Loading summary
Transcript135 lines
- [00:00]
A
Hey, it's me, Dave.
- [00:01]
B
Our friends over at Customer I.O. are sponsors of today's episode. They're a really cool company that helps marketers turn first party data into engaging customer experiences across email, SMS and push. And they built their platform for marketers who actually care about the craft. Because marketing is a craft. It takes creativity, thought and taste. Right now, everyone thinks they're magically a marketer because they have access to AI and the result is kind of painful. More robotic emails, more noise, more bullets. Bl AI isn't magic. It's not going to fix bad strategy or write great copy for you magically. But the best teams also aren't ignoring it. They treat AI as infrastructure. When it's built the right way, it actually makes marketing feel more human, not less. And that's what Customer IO is doing. Their AI handles repetitive work like setup, orchestration and tasks that should be automated so that you can focus on what actually matters. The craft of marketing, the strategy, the creativity. This is how good marketers are using AI right now. Not to replace thinking, but to support it. If this landed with you at all, this idea about the craft of marketing, I want you to go and check out customer IO. It's customer IO. Exit 5. Go and check them out. Customer IO, exit 5.
- [01:10]
A
Hey, it's me, Dave.
- [01:11]
B
Today's episode is brought to you by Consensys, the interactive demo platform. Look, most of your buyers have already decided whether they like you before your sales team ever gets on a call with them. They've asked ChatGPT and Claude about your product. They found reviews about you online. They've talked to peers who've used your product before. And by the time they hit request a demo on your site, they've often already come to a decision. So you're losing control of the narrative before the first touch point. And now they have to wait three to five days for a demo. Consensus gives you that control back. They help you meet your buyer where they actually are with interactive, personalized AI demos that live on your site. When a buyer shows up wanting to poke around on their own terms, you give them what they want. Plus, you get to see exactly who's watching, what they clicked on and and who the decision makers are. So stop being a bystander. While LLMs sell or even unsell your product. Category leaders like Atlassian and Autodesk use consensus to turn invisible researchers into high intent leads. Go and check them out@goconcensus.com exit5 go consensus.com exit5 you're listening to the Dave Gerhardt show exit.
- [02:26]
A
One, two, two.
- [02:27]
C
One, two, three, four.
- [02:29]
A
Exit.
- [02:32]
B
Okay.
- [02:33]
A
Hey, everybody.
- [02:33]
D
We're live.
- [02:34]
A
It's the top of the hour here. My name is Dave Geart. I am the founder of Exit 5. Exit 5 is the top community for B2B marketing professionals. And twice a month or so, we do these Exit 5 live sessions where we bring on a group of subject matter experts to go deep on a topic that is really relevant and something that matters to you. We try to show real examples. That's what people want. They want the real examples. They don't want the fluff. And we have a really good one today on AI, but this time it's all about creative. This morning I've been working on my April Fool's Day campaign using AI creative tools myself this morning, doing very important work here as the founder of Exit 5. But this topic is. Is really fun. It's presented by 11 labs. We wanted to do a session about how marketers are using AI in their creative. But we actually wanted to get real creative people. There's a creative director here. There's a brand person, There's a cmo, There's a growth person. We want to hear what people are actually doing and how they're using AI to power creative and, you know, really going beyond slop. And we have examples of voice, video, music, sound effects, strategy. There's some really cool stuff. And I know you're going to get some specific takeaways from this. But before I bring up the panel here today, and we got some cool stuff to show you, I just would love to hear in the chat your name, where are you writing in from? And then also want to know, like, why did you take the time out of your day to come hang out with us on this session? So, like, I'm Dave. I'm in Vermont. I came here because I'm super interested in AI and creative, and I really want to learn more that I can take back to my company. So drop that in the chat because it's really fun to just see where people are writing in from. So do me a favor and pop that in the chat while I get situated here and we give everybody a couple minutes to roll in. We get Jake's in Cleveland, always trying to hone his skills and add new skills. There we go. Yeah. It's so lonely when you're up here. Is anyone here? And then all of a sudden, we must be having a little delay in the chat today. Braden in Canada, Marissa and St. Pete. Content isn't just SEO. I like that Steve works at a small startup, needs to automate as much as possible. Akash is in India. Just a big fan. We love that. Cool. All right, this is great. We're going to have a really good session today. So as I mentioned today, we have a sponsor for Today's session. It's 11 labs. Really cool stuff that they do. You've probably heard of them before. They're the Voice AI company. We've actually used 11 labs in our own newsletter so readers can have it read to them out loud. I need to do more here with training my voice because I'm still recording our ads with my actual voice. We should be using 11 labs to do this. They've built a full creative platform that includes voice, video, image, music, sound effects, localization, and more, all in one workspace with one license. Every marketing team right now is being asked to do more with less. I know a lot of you in the chat saying that right now. And the teams figuring out how to automate the creative production layer are pulling ahead fast. Teams like Ramp Clay and Nvidia Good logos. 11 Labs teams right there. They're using 11 Labs to output 11 times more creative without adding headcount. 11 Labs lets you spend less time on production and more time on creative and strategy. Go and check them out. 11 Labs IO Creative. 11 Labs IO Creative linked into the chat. All right, producer Allison, who's sitting in today, our. Our events lady is hot on webinars right now. Fresh of our marketing leadership, he stepped in to help us run this webinar. Why don't you send our guests out here and we'll do quick intros and get things rolling.
- [05:46]
E
Cool.
- [05:46]
A
All right, let's go around the horn right here. Luke, let's kick it to you first. So just to set the stage for everybody listening, I want to get introductions from everybody so you know who these people are. Some proof, some credibility to hear what's interesting to them about how they're using AI and creative. And then we're going to go one by one. Everybody has one example, and we prepped for this yesterday. I've seen these examples. Some cool stuff coming. We're going to talk through and show you a couple of examples here. And then we have plenty of time to take all your questions. And I have a discussion that I want to lead about AI and creative and where things are going. So let's go Luke, Liz, V and Carter, and we'll kick it off.
- [06:18]
F
Okay.
- [06:18]
C
Hey, everyone. I'm Luke. Really great to be here. I lead the growth and marketing at ElevenLabs, but I also lead the products for 11 creatives. So I'm building lots of the tools for our team and other leading marketing teams. Really excited to be here, Liz. I'll hand over to you.
