AI Panel

Okay, we're going to get started on this panel talking about AI. And really, this is a unique panel. I. I will say I have not confronted this issue before. We are talking about technology so deeply, and everything involves so many passwords, so many usernames that our panelists are all connected via Zoom so that each one can share their computer as they go through their slides. So if I get technical issues, y' all gonna have to give me some grace, because we are. This is a new one for us because we want to be able to show you what each individual on this panel uses, and we want to talk to you about what you're seeing now. And step just so we can get a really good feel for where everybody in this group is already.

So let's see if the slides are going to change. All right, so how many people on a regular basis does it feel like AI is everywhere? I mean, seriously, like, you're talking to people who are 2 or 100. It feels like everybody is talking about AI. And it sounds like some people truly believe that AI is going to permeate every area of our life in the next two or three years and send us all into pack. And then others are a little bit differently, looking for impact and what kinds of impact that's going to have. And then you have people talking about how it's going to totally shift our business. So whether it's the tools, whether it's the impact, whether it's business, all of us are looking at it from different standpoints. And I don't think anybody on this panel is expects everybody to walk out here with one key learning that's the same.

We hope all of you get some key learnings that fit your job, that fit the impacts that you need to have, that fit your business, that even fit the kind of work you do. Right. So let's talk about where you are in the world of AI. And this is the way. I think there was a moment where I was in, like, a sea of overwhelm, and it's like, I just don't want to think about it. Let me shut it out. Is there anybody in the room that's feeling that way? Because it's a lot. Any hands for anybody feeling that way? Okay, maybe a little bit.

How about people that just have some questions about AI? All right, that seems more reasonable. How about who has a few tools that you use? Okay, that's a lot of people that are using something and who is all in. Across my enterprise doing all the things when it comes to AI really mace I am shocked. It really is a good disclaimer. So I thought I'd talk a few minutes about where I am. Most of the panel is going to be these guys who are talking about it, but I wanted to just kind of give you my view. I'm really in the school of I have some tools I use and I really rely on some of them pretty heavily. So I wanted to just show them really quickly.

Last year, when Casey talked about Perplexity, I'm like, I ought to check that out. Well, thanks, Casey. Because now I pay for Perpletplexity as a paid version. I'm not sure how many of you guys have used it, so I'm going to show you a couple of things I've done. So I went in and, you know, I have all these great classes I've been doing. So I wanted to figure out if I was going to do a demo reel about my classes, what should it look like? And I asked Perplexity that. But I asked it one time where I asked it all kinds of things about demo reels before I mentioned Grounded Nag in a different thread, totally unrelated, I just said, hey, I need to do a demo reel for Grounded and ag. What would it look like? And I have to remember that if I help AI understand the complexity of what I'm trying to do, the answers are so much better.

So where I asked it steps about how to go through a demo reel for classes, what are the best practices? What should I be thinking about? What do I need to know about my business that I should showcase in it before I tell it what my business really even was? That got AI started on a train of thought that by the time I gave it to my URL and gave it my business, it gave me such great information. When I thought that AI could do absolutely everything at once, it was in a sea of overwhelm. And it gave me a simplistic answer. It really didn't reflect the depth of my business that it could. Those are the kinds of things you have to learn with AI. This complexity of the system makes a difference, and understanding how to put information in really does reward you with better information on the app. Another tool I use is Cast Magic and and that's partially because I do so much in video and audio and the way castmagic works. So Perplexity is a lot like chatgpt, but it gives you all the sourcing, all that kind of information along the line.

Cast Magic uses those different sources. Perplexity and chatgpt, they pay different Open AI people to do it. But it takes my video, my audio, and then it applies the AI to it to spot out exactly what I want. So in that you can see, this is a video I did with two farmers in Oklahoma, dairy farmers, that's Logan and Charles. Charles is the older gentleman. And it will give me immediately. It'll give me a transcript of the video. But then I can also ask it for specific things like.

Like I need a good title for a YouTube video. What would that title be? Based on the content in this video, what would the titles be? The cool thing about castmagic is it doesn't share my content out to the Internet. It's applying the Internet to my content and providing it to me. It'll give me the top takeaways from the video in those prompts. I can set up those prompts to run on every single video that comes in. And if you guys like, I'm recording this so we can even run castmagic and send that out to you guys just so you'll see what it looks like. Because you'll have key takeaways from this session, I hope, and you'll have a summary, a transcript, all that kind of stuff. But it does go into things for me.

I use it to do the writing summaries for the videos. Right? Like really what I'm doing is providing access to farmers. But some people like to see a summary of the video. Some people need to see the vocabulary that's in the video. So we put all those kind of things together and we sometimes copy and paste them into canva and make them look pretty so that they're downloadable. I can absolutely tell you the streamlining of this is unbelievable compared to where it would have been before I had castmagic. And what I love about it is it really is my content. It's helping me make my content show up in different ways.

I still have to check it for hallucinations and all those kind of things, but that's one of the tools. The other tool I wanted to mention, just briefly, is description. And it does really cool short social videos. And you can tell it how many short social videos you want. It'll give you the transcript of them and then let you say, I want to have like four videos about sustainability out of this hour long video, four 60 second videos and it will find you four that are concise, that are well edited and that hit the topic you want and it'll drop the captions on it for you. Descript D E S C R I P T And so there's Logan again, one of our dairy farmers. And, you know, it shows up. You've seen so many of them.

If you've looked at TikTok and Instagram Reels and stuff, and you can just make it so that it fits your color patterns and branding, that kind of thing. So that's my introduction for the panel. I'm gonna have each one of these folks because we've all introduced ourselves, but now you're gonna see the work they do, and that's when it's gonna really matter. So let's start off. Yeah, go ahead.

Can I do a landscape video?

Oh, yeah, you can do it either way.

But it take my 30 minute landscape.

Video and give me.

Every once in a while I have to drag it over a little bit, but it tries to line it up where it thinks it should be.

Perfect. Thank you.

Yeah.

How much upfront effort did it take you in terms of giving it all of that?

So with cast magic, that look and feel can come kind of quickly because you're feeding it in what the content is. Right. In anything. How you write the prompts really matters. And so a lot of my friends, we talk about how you organize your prompts, how you look at continually making them better. So, you know, on a lot of mine, I will say this is for an audience that's new to agriculture, that hasn't been out on farms. So let's make sure we use normal vocabulary when we're thinking about it.

Right.

You can do the same kind of thing with, I'm talking to a PhD audience. And sometimes I'll tell perplexity and which major extension services seems to have more abundant resources on pork production than anybody else. Because I don't always know which universities are doing the most work in those. And it goes through their websites really quickly and starts saying, oh, my gosh, did you know Iowa State, North Carolina State. Those are really big schools for that. You may also want to consider this one, though. It's pretty cool. All right, so does that get us for the questions to start? Each one of them are going to go through.

Use cases that are a bit longer and a bit more in depth.

So.

So mine is like the entry level, and we're gonna get a little bit scary for some of y' all. No, we're gonna start how fast? Casey, do you want to start on your. Okay, you go ahead and zoom. Hit Share your screen so we can.

Okay. Okay, here we go. All right, so I'm Easy Mill Swanson Russell. I'm technically VP Director of social media. But I do a lot of different things, including the AI stuff because it's being integrated in all the tools that we use, but also valuable tool that we found to do a number of different things. In my connection to ag, I grew up in southeastern Wyoming. Lots of people growing wheat out there. So that's what my ancestors did.

Moved out there from like Illinois and then they hit like a three year drought. So a bunch of people went back to where it was better. My ancestors were the ones who were either crazy enough or tough enough to stay. And so working out as a picture of the grain elevator that my great grandfather helped build, I worked there for a couple summers. For a while. My wife was a cheese influencer for the Wisconsin Cheese Board. So we hosted cheese parties. Yeah, absolutely.

That was a lot of fun. So yeah, I work across a lot of different industries, so I think that gives me kind of an interesting perspective on, on what we do and how we use AI. Back in February, I did an all day workshop on AI for the week boots. And so we went through and they had a kind of a particular challenge where they needed to adjust kind of their communication strategy. And they also, you know, needed some better research. So one of the things that we did is, is we used AI to help them with like, let's find some better research. Let's make sure like your research is up to date. They hadn't updated their website in a long time.

