How Campaigns Are Using AI For Copywriting & Creative – Brian Athey (Push Digital)

Brian Athey:

Suddenly there is this new standard where everybody has this tool and the new challenges not competing against each other with the tools, but competing with each other against the ideas of how to use it.

Eric Wilson:

I'm Eric Wilson, managing Partner of Startup Caucus, the home of campaign tech Innovation on the right. Welcome to the Business of Politics Show. On this podcast, you are joining in on a conversation with entrepreneurs, operatives, and experts who make professional politics happen. Our guest today is Brian Athe, chief Creative Officer at Push Digital, one of the top digital agencies in Republican politics. We spoke to their founder and c e o Wesley Donahue in a previous episode. Brian is a skilled designer with a long background in both campaigns and commerce. We dig into artificial intelligence on today's episode, how it's being used in the creative realm of politics, what tools he's using, and how you could start to incorporate AI into your campaigns. Brian, when we talk about ai, I think it's very easy to get lost in abstractions and what it could do in the future. What are some concrete examples of common campaign creative tasks that AI is being used for right now?

Brian Athey:

The easiest way to look at it is that you should really look at AI as a tool to perform tasks. So you're, you're not asking it to do a whole job, but you're looking at it in the, in the tasks that need repetition or that oftentimes you need to delegate with the intent of perhaps speeding up your process or accelerating your ideation. So it's really about the delegation and then how good you are at learning how it works, sort of educating it, conditioning it for what you want it to do. You had Ian Hines on a couple of episodes back, and he was talking about just the cadence of writing emails and how many times that he will write them and how many times he's trying to make them fresh or inventive or imaginative. That's gonna take up a ton of time, but that's essentially like a templatized deliverable.

Like if you've ever worked with email copywriting, like there's a formula that you can follow at least that if you're, if you're trying to replicate how the experts do it and what AI will do, if you can kind of teach at that formula, and especially if you can teach at that tone of voice, it will help you scale and generate those ideas much faster for a much sort of like wider consideration of audience or background or segment or demographic, whatever you're really like trying to, to, to write this content for. You can literally tell it to give me a pro school choice version of this topic, but you can also ask it like fun things like, you know, write this like a Texan or, you know, put this in the, put this in the tone of voice of Matthew McConaughey. So there's this, there's this great sort of, when I talk about scale, I mean that you can create tons of variations very quickly and you can as a creative, you don't necessarily have to start with a blank piece of paper as long as the idea is already kind of moving.

Eric Wilson:

Right, right. That zero to one, that creating. And if you can use the AI to get you started, I'll just, you know, share with listeners a little bit of the process of whenever I'm writing questions for guests, I've started to, you know, feed through some of the things that they've written, interviews that they've given through chat G p t it suggests some questions. They're not always good, but it starts to give me things to, to think about. So that's just, just one example in my, my own process,

Brian Athey:

And that's perfect. And what it does is, you know, you, you already have the, the catalyst, you know, like you, you've already started there with the idea, and you know that between inviting a guest on and talking to the guest, you've got a series of tasks that you've gotta complete. So it's not, it's not going to decide that it wants to talk to Wesley Donahue about the new book. And it's not gonna send you a message that says, Hey, you should, you should really have Wesley on your podcast <laugh>. But what it's gonna do is it's gonna fill in all the gaps between your curiosity and your execution. And I think that that's, you know, that's, that's a really, that's a really exciting part. But that there are also like all these just sort of like daily tasks that we ask it to do. You know, like you can say you need to validate your code, or, you know, you're building a website and you're returning an error.

AI is great at picking that up. Say you need to sift through a lot of data, say you need to transcribe a video generate captions on your you know, on your video team. I think that there's, there's just a ton of little applications like that, that when people think of ai, they think of this big omniscient talking to God kind of thing. But AI is really like things that we're already familiar with. In fact, like, you know, I was talking to some of my team members Eric Frenchman, who leads our advertising team, and he was like, look, you know, AI is chat G P T and, and all this new stuff. Like, it's, it's really cool, but it's also like, there's a little bit of hype around it. Like when you are doing this professionally, you're sort of used to these automated and machine learning tools, whether it's from Google or Trade Desk or whatever. You, you sort of have the, the benefit of that.

