Unlocking B2B Intelligence with AI Workflows

Drew: Hello, Renegade Marketers! If this is your first time listening, welcome. If you're a regular listener, welcome back.

You're about to listen to a Bonus Huddle where experts share their insights into the topics of critical importance to our flocking awesome community, CMO Huddles.

In this episode, Dave Brong shares how to move AI beyond content and into everyday marketing workflows using a live project with CMO Huddles. He walks through how to spot repetitive work, map the process before bringing in AI, and automate it in a way that cuts out those annoying manual steps. If you like what you hear, please subscribe to the podcast and leave a review. You'll be supporting our quest to be the number one B2B marketing podcast. All right, let's dive in.

Narrator: Welcome to Renegade Marketers Unite, possibly the best weekly podcast for CMOs and everyone else looking for innovative ways to transform their brand, drive demand, and just plain cut through, proving that B2B does not mean boring to business. Here's your host and Chief Marketing Renegade, Drew Neisser.

Drew: Welcome, Huddlers. Today, we're diving into a topic that every CMO wants to crack, but frankly, few have the patience or the technology chops to execute, which is: how do you use AI to actually streamline marketing workflows? We've talked about content a lot, and that's relatively easy, but workflows feel like the next big opportunity. And we've talked about it a lot in Huddles, but getting it there is a little bit more of a challenge. And our goal with this exercise was to take a very specific, repetitive, manual task that we did at CMO Huddles and turn it into what now feels like automagical moments, and it's kind of crazy. So joining me is someone who's been elbows deep in wiring this magic. That's Dave Brong, VP of Technology and AI at Level Agency. Level Agency is a longtime partner of CMO Huddles, and Dave is not only a brilliant technologist but also the reason I haven't thrown my laptop out the window in the last six months. Together, we've worked on a truly transformative project that's changed how we run, analyze, and learn from over 350 CMO Huddles conversations of various types at scale. So we're going to break it down. So Dave, first of all, welcome. Great to see you. 

Dave: Thanks for having me. 

Drew: And where are you this fine day?

Dave: I am a little west of Baltimore, so we're in the cold and warm situation of the United States right now.

Drew: Got it, got it, got it. Okay, so we laid this problem out for you initially as lots and lots of recorded meetings with lots of manual steps, and maybe you could sort of take us through how you wrapped your head around the challenge we shared with you.

Dave: Yeah, it definitely was a challenge to get started because I didn't really know all the nuance that went into your internal processing of your meetings. I only had my experience from running like client calls and internal calls, and yours just cascaded in more detail on top of what I was used to. So it took me a while to even just wrap my brain around the different types of Huddles you had. That was the biggest challenge there. And that challenge, when I work with people on any AI automation projects, always comes down to articulating what the root challenge is in the first place. And this was really just a way to understand what you had going on to know how we could solve those different problems. And I say different problems because it wasn't just one. It was more.

Drew: We have, in addition to Peer Huddles and Bonus Huddles like this one and Career Huddles, we have one-on-ones with lots of CMOs during the course of a month. We have partner calls. There are a lot. And the way it used to work is we'd record them on Zoom. Elie would move it from the recording into Dropbox. She would upload the recording into Otter because the transcription wasn't very good, or she'd take the transcription from Zoom and then she'd put it into ChatGPT to clean it up, and then it went into all sorts of other areas. And when you started adding it up, it was a pretty big challenge. And so maybe you can, you know, when you looked at it, how many manual steps did we end up finding, and walk us through what used to be manual and now how it works behind the scenes.

Dave: Yeah, so there is at least a good 10 manual steps that were repeated for every single meeting you had. As you mentioned, you have the meeting, but then what do you do after the meeting? You have to go hunt through Zoom and find the transcript and then figure out what to do with it next from there. So in those steps too, then you had that aspect of experience—how it comes into play. Like, what do you do next? What type of meeting was it? What do you actually do? So you can reference documentation or just your head and knowing how to do it from there. So a good 10 steps. In walking through that process, it can be straightforward, but it can also be a little complicated because then you get into the recap aspect of things. What type of recap do you need? Or what are you going to do with it next? Is it going to be a podcast episode, perhaps?

Drew: And again, we're being vague, but we promise we will show you this tool and some of the output. But if you take 10 steps, and this is one of the key insights, right—if you're looking for what automation projects should you think about? Well, we had 10 steps, and there were probably 30 times, 35 times, 12...

Dave: Right, right. You're repeating those 10 steps multiple times every single day.

Drew: And honestly, we weren't getting any better at any of them.

