Co-host: Hope you’re enjoying Agency Advantage 2025. You’ve got about 10 more minutes until our next session, which is super exciting, we’ve got a panel coming up here in just a few minutes. But we also have a deadline for the Easter egg hunt and challenge. So if you haven’t yet worked on that, if you’re not exactly sure what to do, there’s that link that we’ve shared several times throughout, we’ll put it in chat. But if you want to grab that, make sure you copy that doc and submit your answers for the Agency Advantage Easter egg challenge, that would be amazing. So, looking forward to announcing the winner for our engagement leaderboard, for our Easter egg challenge, for all of our games and activities we’ve had throughout the event, like our word search. So definitely stay tuned until the end of our program, after the panel, where we’re going to announce all those awesome winners. But if you want to submit for the Easter egg, you’ve got approximately 10 minutes and 30 seconds, so make sure to get those in. All right, we’ll see you here back in about 10 minutes for our next session. Keep those chats going if you want to hang out, sit tight, we’ll be back here very shortly.

Co-host: Hey Brent, just popping in here to again remind people that, you know, there is the Easter egg activity. There are Easter eggs on various sections of the platform. You may see them on the screen, you may see them on the lobby, you may see them in the schedule tab, so there are multiple areas where we have strategically hidden the Easter eggs. And there are approximately 10 minutes left for everyone to fill that document out and email it to the email address given in that document. So, just popping in here to remind people that.

Brent Weaver: Yeah, what were those places again, just to give those clues, just in case people need one of the plays?

Co-host: Yeah, one of the plays, maybe, you know, a background of a host, or you know, placed either on the table or on the laptop, maybe on the mic. Like, I can’t give any more clues, I don’t want to spoil it for anyone, but yeah, I feel like we’re playing like the hot-cold game, you know, like warmer, warmer.

Brent Weaver: All right, so we’ll see those submissions, good luck with that, and we’ll see you all back here for our next session in just about 9 minutes. So get those entries in and we’ll see you back here soon.

Brent Weaver: And welcome back, good to see everyone here. I’ve been loving the chat. For Oreo the Chief Happiness Officer, those of you that are wondering who the heck Oreo is, over at E2M he’s our chief happiness officer and is also a dog, so if that helps to provide some context there. But welcome back, we’ve got an amazing panel kicking off here in just a moment, and after our panel we’re going to be announcing all of our winners for all the various contests that we did over the last couple of days. Been an awesome couple of days, would love to hear feedback on chat how things are going, give us some love. We’ve got some great panelists here coming up, so excited for that. So our panel, this is on human creativity, where should agencies draw the line. That’s the topic that we’re going to be talking about, kind of like, you know, AI versus human creativity, who’s going to win in the end, we’re going to talk about that.

We’ve got three amazing panelists. First up, we’ve got Ciaran Connolly, he is an AI strategist and digital transformation expert helping small to medium enterprises implement practical AI solutions for growth, marketing automation and competitive advantage. Ciaran, welcome to our stage.

Ciaran Connolly: Great to be here, no, thank you, excited to join the chat. Yeah, it’s going to be a lively discussion today.

Brent Weaver: Next up we’ve got Aurelio V, he’s the CEO of WP Umbrella, a big shout out, one of our sponsor supporters for the Agency Advantage 2025. His company specializes in WordPress monitoring and management solutions to help agencies and freelancers streamline their workflows. I think he’s got a fantastic product and a fantastic team, he’s been doing an amazing job helping agencies that have lots and lots of WordPress websites manage those pesky updates and much much more than that. And then finally we got Mario Peshev joining us, he’s the CEO of DevriX and Growth Shuttle, investor and enterprise adviser with a focus on scaling businesses and engaging 20,000 executive readers through newsletters and communities. Mario, welcome to the program.

Mario Peshev: Brent, I’m super pumped up. I’m pretty sure that AI cannot replace you, so you’re totally safe.

Brent Weaver: I was thinking of like, maybe we’d come back from a break and just like my dog would be sitting here, and like if I could get her to somehow be like comic relief, but maybe that’ll be the AI version of Brent next time.

Mario Peshev: Hit pause and bring them on the show in the break, so definitely make sure you do that. So, Brent, welcome to the stage, you got three amazing lads here that know, combined, probably more than most of the world in terms of WordPress web and agency worlds.

Brent Weaver: We’re going to start off here with a nice, maybe an easy one, but, can AI ever truly be creative, or is it just always remixing human ideas? How do we define originality? And we’ll start off, Aurelio, why don’t you kick us off here. Can AI ever be truly creative?