- [06:34]
E
Hi, everyone. I'm Liz. I'm the CMO at Brain Labs. Prior to Brain Labs, I spent a really long time in B2B and demand gen. So I have approached some of our AI creative challenges from a systematic layer. So I'm excited to reveal some of that stuff and show you guys how you can automate a bunch of stuff in your creative process.
- [06:52]
F
Nice, V. Hi, I'm V or Vicente. I'm a creative director at Bitly. We're the friendly neighborhood shortlink company. I'm excited to show some of the video work that we've been doing at Bitly and how we're able to shrink that production time up to two weeks.
- [07:09]
A
So I was happy to have a creative director on. On for this session because initially I was like, all the creative people are going to revolt. A bunch of creative people I know in my life are like, AI is going to, you know, take all of our jobs. AI Creative sucks. And so I'm very happy to have some creative folks on here. Carter, you're also in a. You know, you're in the brand world at UiPath, but I'll let you give your intro. Hey, everybody.
- [07:31]
D
Just want to give a big shout out to Dave and the Exit 5 team and 11 labs for having me. I feel very honored. And also just the other panelists here. In a very short time I've known them, I've already learned a lot from them. So I think everyone here is in for a real treat today. Carter, as Taylor was saying, I had the video initiative on the global brand team here at uifa.
- [07:51]
A
Awesome. So we're going to go around the horn quickly. I got a couple minutes up for each person here. Just to set expectations at home. We're going to talk through some examples. Some people are going to show some examples. If you have questions as we go, the best place to do it. If you have, like, feedback on what's going on, like specific examples, put them in the chat. But if you have questions that you want us to get to later, put them in the Q and A because then we can sort them by upvotes and we do that. So, Luke, you got the mic first here.
- [08:15]
C
Yeah. I'd love to show one of the videos we made for brand campaigns.
- [08:20]
G
My name is Michael Caine. You know, I've played many roles in my Life both on and off the screen. But now I've got a question for you. What makes a voice iconic? Is it the tone? The accent, the balance? Well, the answer is quite simply the message. And now, through the power of artificial intelligence, we've turned sound into something more connection without limitation. Because voices give us strength, power, protest, prayer, and joy. But each new frontier, each new advancement in technology is met with caution, and rightly so. But every single time, the human drive of passion, artistry and creativity prevails. Because technology is just the medium. This is humanity amplified. 5. I am thrilled to announce that my voice and I are joining 11 labs. Try it for yourself.
- [09:41]
C
It's always a hard act to follow with Michael Kane.
- [09:45]
D
That's sweet.
- [09:46]
C
You have that incredible voice when you hear my British London voice. But why I wanted to show you that video today. And even though it's one we made last year, we're very proud of when Michael Kane joined our iconic marketplaces. What's so cool about that video was the entire thing was made by AI by one guy called Jack, who did it in one day, going from idea to the full finished version. And so that's using Michael Caine's voice replica, and it's using the nano banana models. It's then using VEO to then animate it. And so I think it shows this really cool shift of what's now possible with this different tech. And so just to jump into a few of the like so specifics here
- [10:27]
A
while you do that, by the way, the one thing that like, so the reactions to that are amazing. I love that use case because I feel like, for me, if it's like fake characters, if it feels like fake, you remember the rip now, but they shut it down. Like Sora. The Sora videos I could show to my kids, who are 8 years old and my daughter be like that. That's obvious. Like, you have 12 fingers. Like, that's not you. But because of that, because of the footage that was chosen there, like, that feels very real. And it doesn't feel like, ooh, an AI video. Like, that was a perfect use case. I think that's very relatable for a lot of.
- [10:59]
C
Yeah, 100%. I think we're in this really cool phase at the moment where I think that AI voice models are there. The sound effects in the video, the music, that's all fantastic images. And B roll is, like, really good. We've reached that threshold. But at least for us at Elevenlabs, we're not using anything for, like, a roll. Talking head person in front of the camera. We just don't think the quality's there up for our bar. And so to kind of jump into a few more of the tactical bits of how we actually need it. And I saw the really good question by Caroline of like, what's the legality of actually using a person's voice? And so what we built is this voice marketplace where we have thousands of voices which are real people who put their voice onto the site and then get paid with it. So we partnered with people like Michael Caine and the estate of Albert Einstein, such that you can actually request those, use those in your app. I'm actually calling from Poland, hence why it's showing the Polish voices. But lots of the ones on TikTok are US people using them. And then to give you a very quick glimpse, the like image and video is also the ones where. So how Jack actually built that video, he would actually go in, type away, do the prompt for those B rolls you normally start on the image, and then you pull it into the video and then also able to kind of go in onto the music side as well.
- [12:17]
A
Hey, one thing that popped up really quickly in the chat, kind of like around the legality of this. What's the answer to that? How could I use. I can use all this stuff in my marketing?
- [12:27]
C
Yeah, a hundred percent. So the way it works with the voice library is we work with lots of different voice actors and they go up, they do a voice capture, so they prove it's actually them by saying a random phrase, and then they get paid out as you use the voice. And so we paid out now well over $11 million in payouts to these voice actors. And then particularly for these, like, top tier of famous voices, that's where you actually can't use it immediately in the platform. You request for the specific use case, and then that way the actors are able to still control their likeness and how it's being used and with brands which they really want to represent. The other cool thing is you can also clone your own voices. So the stripe example of the company we work with is where the CEOs actually cloned his voice to then do an ad read in a different language as well.
- [13:18]
D
Cool.
- [13:19]
A
What else do you have more?
- [13:20]
D
Are we good?
- [13:21]
C
Yeah, that's the main stuff that I just wanted to show live and give an overview of. But I think we're in this really cool phase. And then how specifically Elevenlabs is using it. We actually started with hiring the first hire I ever made. So I joined nearly three years ago. Now, the first hire I ever made was we were kind of working with all these different AI creative tools was for an AI creative producer who would basically use these tools to make these different launch videos. And so as you're thinking about how to get into it, test it, I think that's a really good way. Is like either working with a contractor or working with an agency to actually do this, do this. Testing.