Eric Wilson:

Well, that's the first thing I point out to people when they're like, oh, is, is politics ready for artificial intelligence? And I say, well, newsflash, we've been using artificial intelligence specifically the branch called machine learning for years to optimize our ad buying. But Brian, what about those more complex graphic design tasks? So like logos, branding, how is AI going to change the way you work if already changing your way, way you work? Are, are you gonna be out of a job soon?

Brian Athey:

I mean, look, I, I think that honestly, that's, that's the big, so that's the big question, right? Like, I think that everyone, it's weird that this started with art. Like, you know, AI didn't solve like hunger or it didn't solve like doing your taxes, which, you know, I think it's kind of coming around to that. I'm talking to my neighbor that's an accountant right now, and he's, he's, but it started with art, which is like the most imaginative thing, and it started with sort of like a very sort of tactical application of art, which is marketing and branding and storytelling. And I think that, you know, it's a, it's this, it's this weird sort of existential thing that happened to this industry, especially creatives in the industry where it's like, oh, we're all outta jobs until we started using it. And I think that, you know, the, the real answer is that like, this is a super profound tool with like, crazy, crazy potential.

But right now it's, it's just not quite there. There's this there's this quote that I saw online the other day, and it's like for AI to replace graphic designers, the client would need to tell it what they want <laugh>. And so our jobs are safe. You know, like, I think I think to answer the question, like, look, AI can already do logos, it can do brand concepts, it can do mission statements and slogans and you know, it can, it can write exposes and press kits and stump speeches and marketing copy. But I think that you know, it, where it falls short is that everything that it can generate <laugh>, it's not necessarily good, right? Like, is that your experience too? Like when you, when you're asking it, you, I mean, you, you even alluded to this earlier, but like, when you ask it to perform a task, it's gonna do it at lightning speed and with a ton of scale, but like, it's not, it's not always on point, right?

Eric Wilson:

Yeah, it, it, it, it's definitely weird. You know, so some of it's in, in the training. So I, I I, I kind of go the other way. I don't want to discount it's capabilities now cuz it's, it's all about how you train it, what you give it to learn from. But, you know, to your point about art is having a, a conversation at a, at an art studio about AI and someone brought up the, the say, oh, this is destroying art or whatever. I said, well, is photography art? Because there were people the early days of photography who said, Hey, you're gonna put portrait painters out of, out of a job. And and so we think we all agree that, that photography is art.

Brian Athey:

Yeah, where I, where I came to really quickly was that it would be like, it would be like introducing steroids into professional sports. Like if you know, mid-season, you know, the MLB came out and just said, everybody can just like unregulated take all the performance enhancing drugs that they want. I think that it would seem existential. It would be like, this is gonna change the whole game until everybody is using it. And then there's a new, there's a new baseline. You know, it's like there suddenly there is this new standard where everybody has this tool and the, and the new challenges not competing against each other with the tools, but competing with each other against the ideas of how to use it.

Eric Wilson:

Right. The differentiator is what's your humanity, what's your point of view?

Brian Athey:

Casey Neistat, do you follow Casey Neistat, the YouTuber?

Eric Wilson:

I do, yeah.

