How fast can you upskill in AI? We did a sprint to find out.

0:00Hi, everyone. It's Rana. The future of AI is not built in isolation. It is shaped in rooms where ambitious people challenge each other. Masters of Scale Summit is back October 20th through 22nd in San Francisco, bringing together founders and innovators pushing the edge of what is possible. If you're building what comes next, you should be in this room. Apply now at mastersofscale.com slash pioneers. That's mastersofscale.com slash pioneers. On Masters of Scale, iconic leaders reveal how they've beaten the odds. Asking really strong questions is a superpower. You want to show up with something radically different and how they've grown companies to incredible heights.

0:46The greatest rewards always come from the greatest risks. That's hit the gas. Airbnb, Zillow, Microsoft, Liquid Death and more. Hear from the founders who've changed the game. It's anything but business as usual. Find Masters of Scale on Apple Podcasts, Spotify, YouTube, or wherever else you get podcasts.

1:15Every company has to ask themselves how they'll be disrupted. Every company will eventually be disrupted. That's Taryn Fixel. She's COO and president of Wait What, the production company behind this podcast, as well as masters of scale and rapid response. So we had to ask ourselves, how is AI set to disrupt our work? And what can we do to get ahead of that so that ultimately we could be successful in this new environment? Countless companies, large and small, are at a similar crossroads. We could keep doing work the same way we've been, or we can really harness AI to reimagine how we do work. We know that AI can supercharge an enterprise, like automate tedious work or streamline operations, boost output.

2:05But how do you actually do that? How do you start? Today, come along with this one small production company, ours, as we plunge into AI to transform work. and your guide will be Pioneers of AI Senior Producer Rachel Ishikawa. Hi, Rachel. Hey, Rana. So this is a very meta episode. It is. So let me tell you the story. Our company took a very specific approach to leveling up our AI game. For three days, we paused all operations and every single one of us tested how to use AI for our work. And for me, somebody who works on an AI podcast, this was actually a lot of fun. So what did that look like? Well, we experimented with out-of-the-box products, wrote code, some of us for the first time like me, developed apps, scrapped apps, and then started all over.

2:59We called it an AI sprint. That sounds pretty awesome. Did it work? I mean, is the company now running at full AI speed? Well, that's what we're here to find out. And to show our audience what we've learned along the way, I'm going to bring you from start to finish lines and all the hurdles in between. I'm Rachel Ishikawa. And I'm Rana El-Kalubi. And this is Pioneers of AI.

3:33Okay, Rachel, so let's start with a little bit of background on Wait What. What does the company do and how big is the team? Give us a rundown. Definitely. We're a media company. We've been around for about nine years. We make three podcasts, this one and two others. This is Masters of Scale. I'm Jeff Berman, your host. I'm Bob Safian, and this is Rapid Response. And we publish about 200 episodes a year on video and audio, plus our website, social media, newsletters. And there's, of course, the Masters of Scale Summit, which I love. Yeah, we hold this big three-day event called Masters of Scale Summit in October.

4:13It's in San Francisco. It's dozens of in-person speakers and performers. There's all this ticketing and logistics that go into it. And we also pop up with other one-off events throughout the year. So I have to say, for a company that's less than 40 people, we definitely punch above our weight. Yeah, that's a lot. So how has the company been using AI up till now? I'll speak for myself. I've been using AI a lot for research, which makes sense since I produce a podcast about AI. But as a company, our AI use has been haphazard. And for a company that talks a lot about AI with leaders in the field on our shows, it feels like we should have better systems in place because we acutely understand the stakes.

4:58There is no industry that this will not impact. Again, Taryn, COO and president. I want to, number one, make sure that the organization is nimble and has the ability to thrive in the future media landscape that we live in. And number two, I actually really do feel that we have a responsibility to every member of our team to give them access to these tools and help them feel empowered to use them. Taryn knew that we needed to retool around AI, but the how was still up in the air. So she phoned a friend, a new friend. My name is Parth Patil, and I'm an AI engineer, and I work with the office of Reid Hoffman.