We took all the information that we gathered from AI and just put it in one data dump document. We used a bunch of different tools and then we used AI to again like take all that information, synthesize it, sharpen it up and come away with some good takeaways. So one of the first things we did is we said, hey, let's use all these different tools and take a look at like, what are the top five health claims, both positive and negative, about wheat from Paycheck, GPT, Perplexity, Gemini. I use open AI as the name for like free chat. And then we use Copilot, which if you have a Microsoft subscription, you get access to Copilot. And then we looked at like some of the differences between them. So you know, chat focused on like fiber, heart, health, like all those kinds of things. You know, they're pretty similar across the board.

Some of them like pull different things, but obviously like gluten is a, is the main challenge for them. So then we started to look at detractors and misinformation and disinformation. And I think this is maybe One of the most valuable things that we did is had chat play our, the AI play our detractors. So put everybody in groups and we said like, hey, let's role play and have AI act as a savvy online activist committed to informing everyone about the perceived health hazards of consuming weak boots. Asked if it could do that. So that's one of the prompt things that like Janice is talking about that I use as I say, hey, like here's what I want you to do. Can you do this? And then it will kind of give me, I can do this, this, this and this. But here are my limitations.

And, and so I understand what I'm, what I'm going to get and I know that the AI understands what I want. So then we did a follow up prompt. I said, hey, like map out a strategy for me, like what does this look like if we're going to soak down in the minds of consumers about the health of the wheat foods? And it came back with a really aggressive, really nasty playbook of things that are already happening to the wheat industry. It was very, very good at doing this. So then we looked at the flip side and said, hey, like, how can we change it up, right? Like given this is what's happening and you're, we're maybe not doing as well at it as we could. From the positive side, how can we flip the script here a little bit? And these are some of the ideas that it came up with. You know, a Rethink we digital campaign where you know, you talk about like we help one of the insights. And this is where I think it's like Janice talking about it's so important of like who is operating the AI and like what insights it has and what things that it's missing.

So all the wheat health stuff that everybody saw, it was all about like whole grain, whole grain wheat, whole grain breads, all that stuff that really dominated like the positive narrative. But they're like, the experts were like, what's missing is like these enriched wheat products that have all these other health benefits. But very few people were actually talking about that. So one of the ideas that came out that was a combination of AI and these experts is we're saying like, hey, well like whole wheat food stuff is just dominating search. And that comes up prevalently in AI because there's so much content about it. There's very little content about like enriched wheat products. And so one of the strategies we're going to start creating collectively as a group of all the wheat associations across the country Creating a massive content about enriched wheat foods to help build up a base level knowledge about that. Because if it doesn't exist in any kind of volume, it's hard for me to surface it and it's also hard for people to find out information about that because it's just not out there.

One of the other things that I thought was really helpful is a lot of organizations that get beat up a lot online, like they're, they, they spend so much time playing defense, they forget about playing offense, right? So, so you know, some of this is about like how you flip the script and get about like talk about things like the truth about we heard how or hit on some fomo, right? Like what are you missing out on if you help me out of your diet? Like what are the real implications of that? You can bring in nutritionists, you can bring in all kinds of different experts, but really start to make it a little bit more of a debate, right? We'll see. This is like watching like a good court trial, right? Like if both sides do their job really, really well, you're like oh well that's a good point. Then here the defense talking like oh well, like maybe there's some doubt there. And so it's about like if you're getting, if it's so one sided, how you come back and start to present your case in a different way with a more compelling narrative. And I think this kind of helped break them out of that thought cycle of what are we going to do about people that complain about like gluten intolerance? Like what are we going to do about people who are talking about celiac disease? And you know, some of the realizations that we had is like look like you don't have to say that celiac diseases like whatever, like it's less than 1% of the population actually have this. So like it doesn't apply to 99% of you. Let's talk about the great benefits of we then apply to the 99% instead of the fear around 1%. So that was a really helpful piece that we did for them.

The other study that I have, this is a little bit more recent, but we're using AI for risk analysis and specifically around like online safety. So we had an issue back in February. We do producer communications for the checkoff and all the stuff that they do is pre approved and planned a long time in advance because it all has to be checked out by usda. And then once ARMS folks approve it, we can't even really change punctuation. Like it's ready, it's ready to go and it's planned out really far in advance. The thing that that doesn't necessarily account for is that the world changes really fast and dynamics change very quickly. So what happened after the first of the year one is that a lot of the guardrails that used to exist on social media about stuff that would be filtered out or flagged as like violation of community standards or like too profane or whatever, those all went away. And so some channels like X became a lot more full contact sport than they were previously.

And so we had this really innocuous tweet that nobody ever paid attention to. Their Twitter X account. We had it, we were there to engage with like industry publications, that type of stuff. And it was a post about how do you, how do you do the assessment if you're selling a cow. Rancher, rancher. And used direct language from the act of the order. It was everybody's green light, green light, green light, green light, no problem. Well, what it didn't account for is how fast the dynamics were changing with things in Washington and those looking into USDA checkoffs.

And so this tweet goes out and it just blows up on us and it keeps rolling because one of the other things that's happening on social is there's a bunch of like narrative driven bots that go out and have a very specific thing that they're pursuing. They are programmed to go find these things, comment, retweet them endlessly. And so this is a thing that should have gone away in like 24 hours and went on for like four days. So we're like, okay, well we got to figure out a way that if this stuff has to be planned so far in advance, what can we do on a daily basis to help alleviate some of the risks? So we built this custom GPT called the Saint Beef Temp Checker. And what it does is it goes out and every day it will go read like current news. I built a knowledge file for it that includes all the, all the things that have happened in the history of the check off of who, who's the enemies, who are our buddies? Like, what are the potential issues? Because some of these things, like people, they grinded an accident 1985 and they're still grinding it. And so what are all those things? So what this thing does is we say, hey, what's the risk of posting as a checkoff today? And it gives us a DEFCON level rating of where we're at. It never goes below moderate.

It's always risky to post as a check off. And what I can show you here is let me switch the share to my desktop. And so I don't know if you guys can see this very well, but this is my team using this this morning. So every single day they go in and they say, hey, what's the risk of posting as a checkout today? So DEFCON 3, moderate risk. What are the key factors? So positive indicator, strong cattle prices. Yay. Beef trade expansion. Emerging concerns.

And this is built to just say, like, okay, well, what could have producers wound up that could cause them to say mean stuff to us? Here's what's going on. So high meat prices, labor dynamics and meat packing. Screw worm threat, the political climate. So it's like, here's what, here's what we recommend that you do based on all that information. Emphasize producer benefits, address biosecurity measures, and some things that checkout can't talk about and other things, like we can, but it gives you kind of a rundown of that. And then Kate, who's our community manager on there, she's saying, well, what about. Because it can also evaluate your post. So it's like, well, what about this idea? I want to make this kind of video post on Instagram.

What's the risk of that? And so it's like, that's DEFCON 4. Low risk looks good. So, you know, where you post on the platform matters. It's aligned with a trend. You know. Despite ongoing transparency concerns, producers remain broadly supportive of the checkup when they perceive it as promoting real, relatable ranching values. A video tapping into this sentiment can reinforce solidarity, not spark backlash. So it's giving us some pretty good information here that helps us think differently about what we're posting on a daily basis and kind of injects this real time element that we were missing in our process to, you know, really understand what the, what the issues are.

So we found this to be a really good use of it in terms of mitigating risk in a space that has infinite challenges in terms of figuring out, you know, well, here's the plan. But like, is the plan gonna get blown up today? Is what we thought was great four weeks ago still a good idea right now? And we found immense use in that. So I'm, I tend to be like the, the naysayer kind of person. And so that's the way I use a lot of these tools. It's just like, you know, and we do a lot of Personas and finally, like, we were using one on another account. And it said one of our, one of our characters that we created to help explain things. The audience thought it was a gimmick. I told my boss, I was like, finally, these things are mean enough that I trust.

Like, hey guys, intended to be helpful. Like I want to meet. Like I want you to help me see the stuff that I don't see. Like social media is not always necessarily a friendly place. So like don't just tell me. Great insight, Casey or nobody else thinks that way. That's a really great question. No, like I want you to tell me like real talk, what are the challenges? And we've been able to do that.

So yeah, I think we'll probably take questions at the end. Go ahead.

