Share
What You’ll Learn
Are you tight on time and resources, but have a big event to plan? NTCA’s Mark Marion outlines how an easy-to-build AI bot helped him plan and execute a recent big event.
Guest Speaker
Mark Marion
Show Notes
Transcripts are lightly edited for clarity and readability.
Intro: StoryConnect, a Pioneer podcast helps you discover ideas to shape your stories and connect with your community. Love this episode? Follow StoryConnect on your favorite podcast platform and YouTube so you don't miss your next great idea.
Andy Johns: How can AI help you on your next event? That's what we'll be talking about on this episode of StoryConnect: The Podcast. My name is Andy Johns, your host, and I'm joined on this episode by the great Mark Marion, who is Senior Director of Member Services and Engagement with NTCA, the Rural Broadband Association. Mark, thanks so much for joining me.
Mark Marion: You're welcome. Thanks so much for putting the great in there. I appreciate that.
Andy Johns: Well, we have talked about doing a podcast for a long time and a discussion that came up back in, I guess it was at RTIME in February where you were talking about everything that you were doing, how much AI had helped out. NTCA obviously puts on great events, and you are right at the center of all of that. So as we got to talking about, you know, the ways that you had used AI. It was like this is our topic. This is what we need to talk about. So I know folks will be interested to hear it. Just in case anybody out there has ever been tasked with coming up with an event and a short timeline and with somewhat scant resources, but with high expectations, that's kind of where you found yourself when it comes to NTCA's AI Summit that y'all did in January. Set the stage for us a little bit on that event and a little bit about how it came about.
Mark Marion: Yeah, absolutely. You know, we've been talking about AI on NTCA programs since, you know, early 2023.
Andy Johns: Sure.
Mark Marion: And it's made an appearance across all of our agendas.
Andy Johns: Right.
Mark Marion: There was really that desire to pull that all together. Can we talk about AI as a topic and really how it cuts across a number of particularly operational aspects within our NTCA member companies. You know, the decision was really solidified in October of '25, and it was we want to get this going and out the door January '26.
Andy Johns: Wow. Okay.
Mark Marion: Before RTIME '26. Now my team, as they are working on RTIME, that's an all hands effort. And my initial reaction was –
Andy Johns: That next month, just that next month in February, right? I mean, that's the 2,500 person conference. It's huge. Yeah.
Mark Marion: It's a big one. Yeah. And so my initial thought was, I don't know that we can do it. You know, I don't know that we have the resources to pull it off. And then I thought, hang on a second. We've got this tool in our back pocket. We've been using AI for a number of things. Let me spin up some assistance here with AI and, you know, a little bit meta to use an AI tool to help plan the AI Summit, but it ended up being a tremendous resource, to allow – I, you know, I don't want to say that it was me single handedly, but a lot of that preparation, a lot of the development of the content and the supporting materials, that was pretty much me and AI. And then I brought in some other folks at the end for really the delivery of that program, really kind of day of and that week prior.
Andy Johns: Impressive because I did hear really good things about the Summit. I know that everybody, and one of the things that we struggle with a little bit whenever you're talking about AI, the listeners out there, some of them may be just kind of scratching the surface and, you know, okay, I just used a little bit. Others may be very advanced. What I like about what we're going to be talking about here is I feel like there's a little bit for everybody. So if you're listening and you feel like, you know, parts of that are, oh, I know how to do that already. There may be other parts when we get into the AI bot that Mark built that are a little bit more advanced and vice versa. So I think it's, I think it's a little bit for everybody. We're going to kind of talk it through before, during and after the event. Because I know while maybe not pulling off a big summit like you did between end of October and mid January, there are other folks out there in the broadband and electric space on the communicator side of things that are certainly tasked with pulling off some big events in a hurry. So let's talk a little bit about beforehand. So what we're, you know, you have this revelation that, okay, it's an AI conference, an AI Summit. I might as well use AI to kind of help me make it happen since, you know, a little short handed. What were some of the first things you did and the stuff before the event that AI helped you out with?
Mark Marion: Yeah. Well, I'll go way back, you know, in AI relative terms I'll go way back to to November of 2022. You know, that's when the public beta of ChatGPT was released. And I actually went back because I was like, what was the actual day I set up my ChatGPT account?
Andy Johns: Okay.
