AI Hackathon 2025: From Ideation to Innovation with AI Agents and Workflows

Full video transcript

What kind of tech app would you build for the universe? You’ve got 30 seconds to brainstorm, go. This weekend we’re building AI agents and agentic workflows that are leveraging AI and automation. A dozen of us are going to be getting together, experienced builders and entrepreneurs who have already developed products and had success in the world of AI and their own businesses. We’re coming together for this private in-person AI hackathon, sponsored by Gen AI University, my company, to see what we can build and to see what we can share together and learn together. And ultimately what can come out of this event, I have no idea. I know what I’m going to be working on and building, kicking off on Saturday 9:00 a.m. Central Time. We wrap up around 6:00 p.m., checking in and sharing what was the final outcome of this hackathon, what have we built, and what did we learn through the experience.

Getting ready to head on out to the hackathon house, a full day ahead of us. Got about 2 hours till we’re going to be kicking off, see what we can create together. Got our coffee, super pumped. Beautiful day here in Austin, Texas, nice crisp morning and beautiful weather. There’s going to be a big backyard where we’re at, a lot of space to move, walk around. But the creative juice, one of the things I’m looking forward to building here at this hackathon is a voice agent that I can use as an assistant for our customer support and our clients inside of our Market Sauce portal. Last night I was using ElevenLabs to play around with it and it sounds pretty good, so I gave it another shot. I wanted to create another voice assistant to see what else ElevenLabs could do. This was last night, and I used the Side Hustle game, play Sidehustle.com, and I used the game instructions to create a voice for Rich Rock, the character who would be an AI entrepreneur game-night host. The idea with that is to have a voice character embedded on the game so teachers, people in classrooms or running game nights, will be able to work with a virtual AI host that could help keep track of score, points, the timer, the names of the teams. It was getting pretty cool, that’s just what it could do.

For this application, with customer support and AI agents, I think it’s going to change how we can deliver the resources and the services and support and answering questions. It’s a really clear and already proven-out use case for how AI can bring a lot of value right away. And there’s going to be some technical setup and integration that I’m going to need to learn today, because I don’t want to have a bunch of things that are still super manual. I’m going to try and make a lot of that go as smooth as possible. So let’s see what we can get into right here at the Chateau, get set up, and people start arriving inside of the hour, get ready to kick things off.

Now we can get started. Get a little tour of what we got working with here. This is going to be the deep work table, welcome to it, this is where you want to put your monitor up there. This is where people can obviously collaborate, but it’s more for that deep work. In the grand room, this is to talk ideas, problems we’re solving, put post-it notes up on the board, use Kanban if we want to track our progress. So here’s for collaboration, I can imagine people sharing screens, talking through ideas.

Come from Phoenix, so this is cold. It is cool. What we got going on here: an AI influencer travel app, basically we’re creating a personalized experience for your itineraries based on what travel influencers post, and then personalizing it to your preferences. First step of the day is to map out what are the small rocks that can end up leading to a great first MVP experience, that minimal viable product I can use AI to bridge the gap between what your goals are for your family trip, your surprise engagement, whatever that experience is. We’re going to use AI to personalize that itinerary for you.

I’ve got two projects: one is for a buddy of mine that does research for a podcast, his intake form, his initial intake values. He takes that, does some research on doctors, health influencers, looks up what their medical background is, what their educational background is, what their product is, what their audience is, so he can say “this is somebody I want to interview or somebody I don’t want to interview.” It’s taking him two or three hours to do all that research, and I’m going to compress that to a 30-second output process for him. He fills out a form about somebody he’s looking at, and he knows, boom, I want to interview this person or I don’t. And on a personal note, I do a lot of email data monetization stuff, taking your website, pulling info about your company, figuring out what those inputs are, output email design based on two or three design ideas, then a 45-day email plan, a narrative strategy of how we’re going to manage your account, welcome sequence, abandoned cart, etc. Stuff that would take me several days to put together is done like that.