Specific what Casey's talked about and then if we have things that are more broad, we're going to have lots of discussion time and questions for that.

Yeah, so this, this very specific what you were just talking about.

So I used to work in CPA on the issues management team. Like I had to go through and read these articles and monogram.

I don't miss any of that check off approval.

But do you have to with this.

You were just showing the screen there.

You have to like push enter and make it rerun it every day. Or it just goes.

You programmed it so it, whatever, 6am Every day, it goes out and pulls.

In all that information and then creates that report for you. Is there some sort of manual work that goes into that on a daily basis to get that DEFCON report, whatever.

You have to click a button. But like, yeah, I haven't figured out how to automate it in like a sophisticated way where we'll do it automatically. Like most AI still haven't prompted. So we just made it a one button click where it's just like, what's the risk of posting as a checkoff today? And then it goes and pulls new sources and it's got custom instructions about like what kind of sources you use, which ones are not great, that kind of stuff.

And I would say so.

On top of that because. So you'll learn more about me in a few minutes. But my team is actually at our agency here with learning new tech in the AI space very aggressively. And one of the things that is now an option is Most of these AI tools have APIs that can be.

Integrated into third party tools.

So if you have a social listening or brand reputation piece of software, there's probably an AI API integration where you can take ChatGPT, plug it in, have your sort of reputation management Software do its daily download of brand listening and have some sort of integration with that now. So on the search side of things, we have crawlers that we send out to crawl websites multiple times throughout the week. We have Gemini and OpenAI integrated into them automatically and it can flag these for us that aren't normally like, readily visible unless we're combing through lines of a spreadsheet. So there's, there's new tools coming out every day. Another option is if anyone's willing to learn code. Python and SQL are ways to integrate AI into other tools. We have an actually really good computer science intern that is doing programming on Python right now to automate some of these like AI driven checks that we're doing manually. So a lot of potential out there.

It takes, you know, the driving curiosity and setting aside time to do it though.

Cody has a question.

Casey, you said you built this account. Is this obviously an account that your whole team has access to that you will go in and do it? But when you built chat, you're saying you just entered in prompts to finally get it to a point where you wanted it, or is there a way that you have essentially, you know, created a, a whole separate folder or tool or something that. Yeah, I can show you this real quick. Yeah, I'm kind of in the same boat, but I'll give you a sneak peek. So, yeah, so basically it's got a description and then it's got all these instructions. It's like a custom GPT, so it's got a knowledge file about all factors that can potentially influence sentiment. And then basically I just worked with it and kept refining the instructions until it would do exactly what I want and give me like consistent feedback every single time. So it's just about like kind of going in and continually refining and then you update it and you run another search and you're like, okay, well, I like this piece, but I didn't like this piece.

One of the other things that's really helpful is I just ask it to help write different layers. So I'm like, okay, here's what I want to do. Help me write the instructions for this GPT in the right order, in the right language so that you do this and you have like. So there's some things in here that are like mandated and it's like, you have to do this every time and if you don't do it, I'm going to be angry. And I literally like put some of that stuff in there because, like in the afternoons when the system gets like really heavily loaded. They can get lazy. And so I shared with this group, I have a screenshot where, like, I was using a custom GPT and it gave me just a bunch of ridiculous stuff. And I was like, you did nothing that I asked you to do.

You didn't follow any of the instructions, you, the knowledge. And he came back, he said, like, yeah, you're right. Bas had all the information and the instructions and the knowledge. I just didn't do it, like, supposed to be better than, like, the intern is that come over Thursday? So that was really frustrating. And it's getting better, but, like, that kind of stuff still does happen, so you just kind of got to be under toast. But, yeah, that's how we programmed it is. You just keep refining and refining and we'll refine it some more as time goes on.

So for your team's process, you were able to just work this into regular content planning. So it's kind of a day of where you're like, I'm just to go check it, then compare it to what you have scheduled or progress, because I'm guessing, like, you might do social planning or something like that over time. And then I was curious how your team responded to just using a new tool like that.

I mean, they were really open to it because the issue that we had in February was extraordinarily painful. So they all got an extra day of vacation. But yeah, so they were really open to it. And I was like, hey, if we. If we use this thing, here's what we build. We think it's like, you know, if it's 85% of the way there on its risk predictions, like, that's way better than where we were before. Right. Because we can do it quickly and we can, like, have a lot of information to make a better job by calling.

And they, like, Kate was using it this morning, so they're using it. Could you guys repeat the questions from the audience?

We can't hear.

Oh, yes, I'm sorry. So she. She had asked us about, like, team integration, like, how. How accepted the team was of the new technology and the adoption. It was very.

I think we're going to move on now and have Haley get started.

All right, well, thanks for the opportunity to participate on today's panel. I am new to the art community, so part of the first timers club at this event, but really happy to be here. And while I am new to arc, I'm not new to agriculture. So originally grew up on a farm in northwest Iowa. Diversified crop and livestock operation and My parents and brothers still farm there today. And like it's already Ben shared a little bit. Before I dive into the case studies that I've prepared to share today, just wanted to note that I do think it is really important to acknowledge that we do all bring different lenses and perspectives to AI based on the work we do and the different experiences we've had. So do just want to share briefly a little bit more about myself to help give you some more background and context on some of the ways I think about AI and some of the different questions I think about as well.

Thank you. Perfect.

Thanks.

So starting out proud. Three time graduate of Iowa State University, studied a Bachelor's in Agricultural Communications and Journalism, followed by my graduate degrees, master's and Ph.D. and AD COM. And first off, boy, wouldn't it have been nice to have a tool like AI. As I was working on my thesis dissertation. Really missed the boat on that one. But as I was reflecting on that, I also think about how you even for my parents in their generation, they didn't have the computer or Google powered search when they were going through college. So it's kind of funny to think about the evolution of that.

But the other reason I bring up grad school is.

Can you pull your microphone?

Sure.

Maybe I'll set this here so I want to hang on to it too. Oh, I think I'm stuck. Thank you. Didn't wear a collar today, so the badge will have to do.

Okay, let's try that.

All right. Is that better?

Maybe on the side.

I just didn't want my hair to like fly into it and give you all some feedback. All right, sounds good. But the other reason I bring up grad school is my research study actually focused on career preparedness and writing education of agcom students. So as employers, I do think it's also important for us to be thinking about not only how are we integrating AI into our workplace, but also what skills do we need to be preparing the future generation of bad communicators to be really successful in an AI driven future. So that brings me to my current role. Oh, excuse me. Before I jump to that, also wanted to note, I did begin my career as a one woman communications team for the Coalition to Support. I Was Farmers.

Some of you may be familiar with the organization. And for anyone in the room that might be part of a smaller team or even a solo team, maybe you're already thinking about ways that AI can really serve as that extra set of hands in your workplace. I think AI has huge potential, especially in those small team environments to be a really powerful assistant, whether that's performing administrative tasks or even repurposing content. So now jumping to my current role. So I'm currently a senior content public relations strategist at Meeks. We are a full service advertising agency serving ag, food and health brands. And I've been with the agency around nine years now and I spend a lot of my time doing long form writing as well as earned media outreach. So that's really functionally what I guess has informed my use of AI.

Slide.

Thanks.

So that brings me to my case studies today. Go back one, please. Thank you. So at Meeks, I am part of a group called the GPT Collaborator Group. That's kind of our formal name, but informally we like to go by the AI Outlaws. Just have a little fun with it. But this is a group that is made up of different people from across our agency serving in different functions and roles. And we meet on a regular basis to share ideas, brainstorm and experiment with different ways that we can integrate AI into our client work.

And so Casey's already kind of touched on this already, but one of the ways we've really dived into AI is creating custom GPTs. So that's what I'm going to be sharing with you today is to specific GPTs that I've created in the content PR space. Thanks, Kelly. Okay, so the first case study I want to share with you today is custom GPT I built as a press release generator. So in working with our clients in the media space, we've been really successful the past few years, kind of helping them transition from sending out, you know, these single comprehensive releases to more customized local versions. But as you can imagine by doing that it also creates a little bit larger volume of work for us to be executing on. After all, if you do want your story to reach local outlets, you need to give them a local story. For the second year in a row, we were actually tasked with sending out 60 plus versions of a press release across a five state geography.