5:41And I spend most of my days working with generative AI and advising startups, entrepreneurs. In short form, I'd say I spend 14 hours a day talking to language models. He talks to LLMs all day, not just chat tools, but coding platforms like CloudCode, to stay up to date on how to use them, how they're changing, and how to teach others about these tools. If you don't become AI native, anyone on your team with high potential is not likely to stick around long term. If they do stick around, it might be because they're like, oh, I automated my job and no one knows. I mean, for a lot of people, they may go build your competitor and like beat you because they're able to go deeper and further and faster because of these tools.

6:27And what's at stake is that you're going to be competing with other new entities that are AI native that move in ways that you didn't think was possible. And so Taryn made the call. She just wanted to learn from me how I use some of my favorite tools. And she basically looked at me after two hours in and she was just like, do you think that you could teach a whole team how to think like this in maybe like two days? Two days, let's make it three. And I suggested that we pause company operations for three days in order to all align around how we're going to use AI. What can we use it for? And I honestly, I was very skeptical.

7:04I mean, I know for me, it's like if I sit next to like a friend for four hours, I can definitely like AI pill them. And then like they'll be changed moving forward. But I was like, how do you do this for a team? This is the Wild West. I think that the most important thing that we can do for our staff in terms of upskilling everybody is giving them the courage to start something new. Like you two can be an expert on AI if you just get started today. Okay, so the motivation was there, but how do you actually pull off a three-day AI sprint? Were people even on board? Well, Taryn and the rest of our leadership, along with Parth, put together this roadmap for us.

7:46They cleared everybody's schedule for three days, signed us all up for clawed accounts, enough seats to cover the entire company. And then they split the company into small groups that were tackling specific questions. Like, how can we use AI to help us make video faster? Or how can we use AI to surface guest ideas? And how can we make planning our three-day summit more efficient? And at the end of the three days, each team would present their results to the whole company. Could AI solve the problem at hand? How? And don't just answer. Build an answer. Ready, set, go. Okay, and so the sprint begins.

8:31Out of the starting blocks, how is it looking, Rachel? Well, we're a fully remote company, so the sprint started the way all of our meetings do, which is on Zoom. Folks, welcome to our AI sprint. We made it. Today is a day about making things, about building things, about seeing things coming up that we have the opportunity to use and iterate, use and iterate. After the kickoff, each group spent most of the day in their breakout groups. We were working together, getting to know our new AI co-worker, Claude, but we were also just catching up with our human co-workers who we don't get to see too much, you know, talking about the important things in life.

9:11Although I hate Gertie Reign. You know what I mean? I hate it with a passion. If you water down, it's better. I find vitamin water a lot. Starting off, not everyone was feeling so confident about the exercise or about using AI for certain parts of our work, including MG, who's a video editor here. I think the first day I was like, oh, no, does this mean that I'm supposed to use AI to video edit entirely? That's understandable. It's so important to draw boundaries around AI and be intentional about what you want to delegate to AI versus what you want to do yourself. I agree. Like there's so many parts of my job that feel really tedious, but there's parts that I don't want to outsource.

9:54For example, making editorial decisions. Okay, so the groups get started. What platforms are they using? Well, most of us were using Claude, but Claude, as you know, is a single-player tool right now, so there's no collaboration layer. So one team member had to screen share over Zoom as they navigated Claude on their own computer. Let's dig into that for a second, because that's a very important point. That is one of the biggest limitations of tools like Claude today. It's changing very fast, But it's basically not like Google Docs where multiple people can be in the document collaborating. It's kind of more like Microsoft Word, really.

10:33Yeah, that's what Parth was saying, too, is that he expected this to change in the coming months. And I know this is something that you've talked about, Rana, how quickly this is all iterating. Totally. The AI that we have today is the worst it will ever be. And things are moving so fast. But, you know, there were some other technical hiccups, too. At one point, Cloud Desktop stopped working altogether. Cloud Desktop failing to open for some users. Oh, interesting. Every time we hit a snag, Parth was available to answer our questions. And remember, most of us were AI novices, so there were a lot of snags.