Aurelio Volle: This is a tough one. If you understand how LLMs are working, then AI cannot be creative per se, if you allow me to say so. But then, I mean, if you gather up all the existing creativity of humanity and you remixed everything, then you make creativity. So I think it’s a very hard question to answer. I think AI can be creative, and you can be creative without creating something new, and that would be my take. So you can be creative without necessarily creating something new, so it’s always remixing something in some way.

Brent Weaver: All right, Mario, what’s your take?

Mario Peshev: Yeah, I’m going to kind of double-click on what Aurelio said. So ultimately, from a technical standpoint, they are trained on large data sets, but at the same time they’re generative models, so they do generate ideas based on the existing data set, right? So there is a certain level of creativity. This is the reason why they hallucinate that much, right, they come up with stuff that they don’t have access to, they are imagining some stuff that does not exist, we see that particularly for video generation or image generation. So there, to a certain extent, there is creativity. However, it’s still limited. When you get the best advertising expert, creative expert, PR specialist, even, you know, content marketer or a digital marketer, they do have that specific twist that’s emotional and, you know, entertaining, that understands how, when everyone zigs, you do zag, and this is kind of what makes humans special. So in that sense, if you’re looking to the top 0.1%, humans are always going to win 100% of the time, but when you look at the top 10%, there’s definitely some AI that’s going to take human creativity to the next level. Like, there’s almost a threshold of, you know, when it comes to marketing or agencies, how much of what you do has to be completely original, 100% creative. Like if I own a law firm in Colorado and we do estate planning, right, probably we can argue that that business might not need to have the highest potential of human creativity when it comes to their marketing. But, you know, maybe there’s some truth to that, maybe there’s not, I don’t know.

Brent Weaver: Ciaran, what do you think, can AI be truly creative?

Ciaran Connolly: I echo exactly what the gents said so far. I see it as a remixing engine in its core, and thankfully that’s where human skills or capabilities come in, with our creativity, our lived experience, our culture, our history, our emotions, that probably drive a lot of the creativity in what we do, and that’s definitely, just to echo, that’s where I think we have the edge. Will it get to the point that it can replace human creativity? I don’t know. I hope not. But yeah, at the core it’s definitely a remixing engine for us, and that’s where we still, as an agency, still add the creativity.

Brent Weaver: Yeah, it’s weird, I mean, I feel like some of my best, personally, my best business ideas, writing ideas, you know, a lot of times they happen outside of the office, right? Maybe I’m on a bike ride or I’m on a walk with my family or I’m sitting on a beach or I’m at a conference sitting in a big room listening to other people. I think what Landon said in chat there, there’s definitely some truth to that, right? Like, a lot of human creativity is some form of repurposing, right, we’ve all kind of have these experiences and baseline fundamental education, but then I still think that, with the human experience side, not our knowledge, not the content, but there’s still these sparks that I personally get where, you know, you just get flooded with this idea, this thing hits you, it’s like the synapses connect, and maybe it is a little bit of a repurposing, but sometimes there’s that little bit of newness that I just haven’t seen AI necessarily get. And I assume that’s because we get to experience things out in the real world, and AI obviously doesn’t. They get a version of that, they get our videos, they get our photos, they get our text about it, but they don’t actually get to see it with their own eyes.

Here’s our second question for our panelists. When an agency uses generative AI, should they disclose that fact to clients or end users, and how might that affect brand trust? So, do we tell clients that we’re using AI, or do we use it in stealth? What do you guys think? Let’s start with Mario on this one.

Mario Peshev: Well, I think the key aspect here is, first off, there are some contradicting beliefs on whether the output is originated, or whether it’s copying using copyrighted content, whether it’s something that you actually have the authority and right to use, right? There are existing lawsuits already, to OpenAI, to Anthropic, and other companies, especially the Ghibli images that came up and started generating lots of copyrighted creative content. So with that in mind, there’s always the risk to publish content online that you don’t necessarily have the full authority or right to do. That’s on the one hand. The second thing, unless that’s fully edited and manually controlled and so forth, due to hallucinations there’s a possibility that you end up quoting the wrong stats or even wrong facts, which is certainly a factor you need to consider. And with that in mind, considering that there’s a lot of time saved by using AI, I would 100% disclose that AI is being used to speed the processes or optimize the workflow or generate ideas or come up with an outline, come up with a hundred suggestions to narrow down and pick from, to an extent. But yeah, definitely disclose, 10 out of 10.

Brent Weaver: Yeah, I mean, I’m just trying to think of, like, AI obviously is a unique tool, but I don’t go to my clients and say, you know, on your project I will be using a calculator and spreadsheets to make my job easier. I wonder if there’ll be a point where it’s just assumed that everything is going to have some level of AI touching it, and assuming that people are using best practices and stuff. Aurelio, what do you think?