Again, a lot of volume of work and you know, each version of that needed slightly different details. So in this case it was things like dollar amounts as well as the counties that were being impacted in each of those versions. And the first year that we did this project, we actually built all of those out manually, which you can imagine was a very painstaking process for writing, reviewing, and then manually uploading all of those into our distribution system. And so for this year, I thought to myself, there's gotta be a better way to do this and so that's where GPT played a role. And we started exploring ways to make that process more efficient. And so my initial idea was to create a GPT that could create all of those releases from scratch. The problem with that was the more information I asked it to provide provide, the more transfer errors there were. So with that and some tinkering, that's when we pivoted.

And instead of having creating a GPT that created the full release, I actually gave it specific instructions so that it would only create the variable parts of the release. So things like the headline, the date line, the specific dollar amounts and counties that I mentioned earlier, and everything else in the copy or body copy stayed the same. And so the solution was a custom GPT again to build just those variable parts of the release. And I was able to do that by giving it those specific prompts as well as uploading documentation that essentially made the GPT serve as a database based on region. So I could, I'll just use the example of the town I was born in, HeyGPT, create the localized version of the release for Emmitsburg, Iowa, and it would spit out all the relevant details that I needed for that version of the release. So of course this approach really reduced the manual efforts, helped ensure accuracy. We were of course still had a human element to it where we were spot checking and reviewing everything, but did speed up the entire process, even with a tight client approval window. So.

But the best part I think for me about this GPT in particular, is it also opened up accessibility to my team so I could actually share this GPT with other people on the PR team or even beyond that. And we could all be simultaneously working together to create these different release versions. And it fell on no one person's shoulders to get all those 60 plus versions created. Anecdotally, I'll just share the release. As far as results, we were able to generate over 100 earn media placements for that approach. And so, you know, really big one for our clients, big one for our team, and obviously a great fruitful experiment with AI as well. Next slide. The second case study I want to share with you today is for a strategic messaging project.

One of our clients is currently undergoing a lot of transformational change across their organization. And so as part of that journey they, you know, they've needed clear, consistent messaging that they can share with their internal and external audiences. So both customers as well as employees. And like with any good messaging project, we've had a wealth of information to work with. So lots of things like Input sessions with their leadership team, stakeholder interviews, strategic vision documents and other internal documentation. But of course, we're all human and sometimes that information becomes, can be kind of overwhelming to digest and remember all of the relevant details when it comes to actually writing the messaging. And so for the solution for this custom GPT, we essentially created one that would be trained on all the relevant source material. So it essentially became a living resource library that we could plug in all the inputs and then as we were developing the messaging, quickly pull from past examples and kind of weave together, you know, that cohesive narrative using the right brand, voice and tone.

So for examples of some outputs from this, early on, we used it to help create some CEO talking points for a series of customer meetings. It's helped us develop some of that primer messaging for market conditioning. So, you know, as these changes are about to roll out, what are those things we can be saying now to kind of get the audience primed and ready for those rollouts? And then as those changes do continue to evolve, you know, if there is additional information or inputs, we can of course add those to the GPT and build on the foundation that we've already started. Next slide. So that was just a couple quick examples of ways I personally integrated AI into my PR and content work, but did want to just briefly touch on a few other ways that we use AI at our agency. Probably the one that's a little more differentiated to us at Meeks is something called sam, which stands for Simulated Audience Mindset. And this is a relatively new AI power tool that we, our Data Science and Strategy team has rolled out to help agricultural marketers test and refine messaging. So it essentially simulates a farmer in real time.

So they, the GPT is our farmer and we can put plug in things like headlines, copy imagery, and have it respond to us and tell us how well that message is resonating. Casey already touched on Persona power copywriting a little bit, but I think one of the best use cases at our agency is we actually had a copywriter who was tasked with writing 400 recipe descriptions. And so we used a custom GPT to help manage all of that. Obviously a huge amount of volume of work for any one person or team. So that was, that was a really smart use case for it. On the video front, seen some really cool things for our team from creating or using still images to create motion video. It's pretty wild to see that take place. But another way in the video space we're also using it is we actually cloned one of Our team members voices and so it presents an opportunity to, you know, have a more affordable option for some of our clients instead of hiring talent.

A couple other things here as far as research synthesis. Our data science team has been exploring different ways to do quicker thematic analysis with AI so being able to find those, you know, patterns in the data a little bit quicker. And then last but not least, team enablement. We've also been experimenting with some onboarding resources and supporting every dot operations. You know, essentially building a GPT that can serve as that additional resource that team members can go to for things like hey, how do I enter my timesheets for that? Or what were the specs again for that print or specific printer digital ad that we needed to create.

There is no AI tool to mop up the water if you were to have a spell.

We have a quick response time.

The final one. So thank you for your time today. Just wanted to close out by saying, you know, I think AIA is really exciting frontier for all of us and I really don't see it as a, as a way that it's replacing human creativity. I really see it as more of a tool for amplifying it. So whether you are from a small team or a large team, there are many different ways to make AI work for you. And really it all just comes down to a little bit of imagination and some experimentation.

Okay, I'm going to take this one from you. And if people have capacity games, you can do the Q and A with this.

All the options.

I'm just kind of sharing what your how far your team is really thinking.

That.

How did you create the custom.

AI for all these options and then.

How are you making sure it's safe with your clients information? I know that's one objection at our company is pointing for text our company information.

And so I could see, you know, just running into that scenario would be.

Like oh, can't, can't run like you know, CEO messaging maybe or something, maybe something that's internal.

Right.

I'm just curious how you handle that.

Yeah, that's a really great question. And Casey kind of already showed you the back end a little bit of what inputs we're using for the custom GPT side. And so that is the same way that we're developing those. Unfortunately I can't share them with you today. Data client confidence. But to also expand on that, we do use the paid version of GPT which does not put your information into the ether. It keeps it all safe and secure within your platform. So and I guess to share a little bit more about that too.

Just as part of this GPT collaborator group, we've had team members experimenting with different AI tools and language learning models. And we really just landed on on ChatGPT because a of the security and then also functionally we just found it to be one of the most responsive tools available at this time. Any other questions?

Hey, we'll put Kelly on the spot. We're gonna have. We have some great conversations to talk across the panels as well. And I'm so Kelly, let's go ahead with yours.

All right. Can you all. Awesome. It's always nice when you start with a mic that works. I'm gonna stand up just to say hi so everyone can see me, but I'm gonna stay at my computer.

I'll fight you.

I guess I remember five feet taller is the reason I'm wearing heels today. So everyone can see me, but I'm gonna say it my computer because we very well might go into the wheat needs here doing some fun stuff. But my background, if any of you were involved in the very first Ag Chat, I was Kelly, the intern at the first Ag Chat conference. I got my career by accidentally getting involved in social media and ad. And then I took some left turns. Ended up doing enterprise search strategy at Hallmark for four years, then was kind of act incognito SEO for five years. As a freelancer, worked on projects with.

Actually multiple people in the room.

My name was never mentioned anywhere public for any of that work and I'm okay with that. And just recently went back agency side. So now I'm with Native Digital here in Kansas City. We're just two miles away downtown. And my role as a senior discoverability manager is getting things found on the Internet wherever users are. So we do what you would consider traditional SEO, working with Google and Bing. But I'm also in charge of understanding how people are using AI to get their questions answered and get their materials found online. So that's kind of the lens.

I'm going to be taking this from all of the tools. Talk has been incredible. We're using custom GPTs in our agency to create efficiencies. But really I want to talk about the external factors of how AI is impacting as well. So I'm actually going to challenge some of you. If you've got computers open, I want.

You to pull up Google like right now.

Do it.

Let's go.

I don't care what you're going to Google. That's between you and God. But ultimately what I want to do is see how many of you get an AI overview at the top of your Google search. Because as of right now, Google is reporting that 15% of all searches are receiving an AI over views. That ain't that, it's more. Well, here's the thing. Google customizes your search based on your your behavior. So if you interact with one AI overview, you are more likely to get served more AI overviews.

I get one almost every single time I search now because Google knows that I'm interested in AI overviews. Whereas someone who goes straight down to the normal, you know, blue links in Google, they're going to get thirst Google links more actively.

I'm not. Everybody got that.