I’m a data analyst effectively, I’ve been in the game for about 7 years, I run a company that consults on marketing data, and I started building tools to help me do that. The first tool I built was launch analytics, I also built a tool called Organized Automation, and I spoke at Measure Summit about that. Essentially I want to create a world where we can see into each data collection system, compare it based on support documentation, and then have the decision maker make a decision based on how good that system’s data quality is, not based on opinion. It’s been a lot of fun, and today I’m focusing on project management, being able to juggle all these projects as a person who’s coming into this development world without a development background. I use tools like Claude, MCP, LM Studio, Deep Seek, some local tools, Ollama, also GitHub quite a bit, Cursor as well as Lovable and Bolt.

Hi, I’m from Phoenix, Arizona, have a long history in business. One of the things I’m working on now is building out a community, an AI-based community for coaches, to help them get their businesses up and running. They have great passion for what they’re doing but they don’t know a lot about business and technology, that’s a huge issue for them. We try to make it easy, take all my business background and put it into AI-based frameworks and implement those, and I’ve written tons and tons of prompts, but what I’m really trying to do now is make those all work together in a seamless package, using multiple agents to get really good outputs and then automate the process so you don’t spend all your time creating and posting and planning and following up and engaging, that endless cycle that burns you out. You want to be able to actually do your business. I’m just starting a new tool that’s been recommended, seems like a pretty good agentic tool, and they’ve got a new version coming out next week, so right now it’s kind of linear but it’s going to have the option to do a lot more branching and keep the human in the loop. Very big on keeping the human in the loop, you’ve got to see what it’s giving you and not just let it run off the rails because you’ll never get the quality you want. So figuring out how to keep the human in the loop, how to set the right benchmarks, and then probably using some Make, and ultimately the big challenge has been getting this content into HighLevel so all our clients that run on HighLevel can get this stuff automated, because even if you create a nice output, you still got to copy and paste, and we need to get away from the copy-and-paste world.

Really I’m working on creating something to solve my ADD brain, I’ve got all these cool little projects I think are fun and exciting, and then I realized I had all this other stuff that actually needed to be done, and I’m looking to create an automation for doing tasks I’m curious about and want to learn about or follow up on. What I’m working on today is going to be crypto, I’ve got a bunch of people that follow a bunch of narratives I’m looking into and researching, and figuring out how I can have something that will shorten that whole process for me, give me the information I want, and then apply that to our next project, which is an insurance company we’re building right now, figuring out how to make it simple for people to buy insurance online but have that process be fully customized and target the older demographic that people think don’t love the internet, but I’ll use YouTube every day.

Hey guys, I was on the early team at Jasper and helped them go from zero to a $1.5 billion valuation in 12 months, getting us to our first 100,000 users. I’ve been using OpenAI since 2020, when it was just GPT-2 and it would pretty much autocomplete your sentences. I’ve been on a 12-month sabbatical, took off in December of last year, got married, traveled for a bit, moved out here, and I started to realize over the past year that the way we book travel is broken. As a travel planner myself, I like to think about my 30 days in Bali and want to know what are the best restaurants, what hotel should I stay at based on my points, and I believe the way I’ve curated my really romantic trips, or curate all my friends’ bachelor parties, I have this standard of the best experiences.

I’ve gone traveling, went on honeymoon, went to Canada, went to Panama, and I found that planning my travel, that process was really broken. I end up researching one place, having Google Drives with docs, Google Sheets, my notes, saving captions that people have on Instagram with recommendations, so I just wanted to create a better travel app that would be personalized using AI based on my preferences. They say to drive across the country you need a car, but to go down the street you just need a skateboard. Today I built an app, but just the skateboard version, not the car to go all the way across the country. I just needed an MVP, a minimal viable product. What we did today in literally 6 to 8 hours is I built an entire coding environment with the help of a friend, where we have Next.js on the frontend hosted on Vercel, set up my GitHub account with version control, using the backend with AWS Lambda, the database is AWS DynamoDB. We’re not doing anything around user logins today, this is really just me as the sole user, and if it solves my problem then maybe we could scale it up to solve your problem too. So this was all just to get the environment set up for me to start using a bunch of AI tools, the main one being o1 Pro, we set up an o1 Pro workspace and that is going to be my head of engineering. I have the vision for where the product’s going, and I need to create an environment in which I can speak to my head engineer to actually write the JavaScript.