11:12I was wondering, since we're starting with Quad code, is there a way that we should sort of prep in Quad first before we start in Replicit and give it a good base of code to begin with? So you can import projects into Replit to work on something that you already have. GitHub is probably the best way to import, but you can also import a project as like a folder of files. And with those snags, there were some important lessons that we learned along the way. There's so much to discuss, but I narrowed it down to three lessons because, you know, everybody loves threes. Let's dig into it. First lesson.

11:47Engage in conversation with AI. Treat it like a colleague. The more back and forth you have with your AI, the more specific you get, the better the results are going to be. So Team Gatorade, a.k.a. Jodine, Littal, and Stephanie, they were working on a guest speaker engine, a way to help us find cool guests for our podcasts and for live events. Rachel, that's such a great example. As you know, I've been thinking a lot about, like, what kind of work do we want to delegate to AI versus do ourselves? In this particular case, we have our weekly meetings where we discuss various guest pitches and whatnot.

12:26That's not going to go away, right? Yeah, I think that's right. We'd still have our weekly meetings. We'd still be the ones figuring out who's going to land on our shows. But we'd have this other tool that could come up with new ideas that maybe we wouldn't think about. We first started out by asking Claude how it would find guests for podcasts and live events and then how it would organize the database. That's Stephanie Stern, senior talent executive. She leads booking guests on all of our shows. But from there, we actually backtracked, asking Claude to ask us clarifying questions before creating this comprehensive database.

13:03They asked using a voice-to-text tool so it's easier to have a natural conversation with Claude. Before we start building, can you ask us some questions so you can get a better idea of the mission of our company and how we typically select guests and speakers for our podcasts and events. So ask lots of questions, ask the AI to interview you, and when you reach a roadblock, you can ask the AI itself for help. This is something Parth recommended again and again throughout the sprint. This is like the first time we have a computer that can use language and that can speak. The idea that you can wield a computer through natural language means that you kind of have a steam engine for knowledge work.

13:53It reminds me of like Harry Potter and like spellcasting in Harry Potter. It's like if you know the right combination of words, like things start happening. Wingardium Leviosa. When I was first learning computer science, I programmed in C++. But what Parth is saying is you basically don't need to learn C++ or any other programming language for that matter to build your own software. You can basically use just plain English or Arabic or Chinese to prompt the AI to build software on your behalf. Exactly, which is why it's so important to treat your AI like you would a coworker. So that's a good first lesson.

14:30What's the second lesson, Rachel? Lesson two, look for areas of your work where there's too much clicking around. All those tedious tasks like manually entering data into an endless field of spreadsheets. Instead, see if you can use AI. To help bring that lesson to life, let me introduce you to my coworker, D 'Angela. Hi, I am D 'Angela Napier, and I am the Special Events Project Manager for the Masters of Scale Summit. Again, Summit is the big three-day live event we produce. Before we started the sprint, I really just thought about, of all the things that I do, what's something that I thought could be better?

15:10Her work involves lots of details. There's a couple of little pain points where you're like, but I wish this was better. So I was thinking about the hotel management because that's a big part of what I do as it gets closer to Summit. it. This is important and tedious, keeping track of everyone's travel information, their hotel and preferences, like what floor they want. And this information isn't static. Travel plans change a lot. Yeah, I'm totally guilty of that. Look, a lot of people are. And D 'Angelo is the one tracking it all on a bunch of different spreadsheets, which means a lot of clicking around, taking information from email or Slack, even text messages, and then inputting it again and again.

15:58Yeah, this is the kind of work that AI is really good at, pulling and organizing data that is often scattered all over the place. Yeah, so she started building a real-time dashboard, and D 'Angela isn't the only one of my co-workers, hoping AI can reduce the clicks and the cut and paste of it all. There's so much backend detail work around registration, ticket codes, tracking responses. Taryn had a really good description here. There is so much invisible work in everyday roles. It is very easy from the outside to look at somebody's role and go, look, what does this person actually do all day? I think this is where AI introduces a really meaningful shift in creating workflows that are observable, explainable, and shareable.

