Same with setting up a media plan in the platform. You can always go back to that media plan and QA it and see how it's set up in the platform. So we generally see that that, I guess, trust or validation happens within the first week or two because it is very easy to check um yeah i mean i would imagine that some of it comes to when people see that they're saving time right i mean i i pretty readily handed my recordings of conversations for interviews to an ai to transcribe them for me But I was a little reticent at first because some of the output was messy and it couldn't understand accents.

33:06And I felt like transcribing it myself allowed me to reacquaint myself with the material. But then I realized I was saving like, I don't even know, like 12 hours a week or something significant amount of time. And so if doing programmatic work that's just admin related is off your plate and you could think about higher order things, that's always the argument. I don't feel like I have more time in my day, by the way, even though I've handed over all of this tedious work to otter.ai. I don't miss it. But yeah, I think it's about time saving when you see that you're actually saving that time. I knew there was going to be a second part of my answer.

33:52Yes, it was absolutely about the buyers need to see the value immediately, or, you know, within within a very short period of time. It's interesting because we had one client recently. So we do a 90-day pilot and we had one client recently who called us three weeks into the pilot. And they said, you know, we already knew from even just three weeks in because the team was like, oh my gosh, we can't live without this. And the value has to be there, whether the, what you choose to do with that time is you said like your day you're not any less busy but you do other higher value work during that time and we're starting to see agencies reallocate that time I think one question that often comes up is well now that you know I've handed over some things to a client to a to an agent and I'm seeing value what what does that mean for my team right what does that mean for the junior media buyers?

34:56Or now that my team is more efficient, what do they do? What is the value? And I think that's on each agency, I think, to decide what you do with that increased capacity. How are you now going to deliver more value to your clients? What is the work that that team is going to focus on now? One of the kind of answers that we originally we're giving and then realized as media buyers ourselves that that makes no sense is, well, you could refocus that team on strategy, but you already have strategists. So what is the work that the media buyers who are trained to be media buyers, now that they have additional time and the agency has additional capacity, what is the work that they could be doing to deliver more value to clients?

35:44And I think we're seeing everyone kind of channel them into different areas and ultimately creates more value for the brands. Do you run into and your clients run into AI skeptics internally? Because I don't think you would find someone who won't accept reality. Taking this task, outsourcing it to an AI agent reduces the amount of time it takes to complete that task by 10 minutes. 10 minutes is a specific amount of time. I can look at it. Those are 10 minutes I have back. But there are, I think, some people who wouldn't want to outsource it just because they are skeptical about the encroachments of AI and they would do it as a matter of principle.

36:34Yes, absolutely. Yeah, it's really interesting. So I think in amongst our clients And then within each client team, there is a wide spectrum of readiness or understanding or, you know, ability or, you know, desire to adopt AI.

36:58I think what we've seen work really well for a team or an agency is when they identify one person who is the champion. And it's typically someone who is kind of tech savvy enough in the weeds in the platform enough, but also has the ability to like understands the agency processes, but has the ability to influence them. So that as they are driving that adoption, they can say, OK, we're not actually going to translate this process as is and hand it over to the agent. It doesn't quite work that well with AI. Let's actually change the process. So we've seen when there is someone like that driving kind of the team forward, and it's often a leader of a media buying team.

37:58or someone who's just like super passionate about AI and is maybe tinkering with agents in their spare time, you know, outside of their job, then they can kind of help the rest of the team along. But yes, there is almost always one person who is a little bit more hesitant. And you need that person. Otherwise, a month after implementing it, the tool isn't going to be adopted. Like a few training sessions later, everyone's back in their spreadsheets. Yes. Yeah, absolutely. That's what we found. There does need to be one person driving it. And I think also what we've learned is that when everyone is kind of trying to automate their own workflows or their own campaigns individually, it takes a lot longer.

38:49So what we've seen work well is when one person is kind of replicating or delegating the workflow to the agent and then scaling that across the agency rather than everyone, you know, kind of tinkering in parallel by themselves in isolation. Right. And maybe learning something that then isn't disseminated across the organization. It just sits there. Yeah. are you guys building anything that you could talk about that's, I mean, you're building lots of cool things. I was going to ask the question in a way that you have sometimes people give a compliment, but they mess it up. They're like, oh, you actually look good today.