Earlier all of these were over here, each time I created a task I put it over in the done column, so now we have a live website, this app is passportsdo-ai.com, and it has pulled in, we can go to it now, so here’s the homepage, all everything is written and created and coded by AI, I just wrote a prompt that created this page for me: plan a memorable trip with your AI travel assistant. Here are some trips, my Bali honeymoon was imported, had the o1 Pro operator go and gather all of my itinerary of where I stayed, my flight information, different restaurants I ate at, and all of that is here in one place. Then I had it look at my Banff trip and create the four-day itinerary based on my notes from my Apple Notes. The Iceland trip, this is something I created here in Google Sheets, so I uploaded the CSV and it automatically created a new trip in my app. And then a friend just wrote about this mastermind he’s hosting, and I gave it just the simple prompt of his invite, and it created the whole itinerary for his mastermind trip, ski-in ski-out, five-star hotel experience in Tokyo in the cherry blossom viewing season, nine days, and all of that was generated by AI off of his vision that he was pitching me on his mastermind.

I wanted to knock out a bunch of use cases where it’s my own app, my own trips, it’s my friends’ trips, all of the details that take a long time to gather from all these places, and it’s in just one app. I see this in the future becoming amazing, where somebody can write their personal preferences and travel, and then the AI assistant will take an existing trip but adapt it to the family, to the business, to the work trip, to the offsite. There’s a lot of directions this could go, but we solved the MVP today, and I’m so pumped because I’m not a technical guy, but I have a grand vision for the way my app could be built, and ChatGPT Pro is the lead engineer that made it all happen.

Was there anything you found helpful from being in an environment like this today, about 6 to 8 hours of head-down work, was there anything you found most useful to get something like this done in such a short amount of time? Well, look, you are the sum of the people that you’re surrounded with, and when you’re in an environment like this it’s a pressure ecosystem that creates diamonds. Everybody around, brainstorming what it could do, how it could be built, basically I didn’t know what the tech stack was for a modern app, but somebody in the room did, and they told me go and sign up for all of these apps, and when you’re done let me know, tap me on the shoulder, and then I can connect them all in about five minutes, and he did. I didn’t know how to get all this up, but it’s something that was like a one-time environment setup that would have taken me months to figure out, reading blog posts and watching YouTube videos, but there was somebody in the room I got to borrow their expertise from and just get it done right away, so I could actually get the value I’m creating for the world, which is my vision for a travel app. So I used the next six hours and actually made an MVP that’s workable, clickable, and look, here it is, this is a real live thing, and I couldn’t have done it without community, the people in the room is where the magic happens, so incredibly grateful for the people and for what we’re building here together.

Back home, taking in what was covered throughout the course of the day. People had some big wins developing SaaS products, their own apps using AI that are designed to get a job done. Some people didn’t make as much progress as they had hoped on their initial goal, me being one of those people, I still made a lot of progress where I knew the direction was going, but I just ended up spending more time thinking through and mapping out some of the problems, so gaining clarity in that regard, but from a technical standpoint not quite as much done that I would show and put on a website or as its own product. But I can see where the process is going much more clearly now, in terms of how we’re supporting our customers and onboarding people into our own products and services, and then just getting the kind of fly-on-the-wall insights into what people were sharing and the breakout sessions folks were having, brainstorming, collaborating, something that is fun and we’ll definitely be doing again, hosting these hackathons. Whether you’re a builder or you’re a founder with a big idea, if you’d like to be involved with another one of these events in some way, shape, or form, reach out to me, let me know, and I’ll be happy to get in touch and chat and see what we can build together. All right, y’all, keep calm on, I’m going to hang out with my pup now, I’ll see you guys on the other side, cheers y’all.