Mark Marion: And it was December 17 of 2022.
Andy Johns: Oh, you were early.
Mark Marion: So in terms of early adoption, I really jumped into it. And that initial use –
Andy Johns: Sam Altman and then you. Yeah, that's pretty quick.
Mark Marion: I kept running into him in there. I was like, "Get out of my chats!" But you know, that early exploration was really just crazy fun. What is this? What can this do? And so I remember some of those early searches, and I actually went back to look at what some of those were. But I remember, you know, just thinking, I want to throw things at this just to see how does it work? And so I said things like compare baking to banking by way of a metaphor. And it just went. It just, all of a sudden it started writing and it started talking about, "Well, if you think about banking as a, you know, a kind of a financial network, and you think about baking as having all of these ingredients, you could consider those ingredients to be components of a financial." And I was like, yeah, well, hang on a second.
Andy Johns: Yeah.
Mark Marion: Well, you know, separate chat, create, you know, a four panel Garfield comic that explains communism. You know, like just kind of throwing crazy things at it.
Andy Johns: Whatever. Yeah.
Mark Marion: And the outputs were amazing.
Andy Johns: Mhm. Imperfect, but amazing.
Mark Marion: Oh, imperfect for sure. But early 2023, that's where I was like, okay, wait a minute there's some capability here that I can bring into work.
Andy Johns: Okay.
Mark Marion: There are, you know, 15 page documents that I can give to this tool to say, summarize it.
Andy Johns: Mhm.
Mark Marion: And as the models got better, and as I got better at what I was asking it, I could start to say summarize this document for me in my context in the field that I'm working in and the topic that I'm exploring, you know, kind of not just the summary, but now recast that to make it something that is going to be relevant.
Andy Johns: Right.
Mark Marion: Carry that through, you know, into, 2025. And as the capability came online, I switched over to Claude pretty much as soon as Claude was available as well. But kind of used both. Yeah.
Andy Johns: And I will jump in right there that, you know, everybody uses different ones. Mark's mentioned a couple already. I use ChatGPT primarily. Pioneer and or NTCA, I don't imagine, are specifically endorsing any particular company or brand or anything when it comes to AI. Just talking about our lived experiences all there, so yeah, go ahead.
Mark Marion: Absolutely. No. And I'll throw out Gemini because I've certainly used Google's Gemini as well. But yeah, no, no, no endorsement at all there. But when the capability came into these models to create customized AI tools in ChatGPT, it's called a custom GPT. In Claude, it's called a Claude project. In Gemini, it's called a Gem. It gave you this functionality to say, here's an instruction set to define how you reply to me in this chat. You're not just chatting with the bare model. Now you're saying, I want you to think in this way, reply in this way. Here are my inputs and push back on them in this way.
Andy Johns: Game changer.
Mark Marion: Yeah. In addition, you could add in knowledge. You can say, sure, you've got a universe of information that you've been trained on. And again, later on in the process with these models, they got web access where they could go out and do some searching and come back. But I want to give you information at your fingertips that I consider to be primary source. Go to these firsts. So that's going way back. But to answer your question. October 2025, I'm thinking I need to spin up a custom AI Summit planning assistant. So I did this in Claude. I built out a project, and there is that empty field where you can just write in instructions for that assistant. I had learned through my experience, don't do that. It's just like prompting. Like, sure, if you want to get into a chat, you could just go in plain and start chatting. But if I want to write instructions for a CLaude project, I'm going to have Claude do that. I'm going to explain to Claude, this is what I need. Now you write.
Andy Johns: You write, okay.
Mark Marion: Instructions. You know, and it's called meta prompting. You know, it's going to do a better job. It's going to put it in that structure and format that is going to work best for that purpose. Then it becomes –
Andy Johns: With your review.
Mark Marion: Yeah.
Andy Johns: Yeah. With your review between each copy paste with you stepping in and reading it and tweaking it, I imagine a little bit.
Mark Marion: Absolutely. Yeah. And that's important of course. You know, looking at that output and making sure, well, is that right? And do I understand the instruction set that it's been given? But to give an example at the very beginning of these instructions.