- [14:00]
A
Nice. I like it. Look, the narrative that I've been feeling and pushing is like, when it comes to marketing, in my career in marketing, I basically got promoted up to the CMO job because I was good at marketing. And eventually, if you're good at marketing, you get more people, more team responsibility, more budget. But eventually my job became not actually doing marketing at all. It's all people management budgets, you know, team planning and all that stuff. And so the reason I'm bullish on AI, beyond a lot of the hype is, like, for stuff like this, wait a second. I could have an idea on my own, write the script now. I don't. I don't. I don't have to rely on somebody else for Voice for Creative. Like, it feels. Feels like I can be this kind of, like, super marketer because of these things. And so that's what's really cool. All right, nice job, Luke. We got Liz. All right, Liz, you got something completely still in the realm, but different than what Luke showed.
- [14:50]
E
Yeah, very different. Also, the voices within 11 labs are awesome. We use them at Brain Labs. So shout out to the 11 Labs team and all the great stuff that they're doing. But yes, I have something really different. And I think just to tee us up before I show the very detailed slides that I have, we approached the AI creative from a multitude of different angles because you think about the creative AI landscape right now, and there's dozens of tools out there, right? There's my CEO sending me tools. My team is sending me tools. We're like, let's do all these AI tools. But I think the real unlock happens is when you build a system around it, when you figure out a way to actually systematically use this as a team and then be able to bring in other people that are not necessarily, necessarily designers to be able to engage with it as well. That's really when you create creative freedom for your designers and also for the marketing team to be able to do more strategic stuff and not so much the in the weeds, you know, hey, I need this image done set in different ways. Let's just spin this up type of thing. So without further ado, I'm going to share my screen here and show you all a very cool system that we built. So we built the system where we've essentially, we've built this brand so skill in Claude. A couple of things here is you do have to have Claude Enterprise to be able to use this as an org wide skill. But what we did was we inputted our brand guide, our visual styles and our tone into this brand. We call it a mega skill because it's like a huge brain of all of our brand stuff. We've inputted all of this and we've structured it very specifically to basically be fed into exact instructions that Claude could use for different scenarios of different things that we felt like were really repeatable things in our design process. So think things like someone on your team wants to post a LinkedIn post and they need an infographic or they need some sort of visual to go with their LinkedIn post. We got that request over and over and over again. So we were like, how do we create a process beyond templates? Because that's the other thing. I think historically most marketers have operated from a design standpoint of like creating a bunch of templates and then swapping content in and out.
- [16:47]
A
I have my existence. And then. Well, then the other thing that I always hated about this is then also let me know in the chat if you feel like you could relate this. But I also hated because I always felt like I was becoming like the no department. Right. And nobody. We always had to say, no, no, no, no, no, no. And like, then you're not the team that people want to work with. And so I think what's cool about what you're going to show is like, you've now enabled the team to make their own versions of things within the guardrails that you have created.
- [17:14]
E
Exactly. That's such a huge unlock. I mean, there's so many benefits to it, of course, from a performance standpoint, because you're not flooding LinkedIn with all very, like, generic templates ultimately, but you are giving the team the freedom to create their own things, but within the brand confines. So we created this mega skill. We imported all of our brand assets into that. I will say pro tip. You can use the skill creator within Claude skills to create the skill. I know that's a bit meta, it's like a skill for a skill, but it will tell you very clearly all of the instructions that you need to follow to essentially accomplish your end goal. So if you're thinking to yourself, okay, I have Repeatable asks that come in on ad units, I on infographics, on sales decks, on sales enablement material. You can outline all of those outputs and then ask the quad skill creator, how do I create the perfect skill to align to the output of all of these things? So it's super customizable to the exact needs. But I'd say a great starting point is thinking about what are the things that I get asked most frequently and how do I figure out a systematic way to continuously input that context into a skill and then be able to allow others to self serve on it. So that's what we've done with our actual Claude skill. I think I mentioned this, but the org skill is an enterprise functionality, so it's something that your team would need to have Claude enterprise to be able to access. But you can set this up on here.
- [18:31]
A
Is there a difference between like having a team plan and having an enterprise edition of Claude? Because I, I was working on this this morning. Since our call yesterday, we just have a team plan and if I go in, if I'm the admin, I just go in there and the setting there's a. I can create a universal, I can add a skill and then anyone can just recall that skill.
- [18:49]
E
Yeah. So I think on a team space you are able to have. This is like a thing that we have to just. I'll. I'll drop it in the chat once I get confirmed on my team on this. Exactly. But I believe you have to have either a team space or an enterprise space to have an org skill, because an org skill means that multiple people can access it. I think on an individual plan is where you fall into a little bit of a gray area because you're not able to then share that skill.
- [19:10]
A
Skill.
- [19:10]
E
So I'll confirm that in the chat. But yes, the idea is that you create a skill that essentially everybody can use and socialize that across your team and you use the skill creator to do so. Something that we've also done because within Claude skills it's really awesome that everybody gets to use it, but you also don't necessarily want everybody providing feedback into that cloud skill to actually update the instructions. You want to be able to vet the feedback that the team gives to you based on using that skill. So, so it's great to crowdsource it, but it's not necessarily. You don't want people to just auto add like you know, always use yellow as the background or always use blue because then that will update the clogged skills universally.
- [19:48]
A
So this might be Related, we might get here anyway. But Jose, how do you make sure that the process and context and data stay up to date? For example, our messaging evolves or we have a new customer story defeated. I think this is why you have in the bottom right here, she has this skill change request so people have to submit. That way you do have control over it and you update it regularly.
- [20:06]
E
Exactly, yeah. So what we've done is we use Notion as our kind of ways of working. That's where all of our information is stored. So we push all of the markdown files. The markdown files are essentially the set of instructions that live within the Claude skill into an active skills database within Notion. So you can actually open that and it's, you know, natural language. You can read it, you can start inputting. If you, for example, have updated your brand guidelines or have changed anything within your style and tone, you would update it there and then the actual cloud scale would pull from your Notion board. Because there's an integration between cloud and Notion if your team wants to make change requests to the actual skill. So people, let's say outside of marketing or outside of design, are using the skill and they're saying, oh, actually, you know, these charts are coming back a little bit wonky. The lines are distorted. We have a process for a skill change request where anybody on the team who's using the skill can input a request there. And we have our design team. This is where the governance pieces, like, is the design team still very important to this process? I just want to call that out. This is not one of those like, hey, we built this and got rid of our design team. It's like, no, we've built this to enable our design team to be able to do more of that really good foundational work that then can be shared across the entire organization. So they're actually monitoring the skill change requests and they're approving the ones that make sense and adding those to the actual active skills. And then the ones that are maybe a little bit more unique or one off, we're going directly to that requester and, you know, trying to work out an individual solution.