16:44The people in our department know the things that I do, but I always get the feeling that most people don't. I come from a military family, and my dad was like, it doesn't matter what accolades you get, just do the job well. And so I'm going to do it well anyway, but it's nice to have that recognition. I asked D 'Angela what advice she had for others getting started with AI. I would just say, don't be scared. Just try. Just try. Then start thinking about how you can apply that in your career. Because a lot of people think that AI is here and I'm going to be replaced. And that doesn't have to be the case.

17:20The thing is, I'm not sure if that's the case. In a minute, the elephant in the room. Are we learning how to use these models or are we training them to replace us? That's after a short break.

17:39Now more than ever, technology is a dominating force in our lives. Then there's the threat of AI everywhere. And yet, tech can be inspiring and help level playing fields. I mean, a YouTuber with a self-funded debut movie just dominated the box office. I thought, hey, if you interview me, it'd be good for your publication. That's not ego. I just have a lot of followers. But it's that stigma. It's like YouTubers, they're not real. Join me, Lizzie O 'Leary, the host of What Next TBD, Slate's podcast focused on technology, power, and the future. Follow What Next TBD now, wherever you get your podcasts.

18:26All right, we are back. Let's call it the second lap of three in this AI sprint. How's it going? Well, the 12 teams are making headway in their projects. Some have even built prototypes. We are working on a web-based app that can help us review applications for Summit much quicker. It has a queue for our applications. It allows us to score them. Some are testing off-the-shelf products. Some of the Descript tools that we hooked around and discovered are actually pretty useful. And others are still in the exploration phase. I feel like we might be a little behind, but this particular process is really wrapped up in a lot of other things that are going on, both in the Sprint and at the company.

19:12Got it. What is this all costing, though? Leadership said that the highest cost at this point is paying for seats to use tools like Cloud and Replit. There's also the cost of tokens, though most of the team members haven't maxed out on that yet. You also have to consider the cost of people spending time learning, which means they aren't doing the other parts of their jobs. Absolutely, Rachel. There's a lot of hidden costs here that we often don't talk about. So, so far, you've shared two of the three lessons you want to highlight. Lesson one, engage with your AI and treat it like a colleague. And lesson two is make AI do the repetitive, tedious work.

19:52What's lesson three? We're going to get to that in a minute. But before we do, I want to address something for a moment. This big question. Is the Sprint about upskilling us or is it about replacing us? I mean, we've all seen the headlines at this point. Companies have been cutting thousands of jobs and some are saying it's because of AI. Yeah, but some of that might be AI washing, right? AI will disrupt some jobs, but it is my fundamental belief that AI will also add a lot of new jobs to the economy. Yeah, I do agree with that. But at the same time, it is kind of weird to see Claude do aspects of my job and do it, I would say, almost as well or just as well.

20:33It's a little bit of an ego hit. And just to put a finer point on this, we used to have an associate producer for this podcast, and currently we don't. AI is not the reason why we don't have an AP right now. But I will say that I have been using AI a lot to do the job that an AP would do. Wait, what? And its leadership is very transparent about this tension. It is uncomfortable as a leader to ask your team to train on technology that they fear will replace them. Here's Taryn again. And also, I feel a sense of responsibility to make sure that everybody on our team understands how to use these tools so that they're well positioned for a long career.

21:19Taryn's upfront that she doesn't know what any of our jobs, even hers, will look like in the future. But she made it a point to address that uncertainty head on. I felt that by creating a space where we were doing this collectively, it countered some of those fears and would make it more productive. Right. So having the sprint be a co-created project rather than something that is imposed on the team top down is giving people agency to think about how AI is impacting their work. Yeah. And that's how it felt for us, too, as we were doing the sprint. It felt like that even to MG, who is a video editor here and one of the team members who started out pretty against AI.