So, all right, if you've googled, raise your hand if you have an AI overview. 3, 4, 5, 6. Raise your hand if you googled and didn't get one. Wow, that's like a pretty good half. That's a good 50. 50. Well and it's just a. So the thing to keep in mind though is Kubo customizes literally every search.

So your behavior, your geographic location, even the intent behind your search is going to impact how these AI overviews show up. We find that they're less likely to occur in things like transactional searches. So if you're looking for a specific service or a product, Google knows that the AI overviews aren't going to be helpful for you. Whereas informational searches, how to's, recipes and one thing I do want to point out on a screenshot before I say anything else, we now, okay, great. There's this little button here that looks like a link. That's because that's what it is. This is a relatively new feature. A few weeks ago if you got an AI overview, it was just Google spitting out information that it stole from other websites.

Now it's actually citing those websites and we are learning how to synthesize the search engine and fast. And the traffic that these links in the AI overviews are actually generating, still very black box. We're still learning. That said, all of this comes back to something that we call zero click search. If you are searching for something on Google and it gives you all the information and you never need to click off of that search Engine results page, that is a zero click search. We know that those AI overviews when they search are accounting for roughly 30 to 70% of the clicks happening in search search on the search engine results pages. So one of our longest standing jokes in the industry is if you want to vary a body, do it on the second page.

Of Google results.

That said, it is now becoming more and more difficult for even that first page to matter because zero click search is pushing things even further down the page. So just know that it's not going anywhere. It's getting like slightly more helpful because they are including links, but it's drastically changing the landscape of what people like me are doing each and every day. So Google's changing. We're trying to adapt the other way that organic.

Yep. Question Is there a way to to search on things to defeat your to to not draw based on Google's group analysis of all your preview searches, get a true unbiased access search.

So he asked if there's a way to get a completely unbiased search without any of your personalization. You can do something called an Incognito window. If you're using Google Chrome instead of opening your existing window, right click on Google Chrome and it'll ask if you want to open an Incognito window. That's how us house pros are getting unbiased search results. And you're a pro. I've worked on SEO projects with you Janice. Don't be honest. We know that's not you.

So anyways, like you can do the incognito and that will basically ignore any existing data that the browser has stored on you and gives you a completely fresh search.

That's an awesome question.

We do know that in the last year or so the percentage of actual searches that are happening on LLMs large language models like ChatGPT pod perplexity, it's anywhere from 5 to 10%. ChatGPT is telling me the average is 7, so that checks out. One thing to note though, if you're looking for information on AI model, they're not self aware. So when you ask them about themselves, ChatGPT goes out to Bing and then searches for itself to get information. And we have tested this pretty exhaustively. We had a working session last week where we were digging into the actual code that is visible publicly on ChatGPT as you're using it and we identified.

A line of program and we then.

Asked it about that program and it baked into its code and it said I'm sorry, I'm not familiar with that. So we know that they're not self aware. That said, you can still ask ChatGPT about itself. It will have useful information. Just know that it's getting it from Bing, not from itself. So with 10% at max of searches now happening in these large language model platforms like Chat GPT and Complexity we know that we have less control over what happens from an SEO standpoint. We are finding ways to impact that. And what that is doing is one making us very uptight about a lot of things, but also it's causing us to think outside of the box and realize how little we actually have control.

So SEO is basically you're a boat on the the oceans of Google. You can prepare and you can do everything right and you could still get swept away by a storm.

You have no income all over.

And my job is, you know, risk management. It's loss prevention, it is reactive, but it's also proactive and LLMs are really shaping that up. So I'm gonna see if I don't butcher this, analyze the five stated in February of this year that we're projecting the market share of searches happening in LLMs to rise to 15% by 2028. The fact that someone's coming up and scaring Google is amazing that they have had 85% of the search share in.

The world for years.

So with the idea that and this includes even Geni, which is Google's AI search platform with 15% of it projected in the next three years to be shifting entirely to chatgpts and clause approach complexities and these other AI bots is just industry shaping for us, but it's also really exciting. So how many of you have ever done anything with paid media? Google Ads? Okay, so just making sure this is not completely irrelevant or uninteresting for y' all, but that zero click search impact also hits on the paid side as well. So just be aware of that as you're thinking about how to purchase your ads in search. The other thing to keep in mind, Google makes more money if brands are going directly to them instead of paying experts. So you will see in the coming years and it's already happening, you will see Google going directly to your clients and saying, hey, why don't you just run some automated ads with us? And they are called pmax. They have a place. Usually that place is in the garbage can. A lot of my job security as a freelancer prior to going agency side was coming in and helping small businesses and nonprofits undo the damage of trusting Google's AI.

Ad platforms are not necessarily bad thing when used responsibly. But you have very little access to your own data. You have very little as far as customization and levers you can pull. So just know that that's a gotcha that the, the big tech companies are doing right now is they're trying to Tell you, oh, this is automated. We use AI. That doesn't always mean it's better than trusting a human. And I'm not just saying that as someone who gets paid by that agency who wants our clients to pay us. I'm saying that as someone who spent countless sleepless nights undoing the damage of pmax ads for small businesses that could not afford to lose that ad spend.

So paid search is impacted most, but we do see it on the meta side as well. With ad buying, pretty much any platform that you pay for ad placement on is going to be coming up with some sort of AI solution. In addition to the ad side of things, though, when you're purchasing the ads, there is the impact that the AI has on the way the ads are served. And we see that here as an example with the Chase bank app or the Chase bank ad being shoved down a third of the page by the AI. But this is where things get fun. I actually have a window pulled up to the side of this, this deck right now where I'm doing this exact thing. We can get into the code of what AI, especially Chat jpt, we know how to do it specifically, we can get into the code of what they're thinking while you're asking the questions. And I can actually pull up what it is asking Bing while I'm talking to it.

And I can better understand if I have a client who wants to be the source that ChatGPT goes to during a question, how do I position myself to do that? So when I say that the SEO industries is really fighting trying to up with this, that's what we're doing. We're doing working sessions where we're staring at code and thinking about what is Chat GPT looking at. And I keep saying Bing that it's because most of these large language models, except for Gemini, do use Bing as their default search engine. So you ask Chat GPT or Perplexity a question, most of the time they're going out to Bing to get the answers for you. There are some that you just Google mainly, mainly Gemini. Some use it in a hybrid with Bing, but they're searching just as desperately as we are for answers and their queries are messed up. So in this case we can see the rational says to help the user choose a bank. I'm considering a mix of local banks like bank of America, Chase or Commerce bank that have physical branches in Kansas City.

That's what it's considering as it's typing up its queries. We know that it's considering my Location. I did not give it my location. It just knows I have GPS active on my browser. But it's asking questions the way we are and then synthesizing that into these queries and they're usually really complicated and change the way we think about keywords now. It's fun. There are ways to create stability. If you're looking to show up in these LLM citations, they're going to be technical.

That's where people like me probably come in and help. I'm fairly certain no one in this room gives two cares about schema markup in their website's code on a daily basis. Whereas that's what I'm going to go have a meeting about that in an hour and a half. Image alt text though, is probably something all of us have thought about this once. That is something you can do to ensure that you have a little bit more stability in a world where LLMs are becoming a more consistent search engine. The other thing that goes so underappreciated in pretty much all segments of digital communication is your content structure. How many of you have standards for brands around the way you use header tags? I have some hands up. I like that.

That makes me really happy. Header tags are one of the most important organizational tools there are for getting crawled by traditional search engines and AI bots. So make sure that when you're out there building your content, you have one H1 per page. Everything that is a second level of content, it's an H2. Everything that nests under an H2 is an H3. We call that sequential headings. Just be a good steward of your information architecture and you're going to have a better eco system to be found by both traditional search engines and AI. The other things to consider are more strategic.

The softer side of things, information, architecture and format still very important. We know that there is an art and a science to the way you structure content on a page. But there are certain types of content that search engines and AI both prefer. But we're finding that in the case of AI, they like research studies, data and summary. So if you have statistics, statistics, if you have just percentages, if you've got charts, tables, anything that is hard information that an AI bot can easily understand, they're going to defer to that more than a softer informational source. Q and A. They love Q and A. Not just because it's something that they can easily spit back out, but Q and A allows their reasoning to better understand the content that is being processed.