Brian Athey:

So he did a, he did a video a couple of weeks back and he, he put in the prompt chat, G P T, and he says write a vlog script in the style of Casey Nesad. And then he goes to shoot that video verbatim. Like he follows that script verbatim. He follows the shot list, he says the words that it tells him to say, and it's hilarious, but it's also terrible, terrible <laugh>. And he, he sort of says at the, at the end of it, he is like, I think that this is a good demonstration. It's like, while it can do this, you know, you can check the box that it did it, it doesn't really have a soul. There's something in us, I think that can still differentiate something that is authentic and something that is contrived. And to the degree that, like, we've got a, a babysitter that watches our kids that we adore and she's in college and her professors are, have even started to adapt their ability to spot <laugh> generated essays and term papers, which I think is remarkable because it's so early. But like, imagine if you had a stack of essays to read through and you start to spot the patterns and the inauthenticities and you start to see like when the human actually writes. I think that that's an interesting thing that I, I certainly hope is pervasive amongst everybody. I hope that we get enough exposure to AI that we can spot it because, you know, it is so accessible that that's really like the, what's crazy is like how it could be utilized and who could utilize it. I think that's the big thing.

Eric Wilson:

Let me, let me steer us back to kind of the creative aspect of this, particularly as it relates to politics. Give us a sense of some specific examples of what those limitations are with the, the visual AI outputs. Where, where is it still struggling right now?

Brian Athey:

The outputs themselves there? There is a, there is a level of inauthenticity that I think that we've gotta contend with. I think, okay, here's, I I, I think I thought about this in, in in the sense that like, you know, if you've got, if you've got a hu a human and you've got, you know an AI analyzer and you're trying to call through a bunch of data, you're trying to do a bunch of data analysis, like the, the AI is gonna perform really, really efficiently. But like if you ask that AI to generate kind of an actionable plan that considers, like, considers everything that a campaign has to deal with cash on hand and you know, news cycle and what the opponents are doing and, and all that stuff, I think it's gonna fall flat. There's this gap between quantity and quality.

There's a gap between how much and how fast it can generate things and how actionable or even like practical or applicable that stuff is. So I think that where it's falling short right now is that like I could ask AI to like revise that same plan 20 times and it's not gonna blink an eye. It's gonna happily rewrite the same thing 20 times, but it still might not nail it. The human counterpart, they probably don't wanna do 20 rounds of revisions if they're moving quick and they've, they've got the campaign mentality, they probably don't wanna do more than one or two, but that plan is gonna be much more zeroed in and much more easy to execute.

Eric Wilson:

So my take on that is specifically like that example of cash on hand, which I just wanna zero in on, which is I think we're about to find out that a lot of human cognition isn't actually that special. I agree with you on the soul element sort of that hard to define thing, but the pattern recognition that we apply as political professionals comes from years of training and seeing this. And the more we can train the data sets where more we can quantify what we're doing, I think the closer it gets. So, so then the, you know, it's a whole new set of challenges. I think it ultimately frees us up for higher, better uses of our time, but the machine is always gonna be better at logical decision making than, than we are.

Brian Athey:

That's, that's if a campaign is always logical. Right?

Eric Wilson:

Well, and, and, and voters famously are not logical. Are not

Brian Athey:

More than

Eric Wilson:

Consistent. Yeah. What,

Brian Athey:

What do you think, what are your, I I mean, just to, just to kind of turn it back on you, like where, how are you using it and kind of like, you know, where are you seeing opportunities in pitfalls? That's what I'm, you know, I, I just level in the conversation.

Eric Wilson:

Well, so I, I think the, the best way that, that I have for, for describing this to folks is, is large language models, which is the technology, the branch of artificial intelligence that chat G p t represents. And what we're talking about here is really, really good auto complete

Brian Athey:

Mm-Hmm.

Eric Wilson:

<Affirmative>. Yeah. It knows what the next word to put in the sentence is. And that is really powerful. But it, it, it does have its limitations. It can only look, look backward. It's in the rear view mirror. You know, I think again, the, the smarter it gets, the more it's going to appear to be forward thinking, but that's gonna be really the, the dividing line. So I, we need to tackle the ethical questions of AI when it comes to campaign creative. We, we don't have any laws or regulations yet. So it's on a case by case basis, by professional, by campaign, by candidate to decide what is right and what's wrong. So I'm curious to hear what your framework is right now, how you and your team are approaching this to decide what's appropriate and what crosses the line.