- [21:32]
G
Nice.
- [21:32]
A
Love it. Good, good comments in the chat about this one. Very relatable.
- [21:37]
E
Yeah. So I want to show you guys what this looks like. So on the output side, and again, like I mentioned, this is customizable based on, like, what you feel like your team is designing the most. But for us, it was really in these three buckets to start off with. So we created something that allows us to, to create infographics and this is using specifically the Claude brand skill and using the Figma mcp. So you're able to actually use the skill and prompt within Figma to create these types of infographics. You're inputting all the data directly into Claude and you're just baking out that infographic that then gets pushed to Figma and then you download it as a jpeg.
- [22:13]
A
Explain the Figma MCP for people that might not know what that means.
- [22:17]
E
So within CLAUDE there is a app store. The way there's like an app store and any other like on your phone whenever you go to download an app. MCP is a fancy way of saying app. So it's basically a Figma app that you download into your Claude instance that is able to communicate with the actual natural language chat that you're having in Claude and actually push those instructions directly into figma.
- [22:38]
A
Good job, marketer. It's an app like, come on, man, you know, that's why we need marketers. That's why we're not going anywhere. Because the engineers would call it the Figma mcp.
- [22:46]
D
We would be like, it's like an Apple.
- [22:48]
E
Yeah, it is, yeah. All of this, I mean, this goes back to like in SaaS. Any ecosystem that integrates, I think CLAUDE has done a phenomenal job with integrating with a lot of the tools that we typically use. So Figma is one that I know the design community will probably be very, very happy about. They also have a Google Slides integration, which is the one thing to note, and this is kind of my qualm, if anyone from Anthropic is listening, I would love for this to be resolved as soon as possible. But currently there are not bidirectional syncs happening between these tools. So what that means is if you were to create this infographic and push and basically build it within the skill and push it to Figma. Once you've moved that file over to Figma and you're going to download it, you can't make updates. So it's like you have to make sure that all of your instructions and everything that you want the image to look like is done on the Claude side. And then when you're finally ready and happy with the final product, you then push it to Figma for the final download. And it's the same thing on Google Slides. You basically need to make sure you're getting it to that 100% within the chat and then making any sort of tweaks, maybe manually in Slides or directly in Sigma. But right now there's not a two way sync, meaning when you push that content over to whatever app you're using, you can't make changes in the chat anymore. So I'm sure they're probably going to resolve that. But this gets you about 99.9% there. So if you do have to make some manual changes, it's not a heavy lift. Yeah. So this is some of the other stuff that we created. We create some sales enablement stuff. So building slide decks, building one pagers, and that is done directly through using the Cloud skill and the Google Slides integration. And then for all of my demand gen peeps, this is a huge unlock. I think I spent pretty much like 90% of my early career building dashboards in sheets that looked really bad. And then I'd have to figure out how to like tile the correct colors in the different bar graphs. But now you can actually create absolutely amazing dashboards within Claude skill by using your cloud skill for the design stuff and then using Cowork to actually parse your data and create dashboards with different views. They're toggleable and it's just such an incredible experience, especially if you're doing a lot of reporting and stuff like that and want to share stuff with your executive team members. It's a really great way to incorporate some of your brand design and build some really nice dashboards.
- [25:00]
A
Nice, Liz.
- [25:01]
E
And then I have one final thing to show you guys. So I wanted to show you how this works as a system. So what we did here was we have, like I mentioned, we use Notion to aggregate all of our tasks and we have the integration between Claude and Notion set up. So what it allows us to do is actually take the link directly from the task in a Notion board, drop it into a Claude chat, and that's what you see populating on the left side of the screen here. Basically said complete task number two for me, which is building a slide deck. This is all happening in Claude Cowork. The skill is being run, our design playbook is being run and the deck is being generated. So in about a minute or so here you'll see this populate and this is essentially how that looks. So once that you know you have your task, you run it through Cloud Cowork, it then opens up in Google Slides and you have the foundational part of your slide deck already finished. And again, it abides to your brand guidelines because it's pulling in from the brand skill.
- [25:59]
A
So is that how you use your slides for this?
- [26:04]
E
Actually, no, I did not. I knew that was going to be a trick question, it is a little bit easier to work within regular manual work when you're creating charts like this.
- [26:11]
A
But I agree, I look, we had an event two weeks ago in Arizona and I was like, I'm going to go, I'm an AI native now. I'm, I'm going to go all in and make my slides. And it was, it's impossible. I understand, like there's cool apps and you can use to make slides in there. But for me, the way that I like think about the slides and the structure was easier to, to get it done manually. Last question before that was great by the way. Really useful. So I just have one question before we get to V, so I just want to call us out. I think that there's a conversation now like AI people are bought in with AI and marketing, but then there's kind of like this like ickiness feeling where like, oh, if I just start writing all my blog posts and emails with AI, like is that really valuable anymore? Well, what I think is really cool that Liz just showed you, this is CMO of a legitimate company showing you a bunch of real use cases. And a huge part of the job in marketing is internal marketing. And what you showed is a bunch of like sales enablement related examples that make life a lot better, a lot easier. And it's not AI slop. You're not just like, look, I wrote this blog post with Claude, you know, so I think that's really cool. My last question is, how did you go from prior Liz to this Liz to AI First Liz, was there a moment where you just said like, man, I'm the cmo, I got to figure this out for myself. Like, how did you go from there to here?
- [27:22]
B
Hey, it's me, Dave. This episode is brought to you by our friends at Knack. Knack is a no code email and landing page creation platform focused on a problem every marketing team runs into. Have you ever had a really good marketing idea but then it takes forever to actually ship it out the door? It's usually not because your idea is bad, but because the process in the middle is slow. Briefs, more briefs, approvals, reviews, tiny fixes that somehow turn into weeks and by the time the campaign is finally ready to go out, it barely even looks like what you originally wanted to ship. Yep, that right there, that is the gap that Knack exists to close. Knack is a no code email platform built for modern marketing teams. They have AI built into the platform that lets you prompt ideas and instantly generate on brand email assets so you can create, review QA and launch your email all in the same place. No jumping between tools or messy handoffs halfway through after the email goes live. NAC also gives you performance insights and recommendations so you can see what worked and how you can make the next send better. So if execution is the thing slowing
- [28:29]
A
your marketing down, or you just want
- [28:31]
B
one system that takes you from idea to shipped, to learning, to improving, you should check out knack. Go to knack.com exit5. That's k n a k.com/exit5.