Someone asks a question, they get an answer. The, the reasoning, the actual programming, the, the thinking that is happening inside of the ghost computer is able to wrap its head around that and understand the context of, of the information better. Timeliness. We found that most of the queries that ChatGPT has been doing in our research actually include the year. So if you have say, a listicle or something, it's like, you know, best grocery stores for high quality local meat. I don't know, I'm awful at tossing out these off the top examples. Please forgive me. Best grocery stores for buying local meat or whatever in Kansas city.

Add that 2025 on you update that list every year. Making sure that it is timely and, and frequently updated and current is one of the best things that we found, especially in our finance industry clients, because that is where we're, we're doing the most of this work right now. And then pr. I tell you guys how important that shit is. That's why we're all here. But if you're doing good pr, you're building top reputations, you're getting linked to on a third party site. Even though we know the bots are going out into search engines searching for things, they are more likely to cite a website that has been linked to on other websites. Now that's an important part of SEO as well.

We want backlinks because that's how search engines also understand that your website is trustworthy. Their site is linked to you. That must mean that you are worth listening to. So some things change, some things stay the same. Backlinks, if you are one of those things. Now I want to hit on something kind of funny that came up while I was prepping for this. And this is more about choosing the right tools and also not trusting Chat GPT as much as I think we all want to. And I don't know if I showed this to you guys, Janice and Seamus, we were texting real time while this was going out.

So generative AI is awesome. There's, you know, obviously some ethical concerns about using it. There's, you know, compliance and quality concerns. This is all about quality and mostly less. I needed a head. I haven't gotten new headshots for my new job yet. I just started three months ago, but I had this nice picture I took on the day I interviewed for it. So I was like, chat GPT, do me a solid.

I need a headshot. Here's the exact query I asked. Here's what Chat GPT did. All I wanted was a white background and it made three different people. I know I like Your color. I don't own it. I would like to know what that is. So I then turned around and I didn't include it here.

I forgot to drop that into another slide. I turned around and went to Canva. Now, Canva has its own AI solutions as well, but Canva has built some specifically for image editing. So you can go in and say, hey, Canva, I need you to do XYZ to this image and it gets it better. So what did Canva do? It gave me the headshot that is included in the program. That headshot with a white background based on a selfie I took on a.

Day I had a job interview.

So that is the lap I wanted to finish everyone out on. That's my last slide. But if you are going to be using these tools, which there are some incredible case studies talked about here today, we actually have a compliance bot that we built for one of our heavily, heavily regulated finance clients. You know, when you're working, I, I joke that I feel equipped to work in the finance industry because I worked in ag and Vet Med in the past. Like, I know a thing or two about clients and my bank clients are always like, great guys. I love that because I know what it's like to jump through those regulatory hoops. So we built a bot and we can drop ad copy or website copy into this bot. It that's it for brand.

Brand standards. If that's it for legality, it sets it and tells us whether or not we need to add additional disclaimers to creative. And we have it built so that it actually gives us like a great check marker. A red X is like a pass fail. It's awesome. So obviously, great tools, awesome efficiencies you can build from just getting stuff done standpoint. So make sure you're using the right tools because sometimes you just want a head shot with a white background.

Brandy.

All right, I need to get my face off this screen now.

Take a seat. Questions for Kelly. And then we'll move into broader questions. This is very, very detail specific. I'm asking because you manage my website. So with the. You say the H1 tag, like a header tag. Is that like when you're creating content, you have like some text box or the formatting box.

And.

I'm asking about the H. Is that on. I didn't. I'm asking about the H1 tags and the header tags. Is that like when you're writing a blog post or content on a website and there's the formatting box and it gives you the option heading one, heading two.

So Brandy was asking, when I'm talking about H1 tags, I should have checked my jargon in store on that one. It is heading tags. So in her case, she has a WordPress site that I know very intimately because I felt it, but.

I broke it.

I built its current iteration in WordPress. There is actually a text formatting drop down that says, you know, regular text heading one, heading two, heading three. That's what I'm talking about when I'm saying H1S. The reason we in the industry call them H1S is because that is how they're coded. When you're looking at the HTML, the tag is an H1 tag.

Thank you.

Now I know that.

For the mic runner. So, Kelly, from your experiences with clients and freelancer on your own to now, do you have a favorite AI tool for checking the authenticity or rich natural of copy?

Oh, man, that's an awesome question. Did you any help here, Aaron? Okay, so she asked about authenticity checkers. What's the AI tool for? For verifying that something is written by humans and authentic. We. We actually pass it back to Chat GPT for that. I mean, don't reinvent the wheel. You have to. To engineer your prompt really specifically to say, you know, this piece of content needs to be audited for authenticity.

Can you check for the common, common indicators that this piece of content was written by AI? Sometimes it is incredibly obvious. I actually, my boss dropped a article on my desk a few weeks ago and said, hey, can you do me a favor and like mark this up?

It's about your client.

And it was obvious within like two paragraphs that it was. Was written very, very poorly by AI and he wanted to see how fast I got it. So I had a glitter pen, as one does, and I wrote what the actual you know, what across the page to prove my point. It's like you got that faster than about people. But yeah, we're having fun with that.

Keep it simple.

Yeah, yeah, Chat JPT will do it. You just have to give it very clear instructions.

Oh, good.

I'm gonna get my steps in.

I think it's on. It's very, very directional. Is it? It is directional, yes. You have to like be like directional. So can you. Is this. Can you hear me? Is this working? I don't think it's. I can just see it.

So there was a reason. So there was a recent study that AI is actually making people dumber.

My husband and I just talked about.

This because it is saying people are going to use you know, cognitive thinking, critical thinking, they're giving us those skills. So what do you all have to say about that?

They're not wrong. I don't think they're wrong. So full transparency. My husband is also a digital strategist. So we come home, talk about search engines over dinner. My kid actually talks trash by asking people to show him the data because he was raised at home. Digital strategist. So we have talked about this anecdotally.

We both agree that that is a very real risk for our society, but it's also not going anywhere.

I would say calculators were supposed to make us stupid when I was, you know, like there was a time when nobody would know math because calculators think it's stupid. Some math is needed. So math wasn't necessarily as important. You can't not learn math.

Right.

But you can use calculators and computers. I think people are going to learn that AI. You can use it to let yourself become an idiot or you can use it in ways where you're getting smarter as it gets smarter at the same time. Right. So I think the balance is up to people more than it is the tools. I would, I would agree with that.

Not just to expand.

Expand.

I missed the nick on the panel.

There you go.

Okay, just to explain about a little bit too, even in our oh, GPT collaborator group, we've talked about how you can almost get so used to using it that for even simple tasks, like a coworker, give example, like he needed to send a text to a group chat and he actually just had GPT write the text and he realized it probably would have been faster if I just wrote the text in the first place. And so I think it is a little bit that self regulation as well in terms of like how you're using it and in what applications because you can get in that rut, I guess of kind of depending on it too much as well. So again, the kind of self regulation piece of that I think is, is important. I love the, the word self regulation in regards to it.

All right, so Molly. Yeah, and this might be a question for everyone. So I don't. But Kelly, you were talking about this. So we're talking about how we're optimizing websites for AI. Has anybody been optimizing websites in the back end for AI, but keeping the human experience on the front.

So I want to stress just because we know how to optimize for AI does not mean we ever stop building sites for humans. My entire career is based on getting things found on the Internet. I live and die by the Google overlords. And I still like the theory of human first. User first. SEO is what we build our entire practice on. So, like, even if we know how that is, how, how we can appeal to LLMs doesn't mean we aren't putting humans first.

I guess my question is more so, like, do we even have to make that LLM optimization on the front end?

Not necessarily. If you're, if you're doing, if you're building a technically sound ecosystem, you're already like, hedging your bets to perform better in AI. It's just this constant balancing act. Because, like, I, I have actually there's a practice of over optimization in SEO and it's where people lean so hard into the technical side of appealing to AI and the Google bots and everything. I made a lot of money as a freelancer undoing over optimization. So the pendulum is swinging back towards user first. Usually the people who were in that side of things were very unapologetic and didn't care about the ethics or the accessibility of it. There's always going to be bad actors in every agency or every industry.

And that was SEO's most recent group of them.

Yeah. I think the part of the opposite end is for a long time, Google was going to sites because it got a lot of traffic and people read a specific page or two. Yeah. And those pages don't have good headers used in the right way. Like, people choose them just because they like the size of that font better. And I'm not joking. Like, some people go, oh, that seems too big. Well, if you start your page with an H3, then it never had a headline.