Aurelio Volle: I mean, it’s going to be a very dull round table because I fully agree again with what Mario said. Overall, I think whatever you do when you’re an agency, you need to be as transparent as possible with your clients, because it’s a way to build trust. And if you don’t mention that you are using AI because it’s so involved in your current process, and it is so obvious because you are kind of AI advocates everywhere, then it’s fine. But if you are using AI on the workload you are doing for your client, and you are not being transparent, I think you are at risk of breaking trust and then losing your clients. Because in the end it’s like a plumber, I mean, he’s using tools, and you are using tools to build websites, and that’s it, and that’s 2025, and if you’re not using AI and you’re running an agency, it’s super strange.

Brent Weaver: Yeah, so being transparent about it, not necessarily hiding it. But I imagine a lot of clients, I mean, for a lot of small businesses, I’m just thinking about like your mom and pop kind of local businesses, I imagine they might not be that interested in what is making the sausage, so to speak, right, they might be a little bit less wanting to dig into it. Whereas, like, enterprise, I imagine when we’re talking about data policies and intellectual property and how that stuff is handled, I imagine they might have much stronger views on that. Ciaran, what do you think, do you have a consensus view here or contrarian?

Ciaran Connolly: Yeah, no, again, totally. We would write a lot of content for clients, for example, and clients would have, when the AI wave rolled out, clients would actually be asking us, are you using AI for this? And we would have handwritten content and then AI assisted. Now we actually put it into our proposals and say, for example, AI assisted, and be very transparent. So we would have in our terms and conditions on our website, we would have it in our proposals, this is where we’re using the AI elements. And maybe in a couple of years’ time we’ll not have to do this because it’ll become part of every aspect of what we do in our business, and it’ll not be such a big thing. We are seeing as well, based on the EU laws coming in, and word of laws where there’s going to be a requirement for companies, especially large companies, we see it in California as well from next year I think it is, so large companies where a large proportion of what they handle with AI has to be publicly made aware to the public. So again, that might roll in and change things too.

Brent Weaver: Yeah, so they’re looking to make like a public disclosure that certain things have been made with AI, is kind of the idea.

Ciaran Connolly: Yeah, there’s the, I was reading that there’s a California AI transparency act which should come in from 2026, so you need to be publicly making it aware, or you know, make it a disclosure, if there’s a certain percentage of your business being run by AI or it’s influencing it. And there’s word or talk about something similar for the EU. So I can see, but that’s for, it is for large companies, so it’s going to be very interesting. The lawmakers are slow in catching up, so we’ll see it in a year or two and then we’ll all have to react, but definitely for larger companies there’s going to be requirements coming in.

Brent Weaver: Interesting. I don’t know, maybe I’m too much of an American capitalist and free speech advocate, but I just don’t see, yeah, having to disclose that your business is using AI, it seems to me like an interesting way that the government is getting involved. It reminds me of, like, is it the same as using hormones in food, using AI to write content, like, are we talking about the same thing here, like a health blog post on AI, I don’t know.

Ciaran Connolly: Yeah, probably same principles.

Brent Weaver: The lawmakers need to keep busy, the politicians.

Ciaran Connolly: Right, the hammer that only sees nails.

Brent Weaver: So, what do we think in terms of our ideal level of human involvement in AI-assisted creative work to maintain ethical outcome and integrity? And it was actually really interesting, we’ll get this question up on the screen, team, it’s question number three, but we did our AI image contest yesterday, and I’m gonna say, I think our audience could have been a little bit more creative in their prompting on this one, because I actually saw a lot of the submissions looked very similar. Like, people just took our prompt of, like, imagine what agencies are going to look like in 2050, and they just put it into ChatGPT, and so it was like everybody, all these agencies lived in different places, they’re all obviously attending our session, but like seven of the 10 or 20 or 15 images that we got kind of looked almost identical, because everybody took the same prompt and they put it in there. Now, they were technically different images, but they just kind of all had the same kind of vibe. And so I feel like, in a way, it negatively impacted the creative output, like the human didn’t take the time to really think about that creativity and put some spark in there. So, Mario, what do you think, what’s the ideal level of human involvement, how involved should humans be in making creative works by agencies and marketers?