I just have them open here. It says "Role: you are an expert virtual event planner specializing in attendee experience for Zoom events. Your client is NTCA–The Rural Broadband Association. They've brought you on to assist with marketing copywriting, script development, event optimization, technical guidance, and perspectives on the overall attendee experience." So instead of chatting with that Claude based model, all of a sudden there are little weights that are placed on that model, that fine tune it to say, keep in mind you're going to be replying in this way, and then it goes on. What should your response style be? Here are assumptions you should not make. Here's background information that you have in your knowledge base to draw on. And in that knowledge base, you know, this is part of the human influence. I had put together an internal brief about a three page brief to get buy in internally. Does this seem right? Had a basic wireframe sketch of what the content, you know, structure would look like. We were at the same time running our 30 day AI challenge with the Innovation and Business Opportunities Committee with NTCA members. And so I definitely wanted to integrate that competition, you know, that, that, that inspiration, that, that other thing into, what we were doing with the AI Summit. So that three page document that I wrote that became knowledge. Our engage book that has all of the NTCA member benefits in it and kind of explains who NTCA is, that's in the knowledge. The FRS glossary, that's in the knowledge. I had a couple articles that I had saved through our venture into virtual programing during the pandemic that I really liked about virtual conference attendee experience that went into. Like, this is all stuff that this bot should have at its fingertips so that when I'm firing a question at it, even if it's, you know, I'm not saying go find the answer in this documentation and deliver it back to me. It's this is all going to influence the quality of the answer that I'm getting back. That it's not going to be a generic answer; it's going to be my answer for this specific purpose.
Andy Johns: Yeah, I can definitely see somebody getting ready for their annual meeting with the same setup you're talking about. You know, you bot are our expert event planner planning an event for the name of the cooperative and then feeding in things like the style guide. And, you know, some of the other kind of foundational, you know, about us pages, that sort of thing. And, and really being a powerful tool for somebody as they're setting stuff like this up to keep it consistent and make it happen.
Mark Marion: Absolutely. And if you particularly liked the agenda that you had last year for your annual meeting, load that in as knowledge. You know, if you've read an article or if you attended a session, and you've got maybe notes from that session or a transcript from a recording, boom. All of that can just sit in it, and it can hold a pretty good amount of knowledge. But yeah, that was very much an early step because I knew that, again, resource wise, I knew where my limits were in terms of time, mostly you asked about time. And I had a sense of where this was going to buy me that time, where this is going to take something that would take me, you know, maybe three, 4 or 5 hours to do something that it could return quite literally in seconds. I mean, sometimes, you know. 30 seconds.
Andy Johns: Yeah, there's still that crucial human piece of it right now to know when AI can save you the time, and when it's not. Okay. So you're using that. Once you've got the bot built, then what you start using that to for agenda ideas? Or because this was so new and so quick, I don't know that y'all had a ton of time for speaker submissions. Or what were the steps there in terms of building out the agenda and getting it moving?
Mark Marion: Yeah. Well, we did have that AI challenge running. And what I had already decided was we were going to take the entries into that AI Summit or into that 30 day AI challenge. Turn those into case studies for the Summit.
Andy Johns: Smart.
Mark Marion: What I didn't know was exactly how I would do that. So one of the very first chats that I had with it, I gave it the submission form, and I gave it the instructions for the challenge.
Andy Johns: Mhm.
Mark Marion: And I said, I want to take the submissions from this and turn these into five minute case studies. Give me the structure for what those case studies should be. Mhm. Again, seconds. I mean, you know, probably about 10 seconds.
Andy Johns: Sure.
Mark Marion: It comes back with that outline to say great, you would pull basically what they're saying here and put that here. Have them talk about this, and you're done. Great. Capture that as an artifact, you know, capture that template. Load that into the project knowledge for later.
Andy Johns: Mhm.
Mark Marion: So that when we did have submissions, and when I did reach out to someone to say, hey, would you be willing to hop on a Zoom call? We're going to record for five minutes. I'm going to make it very easy for you. And I'm going to basically give you a bulleted script of your own writing.
Andy Johns: Right.