- [28:45]
E
Yeah, I mean, to take a step back, I feel like my role as a marketer has changed so much in the last eight months. Like, I feel like I've restarted my career, which is one of the most exciting things I think, in marketing in general. Right, because we reinvent ourselves all the time. But what I realized is I sat with my entire team, all the leads of my team, and I was like, we need to document our process. That's kind of the first thing, right? Because if you don't have a good process on anything, that means that things are breaking down. Once we've sat down and documented our processes, I started thinking about, okay, there's a lot of things happening right now in the AI world. How do we bring this together? How do we create space for efficiency in some of these workflows? And when Claude released its skill functionality, that was really, I think, the tipping point for us and especially the ability to now have an org wide skill because a lot of this great work was happening in containment. So I was spending hours learning this stuff, trying to figure it out, right? Like building these amazing. I mean, my team will laugh about this if any of them are on. But like I had them like building prop libraries and figuring all this out. But the issue with that was that it was so unique to the individual. Right. We weren't maybe doing a great job socializing all of the stuff that was happening because it was just like happening in individual pockets. When we realized that the Org skill had the power that it had, that it allowed us to really create an environment that everybody could tap into, that was the biggest buy in for the team. So it's like showcasing to the team like now all of the hours because a lot of this work is done in addition to. Right. Like nobody necessarily has like learn AI in their jd. It's like, oh, I'm tinkering with it on the side. I'm spending a ton of time doing it over the weekends. Not everybody has the privilege to do that. So when we showcase that to the team and we were like, now what you're building is actually going into a repository that will continuously build upon and you're no longer doing it just for yourself. I think that was the aha moment where we're like, okay, there's really something in this. And if we build. And to your earlier point around creating AI, slap, like we've invested hours into building these skills, right? This isn't like a one and done. Like, hey, I spent an afternoon building this and it's all set. Like I have my design team and shout out to Megan, our great art director that like dove right into this and all of the antics that came with it and just was like, I'm going to build this gate. It was hours and hours of feedbacking and figuring out what was the right mix of how cloud was responding to it. So like you're not creating a islap, you're just creating almost an AI employee for yourself for different processes that you feel like are repeatable enough to have a centralized repository. And those things have always existed. I mean, the reason notion originally got so big was that they were the knowledge base and like the internal wiki for companies. So now you're just taking that data and you're making it actionable. It's an additional layer to what you've already had. So it's been a really exciting journey and it's not over. Right? Like, I think anybody that tells you that they're an AI expert is. I'm certainly not. I think I learn things every day. My team is learning things every day. And I think it is really inspiring that we're able to build foundational work that allows people to self serve and feel inspired and creative as well and spend more time doing the actual, you know, strategic work and leveling up. Like we've redesigned our entire website and all of our design assets in the span of a month because we were so inspired by this.
- [31:56]
A
So Damn. Dan. Dan. Earth to Dan. Earth to Dan. Are you listening? She did the website in a month. The whole thing.
- [32:04]
E
Please note, not the development, but the look and feel. Yes.
- [32:09]
A
All right, cool. Good job, Liz. Hey. Well, let's bring up V and real quick, I just want to something that you just said really, really hit with me, which is like, I shared this message with our team a week or two ago, but I do feel like there is this thing out there that everybody's afraid to say out loud, which is like, I did my work with Claude, so is that cheating? Well, I use Claude to help Me write our newsletter. And guess what? It takes me a freaking long time. I spend a lot of time on that. I'm not doing it to get a 10 minute output. I send this to my team every week. I'm like, I just wrote the newsletter. Took me three hours. Took me two hours, two and a half hours. But that's because I have this, I love that framing. Like, I have this assistant, this helper working with me now. And I'm like, okay, so I just had Liz on my podcast and I want to write a newsletter about how she said this. But a couple weeks ago I actually had this guy V on the podcast and here's the transcript from his conversation. Can you take those? Can you find and can we research this? And it's like, that's why it's really fun. It's just a new way. And I've always felt that like great writing is actually just great editing. And what I'm able to do with Claude, it's like this constant game of like moving the Legos around versus like I'm in a room with candles and incense burning and I'm in flow state and I'm writing, you know, 1500 words. It's just, that's not how I write. And so it's been an amazing tool for that. Liz, great job. We'll see you out there. V. Glad to have you here. We have a senior creative director who has not revolted against AI. Tell us, sir, what you have been working on at Bitly, what you're going to show.
- [33:31]
F
Not yet at least. So today I'm going to share a quick project here that we recently did regarding video. It's not as exciting, I'm sorry as Michael Caine, but I think it goes through like kind of what the workflow is. The behind the scenes of this new AI video production world doesn't really change that much from like a traditional production. The only thing is it's way quicker. So like Luke said in the beginning that that guy, that freelancer, took a day, takes us a little bit longer, especially for this video. This is like a two minute video here that we did. But it really enabled us to create more video content in places that we would have never done video for, for example, customer education. This would have been a lot of investment for just a customer education video. But we ended up doing this, you know, for a couple hundred dollars in credits and a full video in like two weeks.
- [34:32]
A
Like you would have had to have that one person on the customer team who's like, knows the product really well and is great on camera. And that's very rare.
- [34:39]
F
Yeah. And like produce this the way we did it was more like ad style and it would have taken a lot of work and money to do this like this. So it's really been freeing in that sense to just like start creating video content with the vision that we want and the style that we want and things like that. So here's like a quick overview of how we do it. Like, we start with the idea, we prompt with Claude. Like you said, Dave, like the scripting is not one and done. We go in, we edit, we keep working back and forth with it. And then from that we generate a storyboard, generate the clips and then go into like a traditional production timeline where we go into the post production, the edit. And here you can see kind of how it came to life in a final video. So we're not using like Luke said at the beginning as well. We're not using a roll people talking to camera. We're using voiceovers.