Yeah. Right. So some of that integrity of understanding how systems work is now way more important than it used to, to be. And that's, that's why some people are seeing sharp declines on traffic. It's not because AI is doing something different, it's just because now it's really getting a full picture and Google's taking more of a full picture for its search. Right. So there's those kind of things coming into it too. Yeah, yeah.

The header structure thing in particular is both good for machines and good for se.

Right.

Alt text. So one thing to keep in mind, most SEO recommendations are actually good accessibility recommendations too, because a lot of the technology that accessibility software like screen readers for the visually impaired are using, it's the same way that Google Bot reads our site. So like, we, we often say that if it's, if it's Google friendly, it's user friendly. An accessibility standpoint. What's good for accessibility is good for SEO. My team is actually the accessibility consultancy source for our clients. If they have concerns about getting sued because of ADA guidelines, we are the ones that help fix it because there is so much overlap. So there's very real concerns about whether or not you're doing right by humans if you're.

You're optimizing to search and AI. But to Janice's point, good design very often now aligns with what Google wants as well because Google wants at the end of the day, happy users. And if we trounce all over the users to appeal to the bots, you are going to get penalized for would be so interesting. Casey, you were talking about how you.

Fed into your GPT the don't do.

This or I'll be angry to have.

That coded in the back end.

We have a We teach our interns to say if you don't do this, I'll lose my job or I'll get fired. Because ChatGPT takes that very ser.

My thing on ChatGPT is anytime I give it an input and I give.

It as many instructions as it can, the last thing I always add is.

Ask me any questions you have before responding to my prompt or how, how.

Can you improve upon your last response is another we always go back to chat and output say hey, look at this. How would you improve this if you were reviewing for someone else?

Yeah, I love that. Thank you. Sorry to monopolize.

So I'm still very new in my role at my organization and my biggest thought is AI and ChatGPT feels like a dirty word or a secret that I have to keep from everyone that I'm using it. So my biggest question in case you touched on this with your team, but how did you or do you go about getting organizational buy in for utilizing these strategies without people that are higher up coming back with well, we don't want to do that or we have a certain kind of rule set that we have to use. I guess the biggest question is just that organizational buy in and oh wow, organizational buy in and making sure that everyone's on the same page when utilizing. Yeah, that's a really great question. And I think for us, at least in our agency, we decided early on that we're going to be, we're going to embrace AI very quickly and there are guidelines around how we use it. But I think one you like want to be upfront about using it, right? It's one of those things where like, if you're using it and you didn't disclose that you were using it, if somebody finds out, then they're going to be a lot more upset than if you were just straightforward about putting together a little proposal and saying, hey, I, I think that we need to use AI or that we could use AI to do some of these things. Here's some of the protections that I recommend putting in place. Whereas like, you know, we talked a little bit about using a paid version that, you know, toggles off that it's using your data to train the GPT and that it's in more of a closed system.

That was a really big thing for us. A hurdle to clear to say that like, this isn't just going out publicly on the Internet, you know, it's in a closed system, we have control of it, it's password protected, there are safeguards in place. Also maybe defining some of the use cases on like when you would just use it when you would not use it can be really helpful to kind of get some of that buy in and then just like demoing what it can do. But I think it is both fortunate and unfortunate. But one of the things that really drives adoption is that like your competitors and your detractors are using it, right? Like, and this is the biggest challenge with s with social media. The people with the worst intentions are the best at using new technology. And so you know that there are people who are already far, far ahead of the game in terms of anything that we've talked about. In terms of like a guy in China faked a zoom call and got a CEO to transfer like $25 million to an account and he used like AI powered avatars of actual people and voice cloning to do it.

So there's a lot of stuff happening very quickly and like, if you're not embracing it or using it to do some things, you're just going to get left behind. So it's better to one, be upfront about it, to really outline your use cases and explain the value and then three, help them understand that like your competitors, your detractors, your allies are all using these kinds of tools already. And so you don't want to get left behind on this because that gap is going to continue to work. Enlightenment. And this is a technology that's moving faster than any technology has moved before.

I want to piggyback on that and just say in my blog and in my podcast, I say that the transcript was created by AI right when I write it down. And sometimes it's like the brief at the end says AI was a tool used in this. And I don't have any problems with that right now. If I go back and I rework a lot of it, I don't bother to put the AI stamp as much on it because it feels more like if an intern had written it and I rewrote it, if it was a bylined piece, it would just not have a byline, right. If it was totally rewritten. So to me, that's kind of how I use the guideline for myself of when to disclose you're using AI versus when not, and how you use AI versus how you don't. Right. Like if you see videos from me, they're real people that gave me real interviews you're not going to see me using.

And you can track back to them, Right. And I think people want to know, oh, these are the places where you do use AI so they don't assume I'm using them in other places.

I'll use my outdoor voice. Some basic questions. It's basic to me anyway. How can you trust or have you run into situations where you cannot trust the vocabulary of what's being given back to you? Who ask a question, how do you do special effects in videos? And what are some of the fee structures?

One of the. One of the things, I think, Carol, that really gets to that I had a friend recently ask AI about a press release. How do I write a press release for this event? And when she came back and she asked me, how does this press release look for this event? I went, you didn't need a press release. You need a media background, you need. Or an invitation. So like, if you didn't know that there was some. I mean, everybody, you know, everybody we've ever talked to only thinks about press releases, right? Unless you're in the business. Somebody had told her, just get a.

Press release on it.

Right?

The way it wrote it so differently when she put in good information was mind blowing for her, right? And it reminded her, this is why I have PR people I work with and maybe I should do it all on my phone. Right? But I think, I think there's a lot of those pieces. Vocabulary is a good one. You know, mastitis, it might pick up, it may not, Right. And it depends on, on some things. Like in mine, when I use Cast Magic, it lets me put in a context prompt, which is different than the prompt for the output. So it's giving it the context. So I will say this is an interview with a dairy farmer talking about Animal health concerns use information from trusted sources and veterinarians and extension services.

And the vocabulary is far better. But you will find some funky stuff in transcripts if you don't have context to it. And because it hallucinates, it's like the child who really wants to be loved so badly. And if you ask it to write you a press release, it's like, oh my God, I'm so excited to write you a press release. It doesn't know to go, oh, I should probably do a media backgrounder instead that relies on you to tell.

Tell it.

If you tell it you want it to do something and it really doesn't know how, it starts filling in blanks like a kid will do. Right. Like they broke the lamp and you. They don't want to say they broke the lamp. So they start and you can see their little minds working like well I, I was, I was in the room and the lamp just fell.

AI does exactly that.

Would you guys.

Yeah, and I'll piggyback on that. And Haley, you probably have even more to add to this but like for our client thoughts that we have, it took, we're on before of our financial industry compliance spot. So it took iterations and actively trying to break it to like drive down to exactly what we needed to give it. And one of the nicest things about if you're building one of these custom bots in chat GPT you can actually give it static resources to use. So we have five or six PDFs that are installed in it that it reviews before it gives us any outputs. And we say like here's you know the brand standard PDF, here's the words that you're not allowed to use PDF, here's the words that you can use and should use every single time PDF. And then there's like all sorts of we have a 40 page legal document installed on it and it reads it every single time in seconds. And we know that we have like hyper engineered this thing and it's just reps.

We just had to put in the reps and try and break it. So yeah, repetition is definitely key. Some of my favorite hallucinations from like a grammar standpoint, I mean I'm sure for all of us that use a thing like GPT often it's pretty much single handedly ruined the EM dash because it uses it so frequently that you pretty much, if you're reading something, you.

Know, I use EM dashes more than AI.

I do as well.

And they're really great for ADHD and I like Them too. But it is one of those cues sometimes that you're like, oh, AI probably wrote that.

And then the after comma is another one that I see.

Even if I explicitly tell it, do not add the oxford comma. It can do that still. So, yes, the repetitions and then shout out to Aaron's shirt.

I don't use it when I love.

And then last thing I'll just add is I also notice it's really word count happy. So sometimes I just have to be very specific about length of what I'm expecting in the output as well, because it'll. It'll ratiate now olathon if you want it.