Mario Peshev: Sure. I’d break it down into two kind of sub-questions, if you let me do that. So the first part is, when you use AI, how much human involvement should be in there, right? So I’d say that, looking into the fine balance, there should be some AI assisted uses, like research, copying out, pulling some data, crawling, using MCPs to tap into different data sources and stuff, definitely automated, it’s mundane tasks that you shouldn’t be bothered with. But then definitely human assisted on top, and people reviewing, editing, cherry-picking what’s being used and all that stuff. But second, to tap into what you said about the image creative contest, I think that the biggest caveat right now that we see is that lots of people are trying to use AI for activities that they don’t understand very well, while the best way to use AI is to empower and power up yourselves, or work with other people who can 10x themselves or even 100x themselves. What do I mean here? If I’m a developer, I know very well how to write a brief, how to scope out, you know, how many files I need, how to break down and segment modules and functions and stuff. If I’m a project manager, I know how to scope out a project. If I’m a salesperson, I know what my deck needs to look like, what type of slides. But, you know, if I’m not a designer, how do I tell it, hey, I need this to be like these gradients, and the person’s face needs to be tied into this, and the shade falling through this angle and all that. There’s so much unique skill set in every single role, accumulated over 10,000-plus hours of expertise, that trying to mimic that with AI with a simple two-sentence prompt is simply impossible.

Brent Weaver: So it’s almost like the real experts within a discipline, they’re going to be able to give a very robust prompt, they’re going to be thinking about a myriad of things when they put their prompt in. It kind of begs the question of whether or not those skilled people will exist in three to four years, right? If they’ve spent the last 10 years building up their design experience, and now they’re just becoming prompt experts, right, like in five or 10 years will we have that batch of people with that expertise, or will the shift be more on how to become a prompt expert? Kind of interesting ideas there. Aurelio, what do you think the ideal level of human involvement is?

Aurelio Volle: I think AI is mostly an enabler, so in the end it depends on your level playing field. Let’s say if you’re very medium, very average, then AI is going to allow you to produce more medium, average content. If you’re very creative, and if you’re talented enough to write good prompts, then AI can allow you to do tremendous things. And in the end we shouldn’t see AI as a good or bad thing, we should just see it as an enabler, and that would be my take.

Brent Weaver: Do you think, though, it can also kind of mask the bad? So, for some people, I’ve seen this with job applications, I mentioned this earlier in the session, I used to ask these survey questions, and it was always specific on their role, but I’d usually have three or four questions, and those questions would spread the playing field really wide, because you could see people’s thinking, you could see how they organized their thoughts. Like, yeah, the grammar tools helped a little bit, but the AI tools have just totally decimated that as a tactic, right? Like, when I go and get a 100 job applications, if I ask a survey, I mean 80 to 90% of them you can tell they’re writing with ChatGPT, and for me as a human it’s really difficult to separate the wheat from the chaff when you’re looking at a hundred responses that have been written by a machine that can generate the best-looking language. I saw an English professor said, yeah, it looks good, it’s got the structure, but it’s soulless, it doesn’t have a soul. And so, you know, maybe that’s the thing that we have to look for. But I do feel like it’s given people that are subpar a leg up, like there’s kind of a cheap button, an easy button, to do a certain level of creative work. Ciaran, what do you think in terms of ideal level of human involvement?

Ciaran Connolly: I’d say it depends on the task. So, we’re finding now, say content writing, you might be doing 70, 80% of that, it could be 20% for human review and just checking, make sure the quality is there. Coding, it might be 10, 15%. So again, it’s from the AI side. Video creation, again, maybe similar too, but that’s going to get interesting now, I don’t know if everyone’s seen the new AI video editor from Descript, very interesting as well, and we’re all waiting now with our keyboards ready for ChatGPT-5 when it comes out, I think it’s meant to be this summer, so that was the latest update we heard, so we’re all waiting for that. So, definitely depends on what you’re doing, but I can see it getting less and less as the AIs improve going forward, so it’s going to get interesting.

Brent Weaver: Yeah, I saw a post by agency influencer in our space, Jason Swenk, today, he was talking about how the big platforms are getting everybody addicted to using the tools, and they’re going to get so good and we’re going to be so dependent on them, and then at some point they’re going to like 10x the prices and we’re all just going to be, you know, screwed. So maybe five or six will be that point. Mario, I’m gonna throw this question over to you, are there creative tasks AI should never touch, like emotional storytelling or high-stakes brand strategy, are there things that you would avoid using AI for, or you’d tell other people to avoid using AI?