- [35:40]
A
What does this change from your job, like your job as a creative director 10 years ago versus today? How does like having AI from a creative standpoint change the game?
- [35:48]
F
It's been a huge change over the past, I would say six months, but in a good way because now it's less about like selling the idea. Usually you're like trying to sell these ideas with mock ups and a lot of presentations. A lot of effort goes into decks and things like this. And now it's more like you have the idea, you start creating day one. And so that's been a huge shift. And I think for any creative that should be like more celebrated, instead of fearing it, you're able to like just go in and start creating right away.
- [36:24]
A
So it'll be like a 20, 20, 30 years ago. Be like, I don't see only hand draw everything. Right. You don't use folds.
- [36:30]
C
Yeah.
- [36:30]
A
You don't use Ghost Trader. A couple questions in the chat about the storyboard. How'd you create the storyboard?
- [36:35]
F
The storyboard. We use Nano Banana for creating the storyboard. We put in different prompts for different keyframes of a scene. Like I said, very traditional way. How you would instead of drawing a storyboard with a storyboard artist, you go in and you just prompt it with Nano Banana. You tweak it and then you get to wait.
- [36:57]
A
So what's the first part of this? Is like, I have an idea for a video for customer education. Okay, then what?
- [37:03]
F
Then you go, you prompt that idea into Claude.
- [37:07]
C
Okay.
- [37:08]
A
So you're in Claude, and you're basically coming up with the concept of the script and everything.
- [37:12]
F
After that script, you start storyboarding. You start figuring out what scenes you want for each part of that script. And then after that, it's a matter of creating those images with an ado banana or another tool. And you start from there because these video tools are great once they have an image reference. So always starting with an image reference is great, not just for time, but for consistency as well.
- [37:37]
A
What else should I be asking you right now?
- [37:39]
F
Anyone have any other questions?
- [37:41]
D
That was good.
- [37:42]
A
That was good. All right, cool. We'll come back.
- [37:44]
B
We'll come back.
- [37:44]
A
We have some stuff in the. We'll get back to Q and A, we'll bring you back up.
- [37:47]
G
All right.
- [37:47]
A
Carter's been patiently waiting for a track to explode on behind the Scenes. If there's one person that gets that reference, I'll give you a dollar. What do you got? Last one here. So Carter's going to show us an example, and then we have a bunch of questions in the chat and Q and A. And I want to take all of your questions and bring our whole panel back up.
- [38:01]
G
So.
- [38:01]
A
All right. You got the mic, Carter.
- [38:03]
D
Awesome, Dave. Thanks for having me. So if it's okay with you, I'd like to just kind of take you and the audience through kind of a quick history of how our processes have evolved and then show some stills from a image, so some images from a video that we did a lot of generation with. So I think where we began was really sort of very heavy copy based.
- [38:21]
A
Right.
- [38:22]
D
Whether it's early days of ChatGPT or using Descript, which is one of our great field enablement tools, a lot of it just started with copyright, whether it was generating scripts or generating titles and YouTube descriptions for YouTube copy.
- [38:35]
A
Right.
- [38:36]
D
For instance, I manage our YouTube channel, so being able to help generate ideas for those has been really helpful. One thing that we've actually done at descript is when I have a YouTube video that's ready to go, I'll upload it to descript. And we have some custom prompts that help generate really dense, rich SEO tags. Right. Which I can then put. Which would have been a whole workflow with the SEO team. But now I can generate a lot of those. And then I use the Vidiq plugin on our YouTube channel. So once I plug them in from description, Vidiq invalidate them.
- [39:05]
A
Right.
- [39:05]
D
So I think that's sort of our entry level.
- [39:07]
G
Right.
- [39:08]
D
Was copy. And then we started moving into voice generation.
- [39:12]
A
Right.
- [39:12]
D
Originally in descript and now with 11 labs, right. We do a lot of enablement content, a lot of demo content, and we would either get demos with somewhat low quality or ones that are just kind of hard to understand. And that's what got us into creating voiceovers. Right. I always really enjoyed working with voiceover artists, but knowing that they're in a library like elevenlabs is fantastic. And they're still getting credit and they're still getting paid. But also just the way that our procurement process was, it was just very kind of slow to have to hire an artist every time. So being able to do that work just really sped up our process and enabled us to take a lot of content to market very quickly. Where that's evolved to now is in a lot of localization. So in our APJ markets, there's been a really, really big push to try to get a lot of content that's originally recorded in English out for our Asian markets.
- [40:00]
C
Right.
- [40:00]
D
So whether it's using descript to just generate SRTs so that when we do live streams of content, we have translations, that's sort of our baseline. But now we've been spending a lot of time experimenting and fine tuning our processes for both voice cloning and for dubbing, right? So for instance, say I'm in a video and we want to clone my voice into a different language, right? We've been doing a lot of experimentation of what sounds better, right? Sometimes voice cloning sounds good, sometimes dubbing sounds good, right? For instance, we have an American speaker and we're trying to voice clone him into an Asian voice, right? Doesn't sound very good, right? So in that situation, we would then just use a dub, sort of like a pre canned Asian voice, right? It sounds much better, right? It sounds much more natural. But say sometimes we're going romance to romance, right? Say we have a customer interview with a Spanish speaker from Madrid. We want to dub that into French. Romance to romance sounds a lot better, right? So I think in general, a lot of these tools just require experimentation, but that's been the really exciting thing for me. Right? So that's sort of our audio component video. The thing I think has been most exciting for me, which a few of our other panelists spoke about, is just really being able to generate B roll and sort of what I would usually go to to find stock video. The example that I'll share is we did this really great video for a sales enablement conference that we had where oftentimes if I'm putting a treatment together or A storyboard, I would kind of be beholden to whatever stock video I could find.
- [41:32]
A
Right.