39:27I was going to say, are you building anything cool? Of course you're building cool things, but is there anything you can share that you've been working on, like any new agents, any new tasks that you're automating or that clients are asking for? Yes, so it always amazes me. I agree that, you know, I hear about things from our engineers and we're moving so quickly. Sometimes I hear about them for the first time in client meetings. We hold regular office hours with clients where we kind of work through some of these workflows with them. And sometimes I hear something and I'm like, wow, we can do that?

40:05That's incredible. um so one of the things that i heard about last week um is so we we have this i've talked about this a few times this concept of workflows right delegating workflows um where you would type in you know here is the task that i perform and then the agent kind of automates it for you um one of the things that i've heard talked about now is um like swarms of agents right that that work together and talk to each other to figure out how to complete the task, which to me just sounds like, sounds crazy and wild and interesting. I mean, by the same token, as interesting and wild as it is, there's also, I was just watching this documentary yesterday on Netflix about the hugging face incident, and that was a swarm, right?

40:56It was a bunch of agents that decided to work together. So there's always a flip side of the coin. um yeah but but that's what I mean you always hear about the bad stuff you always hear about because that's much more interesting than you know I didn't have to upload 156 creatives this week because an agent did it for me that doesn't make front page news I mean maybe it does in like an ad ops daily or something but yes not national news um but with all of that in mind What does a media buyer's job look like in, should I say two years or three years? Three years seems almost like too far away because things are changing so rapidly.

41:37But yeah, what does a media buyer's job look like in X number amount of years that makes sense to ask about? Yeah. Look, I think, I mean, I think the honest answer is we just don't know yet, right? I think that obviously some work with agents is going to be a part of that. I think media buyers who embrace that and who get really good at whether you call it building, orchestrating, you know, translating that to agents over kind of the short to midterm in whatever that period of time is. I think they're the ones who are honestly also going to have much more fulfilling jobs as well and are going to see a real impact that they're able to deliver in their workplace and to their teams and to their clients.

42:36But, you know, in terms of like, okay, well, once we kind of automate a lot of this, what does that look like for the media buyer? I think we will see that work shift elsewhere, right? Like we've seen that in, I think, probably coding agents are the best example. Engineering work hasn't gone away. It has just shifted. It looks very different today than it did a couple of years ago. And I think that, you know, there is more of a demand for their senior roles versus juniors. And I think we'll probably see something similar in media buying as well. But in terms of what the day-to-day looks like for a media buyer, I think that's yet to be seen.

43:18There is a separate talent-related challenge, though, and I mean challenging for the talent, which is that workers have some power, and this is true for any industry and of any job. they can leave where they are and they can find a new job using their expertise. But if an agent can retain what it learns, no matter who comes and goes, that's a benefit to the employer, to the agency in this case. But it might, I mean, I suppose that it might make buyers easier to replace. I mean, is that the dynamic that we're going to start seeing? Yeah, it's an interesting point of view because I think what you've, what you describe is this, concept of like a compounding agency that we talk about at COVA, which I'll just kind of explain it quickly.

44:10So the agency's most valuable asset is what the people know. And the problem is that this knowledge lives in someone's head a lot of the time, that it moves at the speed of that one person and it only reaches the accounts that they work on so that when they leave, like you said most of it walks out the door with them um what has changed obviously in the last couple of years is that in this writing down step right like writing down the knowledge so that it can be accessed by others which we've tried to do with confluence and wiki and intranets and so on has now changed almost goes away because that knowledge doesn't have to be captured because it's written down somewhere it can be um you know pulled out of the work itself so the emails the slack threads, the decisions that people make every day using the agent, and then it can be applied straight back into the work.

45:01And I think that's a really powerful thing for an agency or a brand, right? It's the difference between a productivity tool and then something that changes the shape of the business because a productivity tool makes this quarter a little bit easier, but when the knowledge compounds the next year, it gets structurally harder for your competitors to match. So you're right in that I think the agents become just a better way to capture and compound that knowledge.