Mario Peshev: Well, what we’ve been seeing, and kind of one of the main reasons I’m on this panel, is I’ve made it to 12 events over the past two months, and this is more than the entirety of 2022, 2023 together, right? And the reason I do that is there’s so much automated AI content going on that it hurts, it literally hurts my eyes, right? As you said, you go through a wave of 100 questions, you don’t even know what’s AI, what isn’t. We talked about whether AI needs to be denoted or flagged somewhere. Actually, if you go on X you see automated accounts, there’s already a flag for that. If you go on LinkedIn and you upload a DALL·E image, it also gets flagged as, you know, AI generated. There are some flags and telltale signs, or metatags when you copy over content from ChatGPT as well. So the world is already trying to delineate AI versus non-AI. And with vibe coding, and obviously the entire AI marketing content generation, email generation, automated LinkedIn comments and posts and stuff, it gets even more important to figure out whether you seek something for the use of it, or it’s more important who’s building that specific product, you know, it’s the brand versus feature. So within a blue ocean environment, meaning there’s a very specific unique custom tool that’s the one and only solving a specific problem, yeah, sure, you can use the entire playbook. Like if you’re in a small village, there’s only one supermarket, everyone subscribed there, they’re reading because this is the supermarket, you know when it’s open or when it’s closed and all that. But, you know, agencies, if you have 10,000 agencies in your country doing the same thing, then the more you use AI for outbound activities, the more painful that’s going to be. This is where brand matters the most, trust, connections, relationships, one-on-ones, warm introductions, all of that is kind of the significant piece. So moving between blue ocean, red ocean, if you’re the unicorn in the space, the one and only, great, you can do whatever you want, you can post whatever. Otherwise, if you’re in a dense space, especially the agency space, you know, be human.

Brent Weaver: Yeah, so just in terms of how much competition, if everybody’s using AI, it’s going to all look the same, right? It’s almost like, at least for job applications, now that everybody’s using AI, they all look like these very polished, crisp, kind of cover-letter-style responses, and I’m almost looking for the one that’s, like, got the spelling mistakes but you could tell it’s real thinking, you know, it’s like the wheat’s all there but it might be a little bit unpolished, right? And I’ve even thought about, for myself, as I write important business emails and things like that, if I want to illustrate to somebody that I’m not using AI, there’s certain language that I might write in that I just know that somebody will be like, oh, that’s not AI, that’s actually Brent talking to me, right, versus the nice polished bullets and em dashes and bolding and the emojis that AI makes so easy now. Aurelio, what do you think, are there some tasks, or even just thinking about your team at WP Umbrella, there’s stuff that you have said, hey, this is off limits for AI within the business?

Aurelio Volle: Yes. But just coming back on the hiring people with AI, I mean the automatic AI answers issue, I’m asking in this screening question, did you use AI to write your answers, and it’s so funny how many people tell no. Like, they lie in the screening questions, and this is also a good indicator. And maybe you should do that, you have a polygraph on the first job interview.

Brent Weaver: We’ll hook people up to the sensors, and I can just see you in the room coming in like, did you use AI for this? No.

Aurelio Volle: But, like, circling back to transparency with clients, it’s very fine to use AI, just admit it, and we are good to go, it’s not like a red line. What is a red line is to lie on the fact that you use AI. And coming back to your question, sorry, there are two kinds of data that we basically never send into any kind of AI. Which is, like, the super sensitive business data, we don’t have a lot because we build in public, but I have many NDAs going on with many companies, and this is the kind of stuff, I’m super careful because this data is going somewhere, and data privacy matters and will matter even more, I think. So on this you have to be very careful. And you have also to be very careful about the data that you don’t own, such as the data of your clients, so for instance, we don’t put any client-related data into AI unless they’re purely, purely agnostic. And I think this is important, and it’s also a good way to build trust actually.

Brent Weaver: Yeah, I mean, I think there’s, I don’t know if there’s like a reckoning coming with how much data or client content, I mean, I feel like the amount of guardrails for an agency to get access to client systems and client data and client content and use AI to augment not just the deliverables but the thinking that goes behind what they’re doing with the business, the strategy, right, I mean there’s almost no way for the average business to know how much an agency or a contractor or freelancer is using AI behind the scenes, right? It’s just kind of wild to think about how much data might be being uploaded to these various platforms on a regular basis. Ciaran, what do you think in terms of creative tasks that AI can touch?

Ciaran Connolly: Yeah, back to, you think of emotional stories or culturally sensitive things, again I can imagine, like, some projects you’d work, maybe in the charity sector, where it could be highly charged, I think that’s where you probably need a person in talking to clients that are maybe very sensitive about things, that’s probably an area that you should just not go. But again, I can see AI being helpful in helping with scripts and things like that, but the core concept and the ideas, that’s probably where I’d keep off limits for most.

Brent Weaver: Yeah, I see this note in chat, I’d rather hire the guy that can finesse AI well rather than a designer, a writer, etc. It’s an interesting idea, right, because, but I do feel like there’s something to be said about knowing what is good and what is bad, because you get something out of AI, and if you don’t have somebody that has the domain expertise of knowing what is good copy, like there’s a lot of stuff that might seem like it’s good copy but it actually might not be, or something that seems like a good design but it actually might have some glaring mistakes or issues that might affect performance or something like that. So it’s hard to find that sweet spot of getting experts versus people that can just seem like experts. Like, at some point, the image I had in my head the other day was, you know, are we all becoming the wizard behind the curtain in the Wizard of Oz, right, like are we all just controlling, we’re just building this illusion where it looks like we’re really smart and really talented but we’re really just controlling these AI tools to do it for us, and then at some point that will, you know, we’ll go beyond our capabilities with that and we’ll do something that really matters, and the AI is doing the thinking but it might not be good thinking. I don’t know, that’s just my theory.