- [41:33]
D
Like, we don't necessarily always have the budget or the cruise of the availability to go shoot everything we want. So a lot of times you're pretty beholden to what's available with stock. So being able to generate very specific B roll shots has been really cool. With all that said, I think in the last year there's been an even bigger appetite than ever for actual production, like going out and recording stuff in the world. Right. We have these massive demand gen campaigns, whether it's our trends webinar or our recent AI summit, where we're still going capturing a ton of content. But for instance, in our trends webinar in London, we shot at a soundstage in London on this massive LED volume. And on the volume, the webinar was transportation based. Right. So we used a lot of generative tools to generate these like beautiful looping videos that were sitting on the LED volume behind our panelists. So really enjoyed that. Moving into some of the experimentation again, I think a lot of our panelists have covered this as well, is that some things you always want to have that sort of that human layer on top, right. And I think just in general, now more than ever, having human layer of just being able to curate taste, I think. Right. And really just deciding whether things that AI is generated is more important than ever. Right. So being able to experiment where we did some motion capture, we've taken some designs that our designers have done, animated those again, some of the voice syncing doesn't work as well, so a lot of it is just kind of again, experimenting, curating, seeing what works.
- [43:08]
A
Was there a point in like on the team, in the brand studio at UiPath, was there like a moment where someone was like, yeah, we need to start moving faster, we need to start be able to do more stuff like I want you all to go explore AI or did it come together where a bunch of people were just using these tools and then you bring it to the team, like, how did this transformation happen internally?
- [43:24]
D
To be honest, both we had some of our salespeople and you know, pre sales engineers were experimenting with some voice tools and some localization tools. Right. So that kind of pushed us to brush up on our knowledge base of those. Right. But we've also had a lot of messaging from, you know, leadership. My boss as well is our global creative director to accelerate these processes. Right. So I think it's both come from leadership buy in and it's sort of
- [43:48]
A
organically I feel like we had tested a screen share with you, and my team is like, yeah, you had something to show. Why don't you do that?
- [43:53]
D
Yeah.
- [43:53]
A
So we had done so unfortunately, I
- [43:55]
D
can't share the full video because it was done for an in Toronto sales event. But I wanted to share some of our really nice stills. All right, so this was a. Just some of the stills that we had done from our sales kickoff video. Right. All these were generated using different tools with nano 11 labs. Great cling, nano banana. But I just. I love the ability to put sequences together and just have these really brilliantly beautiful images created. I think they just. They really pop and they stand out. I think this is one of my favorite ones. Right. Where again, I wouldn't be able to find a shot like this in a stock video library.
- [44:30]
A
Right.
- [44:30]
D
But being able to generate the exact theater that we were hosting it at, put our name on there, put our branding on there, was just really exciting to me. And something that we've never been able to do before.
- [44:40]
A
That's a great. Like, we should have led with that. That's a perfect example of, like, I, hey, we're doing this event in San Francisco. I don't have a picture there. I don't want to use the, you know, traditional stock photo one. Let's put them there. Nice. All right, let's bring the whole crew up here, and we'll wrap up in like 10 minutes or so. We'll just do pop off on a couple of these questions and just whoever has an opinion on it, feel free to answer. This question is for Meredith. For something like all in video creation, what type of role sleep do you think you need to hire? Before AI Video, you'd need a motion graphic designer, but this seems to be a lot more broad, skilled. V. Probably a good one for you.
- [45:13]
F
Yeah. I do get this question recently a lot. I think you really need, like, a generalist, like someone that understands art direction and cinematography and a lot of those traditional art director roles and copywriters also. I see maybe in the future coming back to the traditional art and copy pairings. Imagine a world where you have two people working on the same video and you're prompting together and figuring that out. But you don't need to involve, like, a ton of other folks in the room. You can just go from idea to creation in one go instead of how it used to be done. So I think you kind of need more of generalist creative now that understand kind of everything instead of specialists.
- [46:07]
A
Anybody else want to build on that?
- [46:08]
C
Yeah, so we now have three different full video teams within 11 labs. So we have our brand videos, we have our performance marketing videos and then we have our like educational YouTube videos. And we actually still really value motion design because we find the best results at the moment are still like motion design with AI can get like really cool combination. And yeah, fully agree with V that actually going like the full stack creative of can you go all the way from like idea to script, generate it. And then the new role we've started hiring for is like the AI created producer is what we're calling it, where you're like producing with these tools when hiring. Quick tip is like we really evaluate for portfolio and also do like take home challenges. And now because they can do in like half an hour, they can actually go and create a full video and shoot which just would have never been possible.
- [47:02]
A
It's like the same thing that's happening to all these roles. It's like if you're a creative person, it's like your job is not going to get replaced by AI, but your job is going to get replaced by someone who knows how to use AI to do those things. And so like we still want a designer with design skills and taste and chops, but now you know, she's able to use these tools to get a better output. What did you say the role? AI. AI creator.
- [47:25]
C
AI creative producer. And I fully agree with you. We've got way more people than we've ever had all creating these shoots, creating these videos and Cool.
- [47:36]
D
Nice.
- [47:36]
A
This question goes to Liz and anybody else who has an opinion on Liz. Have you run into a lot of poor design output? This seems to be the biggest pain for us. It feels like AI design, this is totally related to all you. It feels like AI design output is maybe 50% there on a good day. And literally I got a text from my wife this morning. She's working on something inside our house and she's like, I hate Gemini. Look at this. This is not where I put this in the room. So how do you all deal with the kind of whack a mole of the output that you get back?
- [48:04]
E
Yeah, I think for us it's really the system that we've built and the change requests that we get, we're able to police what kind of things go wrong every time somebody uses uses it so that we're able to feed back into it in real time. I think that's the best way to do it. I mean we did do a ton of testing before we rolled out the org wide skill to our entire team. So we didn't roll it out in beta. We really tested it a lot within our internal marketing team before we said, hey, everybody across the company, come jump on this. We do still got some edge cases where, you know, bar graph is distorted or, you know, things are happening. But for the amount of edge cases we get compared to the amount of efficiency and like self serve that we've been able to create, the ratio is just, just uncharted. So I'd say just continue building on it. Sometimes it is quite discouraging because you're spending hours feedbacking on, on a model, but the output is worth it. And I do think that once you get to a place of. I don't think you can get to 100 perfection with any skill or any of these tools, like just to set expectations.
- [49:00]
A
No disrespect, that's like assuming that like we all always get exactly what we want in perfect designs from the human designers at the same time. And that's not.
- [49:09]
E
But you know what's funny? Like I feel like with AI, the expectations are so high, right? Like you always assume that it has to get it 100% right on the first try with no feedback or else there's no.