Mark Marion: And I took their completed form, you know, their completed submission form loaded that in, made reference to the template and said, turn this into bullets using that template. It does that. I turn around and give that to these folks. It's their words. It's their ideas just reformatted and truncated down. We hop on for a Zoom call. They were like, this is great. Like, I can just talk. It's a narrative I can just talk through. We captured it, and it was done. So on that content development side, that saved us all a tremendous amount of time. On looking at the overall program, yeah, there were lots of chats back and forth to say, okay, I have these pieces. I'm thinking about putting them here in this order. Does that make sense? And again, it's being instructed to serve as that expert and as that pushback. And sometimes it'll say, yeah, sounds great. And other times it would say, what if we reserve that for this other track? And here we really build that out with, you know, a vendor session where maybe they could play up this angle. Great. Great idea. Any vendors in mind who might be good? You know, I've got some thoughts, and it started spitting them out. I was like, okay, it knows. Yeah, this is good. These are a lot of the names that I would think to include here. So yeah, I mean, that process was definitely very good. We did some panel discussions on the front end. Shirley Bloomfield, past CEO, moderated a session, Mike Romano, current CEO, moderated one, and Josh Seidemann, our VP of policy. What I was able to do there, we had the panelists established. By way of having the panelists, we sort of had a broad idea of what this panel discussion was going to cover. By giving that AI tool, those parameters, these are the specific people and this is that topic, we've got 20 minutes. Give me the question set. I looked through it. I made some tweaks. I was like, that's not a very good question. This would be a more applicable question if we drill down a little bit. But, you know, we did a little bit of back and forth, a little bit of iteration. I sent those questions out to the various panels, and all three of them unanimously wrote back and said, these are awesome. I have no other questions to add to this. I was like, we must be dialed in. I mean, this is a good sign. And they were. They were very good discussions. I will say that target was 20 minute panels. You get some people together who are experts and who like to talk, and you end up with a 30 minute recording for a 20 minute slot.
Andy Johns: Sure.
Mark Marion: That's not going to work.
Andy Johns: Right.
Mark Marion: So I'm left thinking, great. I'm going to have to edit ten minutes out of this video. Then I thought, Ken, Claude, help!
Andy Johns: Wait a minute. I don't have to do, yeah.
Mark Marion: Exactly. So I exported a transcript with timecode from that recording.
Andy Johns: Mhm.
Mark Marion: Loaded it in, and I said I need to find ten minutes to cut from this. What do you think? Now again, I'm loading it into my bot so it has all of the contents.
Andy Johns: It's got the context. That's important.
Mark Marion: Yeah. Although I have to imagine even if I were chatting with the base model, it would do a fairly good job. But this was really focusing in on what I wanted it to focus in on. Truly seconds. I mean, 15-20 seconds it started writing back with very specific time code references to say, cut this portion out. It's an extra story. That's kind of nice. It illustrates the point, but it's not critical. Cut this portion out because this panelist already said this prior. And it gave me probably, I would say 8.5 minutes of content. And so I just had to find that additional two minutes of really kind of tightening further. But I just followed the instructions, went through, made the cuts as it suggested, and played it back. I was like, that would have taken me the better part of a day to I mean, if anyone's edited video, I mean.
Andy Johns: It takes.
Mark Marion: It's a lot.
Andy Johns: Yeah. Speaking of, we are going to have to move along. When we were talking about this episode, Mark said he was betting on 24 minutes. We're already at 22. So we got some more to do. We haven't even started the event yet. Since you mentioned Shirley and Josh, I will go ahead and, do like we do on this, Josh Seidemann episode 218 of StoryConnect. The Shirley Bloomfield episode 102. So if you want to go back and check those out, we always like to include that in there. All right, Mark, so the event starts, and one of the things I really wanted to get into too, hopefully this isn't blowing your cover. But you know, when you've got a team on an event, it's so important. And we do this with the StoryConnect event too. Just trying to get the whole team on board every morning to know like, what's going on today? What are the high point? Make sure to remember A, B, and C. You had the bots kind of help you build that out and some documentation, some emails for folks as well to keep the team on track.
Mark Marion: Yeah. For this meeting in particular, for this AI Summit, what that really looked like were, particularly for the co-hosts that were doing the live portion of each of the tracks. All of that was was tightly scripted. At these points, we're going to answer these questions. We're going to tee up, you know, the prerecorded segments in these ways. So that was very much a sync point for them. I think along the lines of what you're talking about is kind of even a little bit of an extension afterward. But as we kind of think about how to use these programed assistants, you know, these AI assistants in whatever platform. Yeah, we've built out some tools for RTIME and for fall conference to basically draft staff dailies, to say, here's the format, here's all of the information in advance of what you're going to need. But then in the instruction set, we have it ask us some questions. So it could just draft a daily for this is Tuesday at RTIME. This is what the staff need to know about Tuesday. But the instructions also say ask if there's anything that's come up that needs to be highlighted. Something new that's not in this information. You know, and then at the end, there's always a little, fun fact about the area that we're in. So when we were in Disney, it had all kinds of fun.