So, do you think about which roles are at risk, this question of, as agencies adopt AI, do you think there are certain roles that are just not going to exist in 12 months and 18 months? What do you guys think, what’s your call here? We’ll start with Mario.

Mario Peshev: Good question. I’m not sure if I’m willing to put it as roles, rather than the entire, like, intern is definitely going to disappear the way we know it right now. Because initially when I hear the question I’m thinking, okay, you know, there’s web coding, development is going to disappear, and I’m saying, well, it’s not, because there’s, you know, hallucinations and security issues, scalability issues, yada yada. Then I’m thinking, you know, obviously marketing didn’t disappear, because we see what sort of crap it churns, right, however you want to prompt it, it just needs all that human touch and emotional intelligence just put right in. So, like, just cruising through different roles, they’re all safe in a sense. However, whether you need, like, a virtual assistant to schedule your, I don’t know, Instagram content throughout the month, well, probably not, I mean agents are doing that well enough right now, not just tools like, I know, Buffer or Hootsuite, or just agentic systems are doing that easily in a myriad of different ways. Data entry, probably not needed. So, like, most of these kind of just standard outsourced low-cost roles are going to disappear, as well as juniors that you need to spend a hell of a lot of time to train for two years and hope they stay with you in the company afterwards, right, they just need to raise the bar, upskill, and get to a level where they can assist and operate faster. So, yeah, it’s more about raising the bar for most roles, and once again focusing on brand or focusing on trust. I’d still pay a lot for a secretary that I can trust, that’s going to be trustworthy, pick up the phone, make decisions on my behalf, know me well enough to know how I feel based on how I look and all those considerations. It’s not about the skills, it’s about, again, the human touch, or the context in which the human operates. Just, fewer people are going to be needed for most activities that we need right now.

Brent Weaver: Yeah, I think a couple of thoughts pop in my mind, but that trust word, right, and accountability too, right, like, if an employee screws up you can go, and there’s some accountability, there’s some risk there for an employee working on a project or within your business, right, they have to do a good job because their job depends on it, whereas I feel like if ChatGPT screws something up, I mean, I’m not going to go and sue some 800-pound gorilla, right, I can’t go and attack OpenAI, there’s no accountability there, right, which I think there are certain roles in the business that might be affected by that. And I wonder too if there’s going to be a crisis of talent, because, you know, when I look at our own business, we’re very hungry for senior team members that can oversee AI production to make sure that it’s hitting our standards, right, but then you go, okay, well, where do senior people come from, well, they come from mid-level, where do mid-level come from, well, they come from juniors, and if we don’t have juniors coming into the marketplace in the next couple of years, giving them those opportunities, what’s that going to mean for the workforce in the next couple of years? Aurelio, what do you think in terms of roles that might be at risk, any come to the top of mind for you?

Aurelio Volle: Just like Mario, I don’t think a role is going to disappear per se. For instance, at the moment at WP Umbrella we have two employees in the customer support department, and I tasked one of them to be implementing AI, and he told me, hey man, I’m going to implement AI and then you’re going to downsize, and I was like, no, I’m not going to downsize, because I will always need one or two persons to be accountable for what AI is doing, and I think it will boost people’s productivity. And now with, like, five good persons well trained in AI, you can do the work that probably 50 were doing two years ago. This is, mostly when it comes to development, nowadays the productivity of a senior developer is insane. And so I think the job market is going to be tightened up, but I’m not sure that roles are going to disappear entirely. And what you said about what’s going to happen to the youth, it’s super scary, because our society, it’s going way too fast. I mean, we are already, like, very in advance I think, like the four of us and all agency owners, in our usage of AI, and it’s already going too fast for us, so how is the labor market going to look in four years from now for young people, that’s super scary.

Brent Weaver: I mean, to be fair, probably for our parents’ generation, I don’t know if any of our jobs existed, so, like, you know, when I ask my kids, you know, what do you want to be when they grow up, it’s like, I want to be a professional gamer, I want to be a streamer, and I’m like, what happened to the doctor and the fireman and the astronaut, right? But, you know, it does change. Ciaran, any roles you think are going to disappear or maybe become less relevant in the next few years?