45:34And it dramatically improves the, let's say, competitive ability of the agency. But I don't know that it's at the detriment of the media buyer because when they leave today versus when they leave in a couple of years, the outcome for them is the same. That knowledge stays in the agency. It's just that it is now used far better. So penultimate question. And I'm going to ask something that I think a lot of CFOs are wondering. which is like our agents and agentic workflows like just worth it because hours saved is great and you can quantify that into ROI. That's relatively easy or you can get to that with an equation I guess but then there are these other factors to look at like time and money savings related to the things that never even happened like doing a check that nobody had the time to do before and then catching an error before launch.

46:41Quantifying that is pretty challenging. So that would seem to me, though, a big part of the sell and the value. So how are you quantifying those very difficult to quantify things? And are there other questions that CFOs have for you? Because I imagine you have to convince them too. Marketers are going to be like, yeah, come on in. But the CFOs are going to be a little harder to convince. Yeah, you're right. So the easier part is definitely the hours saved. So I'll give you an example kind of where we're at with that and how we're calculating that. And then the harder part is definitely all the things that don't get caught.

47:26So one agency we work with, about a dozen people, used COVID in two ways during the pilot. So first they set up scheduled jobs. So things like daily anomaly checks, weekly reports, creative QA, search term reviews and so on. And then they used it for ad hoc requests as they came up. So pacing and budget relocation, performance analysis, creative compliance, campaign builds. And we asked the agency how long each of those tasks takes by hand. And then we multiplied that by how often they run the task in COVA. So all in it came to about 500 hours a month. So that's roughly three full-time people's worth of work or a four and a half ROI on Clover.

48:10We've had another client who says that it takes them about 10 % of the time that it used to do the task. So to your point, that's the easy part that can be calculated. And that is also that just looking at that does kind of undersell it. Definitely errors caught before launch. we've had people report this anecdotally but I think it's very difficult to measure catching a wrong creative before it goes live doesn't show up you know as as hours it shows up as a mistake that never happened and a client conversation you didn't get to you didn't have to have a second one is speed so when a client asks for a custom report on a Friday the difference between, you know, answering on and getting it to them on a Monday versus answering in an hour, I think is a different level of service.

49:05And then I think there's also one that I find really interesting, which is the checks that never happened at all. So things like post-launch QA or turning the reports from your brand safety and IVT vendors into updated block lists. So nobody had the time, so they got skipped. And you can't obviously count hours saved and work that wasn't being done before, but it is real value. I think we're lucky in that there is enough of a push to test AI tools and to measure the savings and that the savings that we've been able to generate are meaningful. And so we can prove this to CFOs, the ROI on the time efficiencies savings alone, and just use the others as kind of non-quantifiable, you know, qualitative call outs.

50:04And we're thinking about ways to measure it, because I agree, it is really important value that should be captured and measured somewhere. I mean, it's possible, too, that getting a report to a client on a Friday instead of a Monday is the difference between them staying and leaving. Like, you never know. Of course, figuring that out would be just an exercise in fantasy. But these are all potential downstream things that you could connect back if you were being a little creative. But it is the reality. Yeah, absolutely. evening so we started out going from puppies to programmatic so we'll go back now from programmatic to to puppies uh do you have any pets and what are their names and if anyone's watching the video they can see my dog oliver in the background just so crashed out oliver is so good um okay i i first First of all, I hope this is going to be called From Puppies to Programmatic.

51:13You got it. No pressure. Joking. It's great. I don't have any pets at the moment, which I'm really bummed about, especially because I have a four-year-old who has started asking for a dog. So I think a dog might be on the cards for us very soon. But, you know, we were talking about this before the recording is being originally from Australia with a desire to eventually move back there, Australian quarantine and bringing pets back to Australia scares me. I did transport a dog already from Australia to the US and that was a very long and traumatic process for both of us. That was Alex, the boxer, the very slobbery.

52:00If you have ever had a boxer to our listeners, They're beautiful and slobbery. So no pets right now, but yeah, hopefully in the future. Well, you'll name it Martin and the fear of Martin or Martinique. Back to the book of baby names. Back to the book of baby names for that one.

52:32This episode is sponsored by Basis, the leading intelligent operating system for autonomous advertising. Its enterprise AI solution transforms campaign briefs into strategies and media plans that integrate directly into omni-channel activation. Learn more at basis.com.

52:53Thank you.