Andy Johns: So many fun facts about (inaudible).
Mark Marion: Bits of information. But yeah, I think that extension of the idea of where else could we use an assistant? And the metaphor that I like to use is, imagine you had an internship program with Harvard, and you had some PhD students on loan. You've got to give them very clear instructions about what they're doing, but they're highly capable at following instructions. The more information you give them and the more context that you give them, the better the result is going to be.
Andy Johns: Right.
Mark Marion: You're still going to have to review everything. That's AI. If you think about these programed AI assistants, that's like working with the same intern over and over again, rather than getting a different person.
Andy Johns: A brand new. That's a good way to think about it. That's a really good point. I'm going to jump through the rest of the event and get into the after. Unless there was anything that you wanted to bring up during the event that it was used for, but I know there are quite a few things in terms of after the event summarizing and you know, some of the conclusions there. Anything else during the event before we move on to the after?
Mark Marion: No, I mean, it definitely assisted with some of the marketing copy. you know, coming up with a tagline for the event. That was more creative collaboration that was back and forth. That wasn't just give me this great copy paste, and it's done. But it was helpful again, to have that thinking partner and that little bit of thinking space offload.
Andy Johns: Got it. All right. So the event is over. You are, you know, successful event people got a lot out of it. We sent somebody from our staff to be on the Summit in particular. What are some of the things that you did after the event where AI was able to help out quite a bit? Because you said there were some summaries that it was making connections that that were not really readily apparent.
Mark Marion: Yeah. I think the summaries were really, that was one of the key elements and kind of a little bit of a surprise element for me. But I knew we wanted to publish, you know, like kind of a two-page takeaway of here's what we learned with this AI Summit. You know, here are some great quotes. And here are some great talking points across all of the breakouts that, you know, we had four simultaneous or three simultaneous tracks going. I guess we had four. But again, I had the transcripts from all of that prerecorded content. So I just took that whole universe of transcripts and loaded them all in. And I said, let's get this down to some takeaways. The surprising part was it was making connections across sessions that no attendee would have seen both of. They could. They could go back on the platform after.
Andy Johns: Breakouts?
Mark Marion: Yeah.
Andy Johns: Okay.
Mark Marion: They could go back to watch the recordings later and watch, you know, some track content that they weren't in. But it had all of that information. So it was saying, hey, this person made this point on the marketing track, and this person made this point on the technology track. And guess what? They're the same point from two different angles. Now, technically, I could have, some may say should have, made that connection as well because I was in all of those session recordings, but there's no way I was retaining all of that and thinking about the connection. But when you charge a tool like this to do what it is very good at, pattern recognition and making those kind of lateral connections. I was really happy with the way that takeaway document came out. Again, it was iterative. It was, you know, it took some tweaking, it took some checking, you know, to make sure.
Andy Johns: Of course.
Mark Marion: Is that an actual thing that someone said in a session? Or is it making that up? Hallucinations have gotten better over time. But no, that was definitely a huge takeaway for me in terms of value on the back end of the conference.
Andy Johns: Sure. A couple of, I had two questions, and you brought up hallucinations, so I have one more to add in there. So when you're going back and checking, as you're doing the project, you know, I imagine that makes it better that you're not just sending it out on the internet to go find information. But how often or did you find anything that you'd consider one of the AI hallucinations or just something really off base, off track. If not incorrect, then at least something that wasn't meaningful. Or was it pretty much, you know, nothing too far afield.
Mark Marion: Yeah. Well, from my experience with just prompting overall, but especially in working on that instruction set, it's right in the instructions. Don't hallucinate. If you don't know the answer to something, don't make it up.
Andy Johns: Okay.
Mark Marion: Just state that you don't know what that is, and that's okay. And by naming it and giving the tool permission to simply say, I don't know, it's base programing pretty much says don't say, I don't know. And that's why a lot of times in those base conversations, it can seem overly confident. It's also designed to keep the conversation going, which is why it can seem overly sycophantic. And that's a great idea. My instructions say don't be a sycophant.