Ciaran Connolly: Yeah, it’s interesting, again, I think what we’re all talking about here is that the roles that are safe are probably the ones that can control and manage AI, the roles that will be responsible for automation, the roles that are leading and key parts of our business, and looking after the humans, and probably the roles that are talking to our customers and selling. At the end, there will be AI rolled into all of that. Definitely the juniors, there’s a large tech company here locally to us, they would take on 400 junior placements, students, every year, this year they haven’t taken on any, and that’s obviously, we know why, it’s going to be AI is taking that role. I’m just wondering, we’re talking about juniors now, I wonder in two or three years will the AI move up to take the medium role as well, I don’t know, it’s going to get very interesting. So, yeah, everything’s going to change, I think every job is going to change. Again, we were a content marketing agency, we’re no longer a content marketing agency, we’ve had to change that, we don’t have content writers anymore, we have digital strategists. So that road, and I see it happening to video, I see it happen to website development, so it’s going to be, yeah, it’s going to get very interesting over the next couple of years, but we need to keep moving and just looking for people that understand AI and bring them into our business, and keep implementing, so, getting interesting.

Brent Weaver: I think a lot of people with those junior roles, right, they might be sweating. I just saw in chat, Nicholas Williams must be a very popular person with his Department of Health here in Florida, but they’re currently in process of replacing about 120 humans across their state with AI for health insurance enrollment. I’m sure that that will make Nicholas a popular feature at the upcoming Christmas party. But I think that’s real, right, I mean, I think people are having to adjust to this, I mean, probably saw this, you know, back when cars came out, when manufacturing came out, when the internet came out, there were just certain things that became a lot less relevant with that as time has gone on. So there’s certainly a real impact that’s happening here, and, you know, we’ll see where it goes over the next few years.

Getting a little bit tactical, as we’re getting close to the end of our session here, I’d love to hear a couple of best practices or guardrails that agencies should put in place to use AI responsibly, and if you have specific processes or policies that you’ve implemented in your own companies, I think that would be amazing to hear. So, Mario, why don’t you start us off here?

Mario Peshev: Yeah, absolutely, and just to close up the chapter, to Nicolas’s point, I actually have been working in Medicare for the past couple of months heavily, and automating is definitely not as easy as it looks for professional insurance licensed agents, and, you know, I can dive in but that’s not like a healthcare conversation, but some things look brighter than they normally look, and of course it depends on the people. If it’s a purely admin role just sitting behind a desk, well, it’s really hard, I mean it’s primarily replacing paper with a computer, you know, it’s not a competitive thing to do, but there’s a lot of power in a human being that actually guides you through the process and explains the pros and cons and feels your emotional state and talks about your grandson or something, and there’s this fine level of detail that’s like the human touch that we were talking about. Even though, once again, yeah, do connect, Nicholas, we can chat after. To your point about guardrails, what we definitely do a lot here is we build our own red teams, meaning that whenever we use AI we run a counter team trying to challenge it in a myriad of different ways and find holes and loopholes and ask for details and questions. In many cases we use, like, three different, and pass information from one of them of, hey, you know, I spotted a couple lies here, can you find them, right, or, I lack confidence in what the other two provided here, find, or I don’t believe these details are accurate, dive deeper. So through inference, which is one of the strengths of the deep-thinking models, there’s a lot that could be accomplished by asking further. One popular framework that I forgot who coined it, but the five whys, you know, you need to ask a question five times, why is that, okay, why did you do the next step, then why did you decide to do this, you know, after the fifth why you normally get to the bottom of it, and you uncover things that you can’t simply uncover with the first why. So this is a pretty common business technique for 20 years or so, and we just need to apply the same critical thinking with AI, especially as we know how good it is at hallucination.

Brent Weaver: Yeah, I mean, there’s definitely, I was reading something on Substack the other day, it was like a writer was wanting feedback on several different published blog posts, and the writer said, well, which of the posts best represents, you know, something, right, and the AI response wrote this beautiful elegant response, and then he kept asking questions and talking about it, and he started picking up hints that he was like, wait a minute, I didn’t actually say that, like, the AI was quoting his articles, and he finally confronted the AI and said, hey, you know what, why did you say this, I didn’t say this. And it turned out that the links that he had shared for his five different articles weren’t even accessible by the AI model, but the AI model, instead of saying, hey, I couldn’t access these links, instead the response was what it thought it should respond based on the URL slugs of the post and the title of the post. So it literally was just hallucinating this entire response based on what was in the URL of the links, and inferring what that content might have been about based on what it already knew about this person, but the whole thing was essentially fabricated, which, you know, it took a little while, it took five or 10 prompts before they realized, like, the tool I’m talking to isn’t actually giving me real information, which is kind of crazy. So, having guardrails to protect against that happening with your clients, very important. Aurelio, what do you think in terms of best practices or guardrails, what have you guys done at WP Umbrella that has worked really well?