22:00I did not use AI before the sprint. I think there's a few different reasons. Number one, as an artist, I'm also a writer and a performer, and I feel bad for artists whose work has been scraped from the Internet, copied, pushed through this like sausage maker AI thing. And then I think also, you know, environmental concerns are a big one for me, just sort of worrying about all of the processing power and how that is affecting our planet. Yeah, these are valid concerns we've talked a lot about on this show. So how do you bring somebody like an MG on board? Well, MG, you know, is the kind of person who approaches skepticism with deep curiosity.

22:46They were part of a team that was working on ways to find how they can use AI to get to rough video cuts faster, which is no small order. There's like importing all of the files, organizing everything in Premiere, and then creating the multicam sequence, syncing the premixed audio. And then you get to like, OK, great. Now let me pick the shot, you know, and do sort of like the rough pass. A video editor's role is creative and technical. It's kind of similar to our audio engineers and design team, too. These are all areas where AI is advancing fast. They can make cuts to video, do graphic layouts, manipulate audio files to improve the sound.

23:29But like, it became clear that that was not the purpose of our AI sprint. Because these tools can't perform at the level of experienced, talented humans, the end product quality is just not the same. So we're not looking to fully automate these areas, at least not yet. It was not to like take away the parts of our jobs that we love the most or that are like creative, that are human, that are fulfilling, that are artistic, but rather to like get us to those aspects more quickly. Which brings me to the third and final lesson I want to share with you, how to make decisions. As you bring AI into your work, you'll need to think critically about what to delegate to it, what to keep in your own human hands, and then you'll have to decide whether to build or buy.

24:24Yeah, you want to build a framework where human judgment is irreplaceable. So, for example, for this podcast, we still get to decide who gets to be on the show and what questions to ask. And then, of course, there's the age-old question when it comes to new technology, build or buy? Again, for instance, say you want to use AI to generate social clips for the podcast. Do you build this yourself using agentic AI or do you buy an off-the-shelf product? And what's the cost comparison in terms of time, money, and resources? Yeah, and the team's instinct in this scenario was to test off-the-shelf products for video.

25:02So I was looking at different products, so many of which were like brand new and having new versions like every single day. And did any of them rate as, oh my God, we have to have this? Not really. A lot showed promise, but there was always like one issue. So maybe the program didn't offer, you know, an audio transcription, which we really need when you're working with podcasting. Or maybe the plugin used the wrong kind of file. So much of it is brand new, and I would be super curious to see where these companies and their products are now. I feel like they're probably radically developed, but it was cool to sort of try out and see different things that worked.

25:46It's a tall order, with or without AI. Yeah, so maybe Wait What can consider building their own AI agents to solve this issue in the future? Yeah, I mean, I would think so. I mean, like MG said, it's a very tall order. it may make sense to wait and see what the off-the-shelf products look like in the future. So we'll find out. So what happens as we wrap up day two of this print? Was there any clear winner? Well, Rana, we'll get to that in a minute after a short break.

26:33At the end of day three of the sprint, each of the 12 groups presented their findings. Team Gatorade, aka Team Guest Speaker Engine, made some real headway on their database. So, what do we build? We build a standalone webpage. Ooh, this is sexy, right? It is. It has some really good information on here. Picture a dashboard with different tags and guest suggestions. There's a little context about each guest, too. Okay, but how good were these suggestions? Well, there are some suggestions that were almost too on the nose, like who doesn't love Oprah? Yeah, Oprah, you're welcome on the show anytime.

27:14Yes, Oprah, please come on the show. But, you know, in reality, we're also looking at other people who might be lesser-known names, and the app was pretty good at coming up with some of those people. What about the other projects? MG's group presented, and no surprise, they didn't find any perfect AI tool that they were confident in. Okay, so maybe Team MG didn't win the race, but who did? Look, it wasn't a competition, and there were so many cool projects that came out of the sprint. One of my co-workers, Taylor, built this really cool tool to monitor all of the incoming application and ticket sales for Masters of Scale Summit.