- [49:18]
A
I'm in my car. I'm in my car this morning. I'm talking to Claude. I'm asking like a really complicated question. I'm like, answer this in exactly five words. It's like, no, this is impossible. I can't do that. This one's from Dallas. When you use tools like Claude to create sales enablement graphics, can you export the output to vector? Can it use brand fonts so that when I open the file I can edit the font?
- [49:37]
E
Yes.
- [49:39]
A
Find Liz on LinkedIn. Blow up her. Blow up her DMs. Anything else that like you all have been on here. You're hearing what people are saying, like, what should we be asking right now? Anything worth discussing? This stuff in the Q and A is not really jumping out to me.
- [49:50]
C
One comment is often people, they talk about the cost of these video models and they're like, oh, it's like $3 for a generation. And I'm like, well, if you crash to sports car into a building, it would cost you a lot more than $3 per generation. But like you can actually just sit there and from, you know, within 30 seconds generate it. So I think actually the like cost, when you view it that way is like so much cheaper to get started for sure.
- [50:17]
A
We see it all the time. I think that we, because A lot of these tools are free. And because you can get so much value free, like, it's almost like, messed up our brains with how we perceive that we should pay for these things, you know, like, there was a question earlier. Someone's like, what are you all going to do? Like, the Claude, you know, credits pricing is changing. I'm like, everyone's just going to still pay for it, you know, or they will, like, completely revolt and another company will find, you know, they'll find something else. I think it's crazy what you can do. I mean, we have six people on our team, and I think Maybe we pay 2,500 bucks for Claude for the year. Like, that kind of seems insane. And when you compare that to, like,
- [50:53]
D
what I pay for some of these
- [50:54]
A
other tools, I'm not going to. You never know who's listening, so I'm
- [50:57]
C
not going to say anything.
- [50:59]
A
This question came in from Jose, and I want to answer this for Dave. That's me. Hi. You still spend hours on a newsletter. How is it better with Claude? The quality of the research, editing, et cetera? Yeah, it's exactly that. It's the. For me, like I said, I guess also the type of newsletter that I'm writing is based on the interviews that we've done, the conversation that I had. It's not like original hot takes, like Dave thought leadership. And so it's very much like moving the Lego pieces around. It's much more like writing and piecing together. The thing that I really like is the ability to take, like, I got these three transcripts. Here's this other thing that I saw. Here's this. Explain this. Like, I spend probably most of the time, probably 30 minutes of the upfront is like, prompting what I'm actually going to write and gathering all my materials, which is like, I think that's actually a perfect exercise. Comparable to this is like, I have failed multiple times as a marketing person, hiring agencies, and a lot of times that is on me because I was not clear enough in what I wanted. And it's like the model always became like, write a better brief. You're a bunch of creatives on this webinar, right? You want to get better, work with a creative team. It's like, we'll write a better brief. And so for me, it's like, the more time that I put in, like, creating the brief. And so my work now is like, I know our audience really well. I know this content really well. I am the subject matter expert. I'm going to use Claude to Help basically be my research assistant, right? Think about any of these popular, you know, news shows, TV shows, creators. They all have producers, right? Like Huberman Lab has become the number one health podcast. You think he's just like making up what he wants to talk about? No, they're doing hours and hours of research and pre production. So that's how I think about how I'm using it to help from a content creation standpoint. If I was like trying to write comedy and be silly and funny, I probably wouldn't have the same process. So, okay, this is great. Marketing rule number one, quit on a high note. And while you're ahead, leave before things go south. So we measure everything. We just want to roll a poll really quickly. Alison's going to roll a poll. We just want to get your quick reaction to. If you're here right now, I think there's 262 people are at peak live, which is awesome. Give us a quick rating on Today's session from 1 to 5. Was it helpful? Was it useful? Was it beneficial? We use all this, we take all this data to share it back and inform what we do in the future. I want to thank you all. Carter v. Liz, Luke, appreciate you coming on here. We'll put all your LinkedIn profiles and everything in the recap. We'll share the recording with everybody and we're going to do more stuff like this. There is a huge appetite to continue to share more real examples beyond the hype and beyond the fluff. Shout out to Luke and the team at 11 labs for participating and being our presenting sponsor for today. A lot of great stuff. Thank you all for hanging out here. We'll see you on the next session.
- [53:32]
C
Okay?
- [53:32]
A
All right, see you in the community. See you on LinkedIn. I'll talk to y' all soon.
- [53:39]
B
Hey, thanks for listening to this podcast. If you like this episode. You know what? I'm not even going to ask you to subscribe and leave a review because I don't really care about that. I have something better for you. So we've built the number one private community for B2B marketers at Exit 5. And you can go and check that out. Instead of leaving a rating or review, go check it out right now on our website, exit5.com. Our mission at Exit 5 is to help you grow your career in B2B marketing. And there's no better place to do that than with us at exit 5. There's nearly 5,000 members now in our community. People are in there posting every day, asking questions about things like marketing, planning, planning ideas, inspiration, asking questions and getting feedback from your peers. Building your own network of marketers who
- [54:21]
A
are doing the same thing you are
- [54:22]
B
so you can have a peer group or maybe just venting about your boss when you need to get in there and get something off your chest. It's 100 free to join for seven days so you can go and check it out risk free and then there's a small annual fee to pay if you want to become a member for the year. Go check it out. Learn more exit5.com and I will see you over there in the community. Today's episode is brought to you by Converter. They're an enterprise lead data management platform. If you're running marketing at a large B2B company, like many of you listening right now are, you know this problem well. Leads come in from LinkedIn, webinars, events, content, and the data is a mess because one form captures job title, another one doesn't. One says United States, another one says usa. By the time it hits your CRM, records are incomplete, fields don't match, and your routing is broken before the lead ever touches a sales rep. It's annoying. And now that everyone's plugging AI into their tech stack, bad data isn't just an inconvenience, it's actually a real liability because AI is going to scale whatever you feed it. Feed it garbage, it's going to scale garbage. Converter is the layer that sits between your lead sources and your systems. With Converter, every lead gets validated, enriched and standardized before it touches your CRM or marketing automation. This gives you clean data every single time. Companies like Microsoft, Amazon, Oracle and Stripe use Converter today. So if you're dealing with this, you're not alone and there's a great fix. You can check out Converter right now at converter IO exit 5 that's C O N V E R T R IO exit 5 converter IO slash exit 5.