Yeah. And I think what everybody said is like, correct. And to get to your point about fee structure, I think like Pro Chat GBT is like 20 bucks a month or something right now, so it's not that expensive. And to have that ability to create the custom dbts, that's where the gap is, Right? If you were just using, like, the free version of any of these tools, you're just going to get basic stuff back, right? There's a freemium version to get you addicted to it. But to really do anything from, like a professional standpoint, you need to be using a paid tool and you need to be giving it custom instructions, otherwise you will never get value out of it. And I think a lot of people who have had bad experiences with AI or think that it's stupid and like, you know, yes, the very basic version, once you get beyond the surface level of any topic, it's going to fail you. There's just no way around that. So the gap is widening between, like, what you can do with free chat and, for example, what the Air Force did where they trained a AI that flies a combat wingman.

So that's the gap between free chat and something that, like, the Air force is paying $500,000 a month to access.

So quick, quick kind of pulse check on the audience. Does everyone know what a custom GPT is when we're talking about this? Does anyone need clarification? There's no shame if you do, because I just started. Okay, so custom DPTs are a feature within Chat GPT where you can actually build a recurring function. It's like saving a Word doc that you want to be able to use over and over and over again. So when we say word, we're building a custom GPT, we're actually like, saving a subset of ChatGPT that we can use repetitively for the same function, and we can tune it and Train it to do exactly what we need it to do. So it sounds overwhelming. It's actually super easy. If we have a minute, we could probably build one right now if we wanted to show you how it works.

We're really close on time.

Yeah. I just want to piggyback on that though a little bit because I think sometimes if people are just getting into it, I think Casey maybe on one of your webinars talked about the security of having a low paying account versus you're feeding it every time you get the free one. So you talk a little bit about maybe the security or the client stuff when you pay versus free.

Yeah.

And thank you. Going on and off. I'll use my outdoor voice as well. Sure. Okay. So yeah, the security is really important and it's important to read like the user agreements and understand what you're signing up for. I think in PR when you're working with news before it's news, you maybe don't want to just have ChatGPT write the press release about a big announcement. Regardless of what it is, regardless of whether you think anybody's ever going to discover it.

Like once it's out there, it's out there. Once it's outside of your private network, it's out there. There are people who have demonstrated that they can do some very smart, well, dorking and service free chat, GPT requests based on usernames and that type of thing. So security is important. Yeah, yeah. Something you've shown you see is why I hold the.

We know how to get in the guts.

Yeah. And so I think that's a big piece of it is making sure that you understand what you're doing when you're using these tools and that you have that toggle switch turned off. That's like use my data to train the AI. And that's really one of the major differences between free and paid. Right. It's the same type of thing. If you're getting a product for free, you know, you're not the customer, you are the product. Same thing with social media.

Right. Like you're using it for free, they're collecting all your data and you're freely giving it to them. Same thing when you're using free open source AI tools.

I, I think it's one of those pieces on security. This one's not working either.

Now we're passing the world smallest microphone.

It's. It was one of those things where I wouldn't put information about people. I wouldn't put information, anything that I consider confidential at any Level, I think I love that other people are comfortable enough to put in like things where they're having problems with their family and they're looking for guidance on how to go about that. If you're, if you're doing that on a free model, people are collecting your data and they're learning what problems your family's experiencing or if you're having financial issues and you ask Cat about it or if you're having whatever it is. Right. Like you're having a problem with a colleague at work, it knows who you are. If it's a free version, that data is likely going to show up in some places. Right.

That's why I started paying for it quickly. Right. And I do have some friends who are willing to do some of those things knowing they paid for it.

And I still have friends that are.

Like, it's still too, it's still too questionable. I'm not positive. Right. Because all of these terms and agreements are page, page after page. And unless you take time to really read with like, I mean, who agreed on itunes or whatever it is, right. Like scroll to the end and you. So make sure you're, you're, you know what you're doing. The paid versions give you a lot more autonomy where there's kind of like a one way door.

Your information doesn't go out into the Internet, but the Internet does something and helps make sense of the stuff you put into it. Right. So it goes out to the Internet, pulls that stuff in and then your next generation with it. It's not sharing that with everybody on the Internet if you pay for the tools. Usually if you're not sure you want to pay for Chat GPT, go into your free version that you've been using and ask them who you are. Yeah, that is a good tool to ask Chat GPT D or something you've been using as a free. What did you know about P it it's amazing.

We kind of talk about this in context of accessing things through Chrome or are there any differences in how they act or perform and can do that through say duck Duck or browser or other.

Oh, buddy. Okay.

So he asked how, how is AI interacting with other search engines that are more French like DuckDuckGo? First off, thank you for reminding me of DuckDuckGo exists because like I said, Google has owned 85% of the search market share for well over a decade. It's very easy to forget about some of these other ones. It will say that pretty much every search engine now has its own sort of AI layered in. Now with Google's case, it's very visible because they have Gemini and they're doing all of this AI mode testing currently. They also have the AI overviews, DuckDuckGo and Baidu and some of your other search engines that aren't as mainstream. They aren't necessarily pushing it on us as actively as Google and to a lesser extent being. But it's still happening. And even though right now most of it is opt in or it's the AI overviews type experience, it's probably going to be impacting all search in some way within the next three to five years, I would say.

And honestly, like what was some of the earliest AI out there? It was the algorithms that search engines use. AI is just a series of mathematically based reasoning that computers are able to do and that's what the Google algorithms have been since they want. It's all math, it's all programming. So it's different in that AI is able to do reasoning and make decisions, but all of it comes back from, from this same level of tech that is driving search engines. So it's everywhere. It's everywhere. Not to be like a doomsday conspiracy theorists or anything, but it's, it's layered into most mainstream tech at this point in time. Wow, that was some big silence.

I, I was wondering, you mentioned Kelly Claude. The others did not. Can you talk about Claude versus these others?

Yeah, so Claude is basically, I don't want to say basically the same. It's just a competitor to chat GPT at this point.

They're all.

So the specific model of AI like machines is large language model. That is what we think of when we're thinking of AI chatbots like ChatGPT. Claude is just a competitor.

So I was on a conference call where there was an AI professor from Illinois talking to some of the students working in the AI agriculture space and he said that he prefers Claude. Of the morality behind how Claude was built.

Another one is PI. PI AI. It's touted as the emotionally intelligent AI bot and they're all tuned slightly differently. ChatGPT is just the largest one with the largest data set. So it tends to be considered the smartest and has the most features. Claude tends to be more based on morality and ethics. They're really attempting to do it right, right. Whether or not there's a right or wrong way to do AI.

Whereas like PI is trying to do it in a human, a very human way. I joke that if you want to talk about your feelings, go to PI. If you want to talk about statistics, go to chatgpt. Do I just have one comment?

Maybe ma spe.

I think they built some guard rails using USDA data. There's a stratovation GPT.

There is. It's called ag research and it's spelled kind of weird. We didn't ask AI to spell it for us. It's H E R, I research.

Agro research.

Yeah.

Okay.

And it basically is plugged into all.

Of the chat GPT you can find. R, I, S E, a R, C, H research.

And it plugs into all the USDA stats and everything.

So yeah. Yes, it is a really helpful.

So, so much.

All right, we're gonna have to go.

Oh, yes.

Darren.

Last time report out today. AI willing to let humans die. Oh, no. Or perform blackmail. Way to shut down early. So.

Yeah. Hey, skyet start somewhere, guys.

Well, everybody, I hope you'll join me in thanking Haley, Casey and Kelly offer being on this panel and working so hard to get these presentations approved by clients and all as needed to be able to share this kind of stuff as openly as we were able to. Because I think this is the kind of stuff Bart does that you can't.

Just find everywhere else.

Thank you.

The world's smallest mic.

Thank you all again. One more time.

We would love to get a group picture. I know it's always like full chaos, sports, family fun. But maybe just out in the lobby we can sleep together and get a group picture from this year.

And just a quick fact check too, on anybody's plans may have changed. Aaron organized the tour this afternoon at.

The and the lunch. So.

Who'S still going on the two? We did.

Oh, you did that already.

I'm sorry.

Okay.

All right, so we're good.

We're gonna organize all that is now live.

Wonderful. Okay.

Very good.

So if you guys want to pack up and cut out, let's get a.

Fixer right over here.

Okay. Sorry.

I feel like there's a quota.