Aurelio Volle: There are two things I always do, plus one guardrail. The first one is I try to always optimize my prompt, to optimize my use of AI, to make sure that I can get the most of it, and it’s not going to give me just some random stuff that we see everywhere, and this is a very good practice, because the better the prompt, the better the answer, this is one thing. And I often ask it, when I want to publish content, to quote sources, which is also a very good thing to do from an SEO point of view, because then you have natural external quality backlinks to your piece of content, this is what I always try to do, and this is what our marketing team is doing. And when it comes to guardrails, always have someone having the ownership of the content produced. Pardon?

Brent Weaver: I think you lost me.

Aurelio Volle: I was saying, always having someone who has the ownership, and thus the liability, of the content produced, because then the person will have to review what has been written and to engage his responsibility, which is a good way to have some content that has some kind of quality. And I actually fired someone in December, because he was a content writer and he was publishing a lot of good stuff on the blog, and at one point I went to check one article and I was like, dude, you didn’t even double-check what you’ve published, goodbye, you killed the article. And I think this is a good practice, to have someone accountable for what AI is doing.

Brent Weaver: I think, internally, you know, we have a writing culture at E2M, and so, using a lot of written documentation, and I fully expect team members will utilize, especially being a global international company, utilize AI tools to enhance and enable their writing and make it more clear and concise. We specifically say, if you wouldn’t fully stand behind this paragraph or sentence in front of the leadership team, then don’t let it go into your documents unedited, right? So I think that accountability piece is key, that you can’t let team members or employees or contractors or agencies say, oh, you know, we screwed up, that was the AI that did it, right? Ultimately, if you’re the one that’s commanding the tool, you have to be willing to stand behind your work. I think that accountability is really important for agencies that want to raise the bar. Ciaran, what do you think?

Ciaran Connolly: Yeah, sure, so we’ve, AI champion, so exact same as the team here, so we have someone that’s totally responsible for it, but we’re actually now rolling in AI champions into each department, because what we’re finding is we’re having to push a wee bit more than we maybe want to with AI, because everyone’s very busy in their day job, and we’re having to drive it a wee bit. We want, you know, as an agency leader, we want AI implementation a little bit faster, so we’re finding if we spoon-feed it, go through it, research it, and say, look, here’s what you do, and give the team the tools, so basically we would subscribe to a lot of different paid software so they can go and try it safely, so the data should be secure, it’s a commercial agreement. And what we’re finding is these tools are updating so fast, we’re going back maybe once a quarter, every four or five months, and rechecking it, because there’s so many changes and updates. So, giving them the tools and time to test it and change and check it, so what else can we implement into the business.

Brent Weaver: Yeah, great, I like the, I feel like a lot of businesses have the AI champion, and they become the chief bottleneck of AI very quickly, because it’s like, you know, the one poor dude that has to look over AI for a whole organization, right, so I love hearing distributing that across teams, right, and also even giving your grassroots team members space to do that, right. I mean, I find that in a services business, at least for us, you know, do you focus on client work or do you focus on reskilling and upskilling your team, right, there’s a trade-off there, and as a business you say, look, we might need to slow down a little bit in order to go faster, we might need to give every team member a little bit of time to really think about AI and upskill and reskill their seats. So, well, I think we’re coming to the close of our panel, guys, this has been absolutely fantastic, we didn’t get to quite all of the questions that we had available today, but I think the insights, the discussion, was great, and I’m seeing some love from Nicholas in chat, thank you, appreciate it. So, yeah, folks at Agency Advantage, let’s give our panelists some love here today, let’s fire up some of those emojis, let us know what you thought of the session in chat, really appreciate you all being here, you guys were fantastic for our panel, thank you for supporting this year’s event. Aurelio, great to have you here, Mario, Ciaran, best of luck on your journey with AI, you all are doing amazing jobs leading great companies and looking forward to hearing more about what you do in the future, so thanks again for being here.

Mario Peshev: Thanks, everyone. Yeah, thanks, Brent, for moderating.

Brent Weaver: Yeah, thank you. Salah says, what happened right now is what AI can’t do, all right, magic, four of us. We’ll be safe in our jobs as long as we just keep showing up for panels, we’ll be in the clear. Awesome. All right, folks, we are close to our close of Agency Advantage 2025, we’re going to be right back, we’re going to come back and announce our winners and we’re going to finish strong for this year’s event, so if you’re waiting on the edge of your seat to see if you got one of those coveted gift cards or hosting credits, we’re going to give some shoutouts and accolades here in just a few minutes, so stick with us, keep that chat warm and engaged, and we’ll see you back here in just about one or two minutes.