Andy Johns: Oh, okay. That's good, That's good. And that goes back to both of those are good advice for the intern metaphor you were doing earlier too.
Mark Marion: Absolutely. Don't kiss up. It's not going to get you anywhere. Yeah.
Andy Johns: Right. Last two things I had for you. So I imagine, you know, with that event and with any other events that you're doing, these kind of projects for before, imagine every project that you do, the next year's event will be able to pull from that. Are you, are you thinking in the future, each conference kind of gets its own project? Or are you thinking you've just got a big events project that pulls in all the NTCA events? Kind of, how are you thinking about that going forward? Because I know y'all have quite a few different events every year with very different audiences at times.
Mark Marion: Yeah, we've got a number of these bots. And some of them now that we share. These are custom GPTs that we use on a shared business GPT account. So we can kind of all get into them across the team. But they're specialized in what they're kind of designed to do, not by conference, but kind of by function. Where, you know, supplemental information about a conference can be added in to that particular chat, but that set of instructions that's really kind of focused on the broader information about what it's designed to do. You know, I think we're definitely engaging with AI more. We're also engaging deeply with AI governance and making sure that, as we're approaching this, it's not total Wild West. That we have a handle on what data is going in, how is the data that's coming out being used? And yeah, you know, while there's not a direct successor for this particular event in mind, we've definitely shown we can do a virtual summit, pretty successfully. We drew over 700 people for that conference in January that we can put this together. It was a little bit of a proof of concept and also a demonstration that there's an audience of people who, yeah, they're willing to hop on for a half, you know, half a day and, talk a little bit about a shared topic. And, so yeah, definitely more, more to come on next steps in that area.
Andy Johns: Sure. I mean, I can see this being useful. I know you guys book your facilities way in advance, but I could see this being handy, you know, evaluate these events, evaluate these contracts from the different hotels, proposals, that kind of stuff. There's a lot where you could get into. But what I guess the last question that I'll have here, what advice would you give somebody who's maybe staring down that annual meeting or Customer Service Appreciation Day or, you know, where people are already getting into the fall and winter planning for their Christmas events or those kind of things. What advice would you have for somebody at a telco who's sitting there thinking, this sounds great. Where do I start? What advice would you have for them?
Mark Marion: Well, you know, I'll talk to that part of the audience who maybe you've tried a chat with AI or, you know, have thought about trying a chat with AI. You just have to jump in. You just have to jump in. And, you know, I did a presentation to a director audience at RTIME about how directors should be thinking about using AI in the boardroom and in preparation for doing their job as board members. And I gave the same advice to them, just try it. And if you don't know what to ask it, ask it what you should ask it. You know that as the jumping off point. Tell it a little bit about yourself, about your situation. If you're worried about data leakage or if you're worried about, you know, anything on that confidentiality side, you know, only put in what you're comfortable putting in there. You know, I think those concerns are a little bit overblown in the media, but give it a little bit of an understanding of who you are, the problem that you're facing. And basically say, I want to use generative AI. I want to use this chat to help me, but I don't know where to start. That's the start. I mean, that will really kick off the conversation.
Andy Johns: I guess along those lines, I would say the caveat, what we always say at Pioneer is, you know, if you're in the event planning, I would stay away from at this point, uploading potential attendee lists or, you know, any kind of personal identification, information. And then of course, obviously the human is responsible for whatever goes out there. You got to have that human step. But some of this is great use of it, and a really smart use of your time. So thank you for taking some of that time you saved on that and spending it with us here on the podcast so.
Mark Marion: No, glad to share. And, you know, certainly if folks have any other, you know, follow up questions, glad to, you know, talk AI. I love it. Again I've been using it since 2022. I'm an old man when it comes to my beard and AI usage.
Andy Johns: There you go. You were into it very early on that for sure. So Mark, thanks again.
Mark Marion: All right. Thank you.
Andy Johns: He is Mark Marion with NTCA–The Rural Broadband Association. I'm Andy Johns with Pioneer. And until we talk again, keep telling your story.
Outro: StoryConnect is a production of Pioneer Utility Resources. Send questions to hello@pioneer.com and learn more about our member-owned marketing agency at Pioneer.coop.