Brian Athey:

This is like the question, this is a super smart question. I was hoping you would ask this. I think in politics, you're dealing with people who, whose primary objective is to win, usually at all costs. And so unfortunately there's not a clear black and white line. There's a lot of gray area, especially with the different levels of players in the industry. And I think that what AI does is outsource a lot of what professionals have been able to do in the past and put it in the hands of anybody. I think that, like, what is, what's really promising and exciting about AI is that like, you might not know how to video at it, but you can talk to a chat bot and it will edit it for you. Or you might not be able to generate a great graphic, but you can talk to AI and it'll generate it for you.

So accessibility, I think is really like where that ethical framework comes into play because it's, you know, while the, the industry itself is a very aggressive sort of all fair and love and war industry, I think the public at large, especially when you think about social media and memes and, and stuff that gets generated and, and goes viral, I think incentive structures are, are sort of aligned to encourage bad behavior in some ways, and that, that's something that I'm super concerned about. I mean, do you, do you track on that? Do you feel the same?

Eric Wilson:

Yeah, I, I think there's a, a certain degree of moral panic here that we here with every new technology, and that's gonna be the case here in politics as well. But I also think that political campaigns and politicians are reflections of the voters of the electorate. And so that's where a lot of the tension is in our business. Ethically, it is very hard to l lay out clear lines, but one of my clear lines is if you are using artificially generated media, so audio, video, visuals, it needs to be disclosed what you're seeing. So the, the RNCs add from a few weeks ago after Biden announced very clearly disclosed that these were AI generated images,

Brian Athey:

But the Trump being arrested by the N Y P D, those images that went viral, like those, those had no disclaimer other than the context that they were shared in, right? Like maybe somebody said, check up this AI generated

Eric Wilson:

E exactly. The, the person who shared them and originally said, I had, you know, the, the software do this. You're listening to the Business of Politics Show. I'm speaking with Brian Athe about AI creative and all things tech in 2023 and beyond four campaigns. One of the best descriptions, Brian, I've heard of generative AI like chat g p T recently is, is that everyone has an intern now, which again, i, I, I think is a good framework and, and basically saying it follows instructions really well, but may not have the ability to connect those dots like we were talking about. So in the op-ed that you've just published, you allude to AI being best leveraged as an accelerator for scaling ideas. Give us some more examples. I I'm gonna keep coming back to this of, of examples. So we make this concrete for folks who are listening and worry that it may be over-hyped which I think it, it, it can be, but what are some examples of how it helps you scale those ideas?

Brian Athey:

I do use chat G P t a good bit. Like I keep it open next to my Gmail tabs or my calendar. I talk to it pretty regularly and I really, you know, I've got <laugh> I've gotten into the habit of talking to it like it's a person. So I say like, you know, oh, thanks for that and good job. And hey, do you mind helping me? You know, where like, I, I really do sort of like have these conversations. You don't,

Eric Wilson:

Don't, you, you don't want it to resent you when it becomes our overlord.

Brian Athey:

<Laugh>, I've been a, I've been a director too long and I, I wanna make sure that I've I'm, I'm treating, treating the people that I'm outsourcing stuff to to do really well. But yeah, exactly. I wanna, I wanna make sure I've got some bonus points and brownie points. If in case things go south <laugh>, that's, I wonder if I'm doing that involuntarily <laugh>, that's, that's subconsciously that could, that could very well be it.

Eric Wilson:

I I really believe that you should not be polite or humanize the ai.

Brian Athey:

I can't help it. I really can't. But like, maybe that's, maybe that just speaks to how, how good it is because I really do feel like I'm talking to somebody sometimes <laugh> here's, here's like a really practical application. So I'm, I started as a graphic designer. I, I grew into sort of a, a brand and, and, and work on these big brand systems and stories. I create marketing copy and, and marketing pillars and turn that into websites and campaigns. I think that, like, what's really fun is if you, if you can generate something, and this is, if we take a small example like Adobe Express has just come out with their platform, that's sort of the alternative to Canva. And the idea is that if you load up Adobe Express with your brand kit and you know your fonts, your colors, your photos, then it will very quickly generate graphics for you.

So you can change the copy, but essentially like it, it turns a few small prompts into this visual idea. And then once you settle on the idea, you can scale it really quickly. Like you can turn it into multiple sizes and, and output it for multiple different channels. It's all about sort of that if you were, if you were having a conversation with it. The way that I do, what I try and do is like load it up with prompts at the beginning and I'll do things like, you know, read this press kit or, you know, read this article first, or you know, I'll pull a bunch of copy off of a website and dump it in. I'll try and like, I'll try and get it as informed and dialed in as I can. And then I'll, I'll give it prompts like, you know, pretend you are an expert or pretend you have a goal to, you know, whatever, it's raise awareness or, or raise funding or something.

And then I will say, now ask me questions. Ask me questions for clarity. And when you, when you say, talk to me and get all the information that you need to give an informed decision, and you can also give it prompts with like, and play devil's advocate. So give me a solution, but also tell me all the pitfalls to your solution. Those are really, really interesting ways that can like, really expedite the branding process. And if you think about, okay, so now I've got this big wall of content, if you can give it some type of framework, like there are all types of different branding frameworks. We have some proprietary stuff at push, but like, if you can, if you can say like, I need brand pillars and emission statement and you know, I need these things or these questions answered, it will generate those responses inside the framework based on what it knows.

So what you can really do is like, you can use what you already have to sort of inform chat G P T, and then it can help you iterate and fill in the gaps. And if you're willing to talk to it, you know, what I say all the time is like, Hey, that's great, but this is way too long. Can you gimme some bullet points? Or, this is great, but I only have 20 words on this website. Can you, can you drill this down to 20 words? That is so much faster than trying to do it myself. Like when I'm live in the edit or when I'm talking to when I'm talking to a designer and we need to trim stuff down quickly. But look, here, here's the other thing, Eric, here's like, here's what I, I use it for a lot. So I had this conversation with chat G p T and so I'm, I'm both a like a maker and a manager.

I, I'm in the design programs, I'm making stuff, but I'm also like on calls and in meetings all the time. And what I told it was like, I don't want to suffer burnout and I want to be more productive. So I tell it a lot of things about how I like to work and what my schedule looks like, and then I then I ask it to ask me questions to help optimize my day. So I, I use AI just personally as like, it, it's given me an entire plan. Like, you wake up at six o'clock, you go to bed at 12 o'clock and well 11 o'clock, I think. And you know, here, here's everything from when you should take meetings to when you should take to drink. You know, fill up your water bottle and drink water. You know, because I, I tell it like, look, my goals are to stay hydrated, hydrated.

My goals are to make sure that I take a walk every day. My goals are to meditate. And so it works that into a calendar for me. And so it, it, it really has the ability to kind of like plan my day. It has the ability to plan trips or to like draft talking points or like all these little inefficiencies that can be smoothed out and add up to a lot more efficiency. I even, I, I even use a scheduling app. It's called it's called Motion. And Motion is like, I don't know if you've seen this for like an ai, I played around with it. So it sort of like, kind of schedules your day for you. But I don't know, I'm, I'm really like, I think the, the real promise in a lot of this is automations. And I think that one thing I'm really excited about, you know, I, I watched Google's conference the other day and, and some of the things that Bard can do, and I really think, you know, we talked about this earlier, but like if AI can't solve anything else, if it solves email by just automatically generating responses for me and replying, like that'll be enough for me.

I think that we'll count that as a win.

Eric Wilson:

But that's like a big red flag and an indictment of email culture and behavior right now. If, if, if the robots can just start emailing each other and, and our jobs are done. So for, for folks who are listing, you know, you've mentioned a few tools chat, G P T, which everyone knows about Adobe Express has some AI loaded in, you mentioned Motion. What are some other tools that you're using on your day-to-day basis that our listeners could use to start helping out with their own campaigns or projects?

Brian Athey:

Adobe Express, you definitely keep an eye on Adobe Firefly. You know, Adobe Firefly is, it's a response to mid journey. You know, you can, you can do a lot of interesting things with photo editing with ai. There are some great time savers. Like if you've got like this podcast for instance, are you about

Eric Wilson:

To tell people they don't need to listen to the podcast?

Brian Athey:

It's not at all listen, but it's, it's, you might not need to edit it a bunch as much, you know, like editing

Eric Wilson:

A podcast, you get a summary of the podcast. Yeah,

Brian Athey:

I've been dropping so many F bombs that you're gonna have to cut all those out, right? So <laugh> no, but, but what I mean is like you, you know, if you wanna do quick video editing and transcribing like their tools for that, if you wanna do captions and resizing their tools for that, I think that the big thing, like I was talking about the automations, I think that what you're gonna see is, see, our, our limitations right now are time. We only have so much time to invest in something that is untested and unproven when we are working for clients in very high stake scenarios that we have made promises to, we know that we can deliver with our method and our approach. And if we automate all that away, or if we give all that away to a robot, who knows what kind of outcomes are gonna happen.

And I, while, while they do show a lot of promise, I think that we are not willing to risk clients campaigns and money and everything else just with experimentation. So we only have so much time to really build these systems and automations out. But I think that that's where you're gonna see things kind of really take shape. You're gonna see like people using things like make.com where you can, you know, you can get chat g p t to randomize topics, generate a blog on them, but then you can connect other apps where, you know, it can post that blog to your website, it can post that blog to LinkedIn. It can, you know, randomize and generate and all that is automated. You can literally just give the prompt to the machine and let it run a marketing cycle for you. And that's not guaranteed that it's gonna be good.

And I would, I would highly recommend that you double check you proof that and spot check it. But like, that's the, I think that's the big thing, Eric, is that like, this industry is so predicated on the notion that like, you have to you have to repeat and you have to, you have to inundate and you have to saturate and you have to be ever present everywhere on every channel, every TV spot. It's just quantity, quantity, quantity, quantity of message. And I think that you're already seeing some real downsides to that because, you know, you just look at the last cycle where people are just spammed to high heavens and they, they completely check out of the process. And to go back to the ethics, I think that if we have any ethical obligation, at least the one that I've, I hope I'm that I adhere to is that like I want more people engaged in the process.

And if we are, if we are outsourcing our jobs to AI that is mostly concerned with output and scale and not necessarily concerned with quality, I feel like the real implications are that we're going to drive people out of the process. And whether that's like a small thing, like you know, they get too many, well, depending on who you are, get too many emails or, you know, you're just checking social media channels and the bots have so taken over their fee, say, taken over your feed that nothing is real anymore, and you're just, you don't have a place to go to, like, whether it's to get the news or just disconnect, like, you know, I, I think that those little things, people aren't really thinking about how, like this overabundance of content and everyone's ability to access it and produce with it. I, I don't think people are thinking about the real implications of that. Because I mean, if anything, it's almost going to, I, I predict that it will have sort of this crescendo and then people will have to realize that like, I don't know if just, you know, complete and total message saturation is the way to get anybody interested. I think that there's a, you know, a little bit of mystery and a little bit of brand exploration and a little bit of brand ownership is, is just as important when you were, when you were building these campaigns.

Eric Wilson:

Well, I wanna thank Brian Athe for a great conversation today As we dig into this question about artificial intelligence and how it's being used in campaigns. You can learn more about Brian with the links in our show notes, including a, an op-ed that he has written that share some more of these ideas that we've discussed. If this episode made you a little bit smarter, gave you something to think about, it definitely did for me. All that we ask is that you share it with a friend or colleague and it's a win-win because you look smarter in the process. Remember to subscribe to the Business of Politics Show wherever you listen to podcasts, so you never miss an episode. You can also sign up for email updates on our website at business of politics podcast.com. With that, I'll say thanks for listening. We'll see you next time.

‍

‍