Dave: You're waiting on the systems half the time. You're waiting on Zoom to finish processing or for Claude to give you the first draft of the summary, and you just sit there staring at your screen, waiting and waiting and waiting. So it's not only manual repeated tasks, but it's also wasted time in which you end up just sitting there.

Drew: And so this wasn't just about taking Zoom transcripts because ultimately there was a fair amount of integration involved. I mean, we do have HubSpot. There was an opportunity to, for example, track so that we could associate someone from Huddles—they attended the peer huddle. We could sort of put that on their profile. We may even be able to add some knowledge to the HubSpot and actually have a customer relationship management tool. But then there's Dropbox and Gemini. Can you talk a little bit about this integration and which pieces were trickiest to connect, and how you solved those?

Dave: Sure, so there's four main pieces. So you mentioned Zoom, and then you have HubSpot. Just about everybody has HubSpot nowadays, it feels like, or some CRM. That CRM is your lifeline to everything. When it comes down to it, it's your database. So in connecting Zoom meetings to HubSpot, the meeting schedule tool is perfect because we can set these up. Those meetings are associated with the contacts. Great. But we don't know what contacts actually talked about within those meetings, so that was the first step for the challenge—relaying some of that intelligence back into HubSpot after the Zoom meeting is over. So that's part of our processing there, and actually extracting that intelligence from the attendees that rates the timeline within HubSpot under their contact records in HubSpot. So that's the first part of the big challenge there in processing. All that we use, Gemini, to do the AI processing flow from there. So with Gemini and their Pro model, it's very good at statistical or structured type of data where it's not going to go off and just make stuff up for the sake of making it up or making it sound good. It's more grounded in facts than anything else. So those are the three big ones, and then Dropbox, as you mentioned, that's your backup. That's your storage of everything. So we could store it in HubSpot. We could store it on Google Drive, wherever. You use Dropbox. It was already used by your team, and everybody knows how to use it. So in the processing, we're just automating the step in taking Zoom's data and putting it into Dropbox for you, but also all the processed results we're also putting in your Dropbox. So now one meeting ends up having 10 different files associated with it.

Drew: Yeah, and finding those files used to be really, really hard. And you know, if you think about the challenge of just finding the information on your laptop or in your email, that's just what we ended up having was a centralized location. And again, that was a sort of unexpected time saver. It occurs to me that one of the things that we needed to do, which I think is worth it—seems small, but I think is—we, in order to trigger the system, we renamed certain meeting types, right? So we knew what a huddle or one-on-one was, what a partner one-on-one was, and so on, as well as peer Huddles and bonus Huddles and career Huddles and transition team Huddles, right? So the system would be able to identify that. And I think that's an important, sort of small, but really good starting point, right?

Dave: Yeah, it's a foundation. It's a good data cleanliness foundation that we had to start with. Meeting IDs change all the time, so you don't know what meeting type it is when the IDs change and your HubSpot meeting scheduler schedules a different meeting event there. So we just standardized topics. And it's topic prefixes, really, so "one-on-one" or "transition team," "peer Huddles," things like that. And that's what's actually used to determine what type of processing—simple as that, you know, not overly complicated. We just prefix based on the Zoom title.

Drew: Okay, so, you know, our main goal was to just initially get these meetings into a place that we could find and eliminate a lot of manual steps. But, you know, I use this word automagical. And can you talk about an example of a moment where this project even surprised you at what the system can do now?

Dave: Once we got that first layer of the processing complete, and this was even with a small subset of meetings, maybe the past 30 days, I just randomly asked a question to it. I was curious. I just asked it like, "Who has experience with Salesforce?" And I was able to see the intelligence come back from the system that relates to conversations you've had about Salesforce with your community. So that was that magic moment for me, where you can't just search files and files and files and find that result. You could, but it's going to take you hours. It took two seconds to return that result for me. And that was pretty cool.

Drew: And what makes me so excited is one of the features of CMO Huddles Leader program is that we will match you if you have a question about a particular thing. Let's say you're migrating from Pardot to HubSpot or something like that, you would want to know, well, who's done that and who could I talk to? Well, for the last five years, that information has been in my feeble brain. And so I'd have to go back and think, "Oh, I remember talking to Peter Fincher about this. Maybe I could get him to talk to Ryan about this." And that is not a very efficient or scalable system. This was a real big unlock for us, and it's funny we're working on it now because someone recently asked, "Hey, I need to talk to someone who owns 100% of pipeline." But pipeline comes up in so many calls, it's a little tricky for us to find it, so we're going to have to fine-tune how we ask the question. And what you told me today is the system's just going to keep learning, right?

Dave: Yeah, that's exactly the point of it. So over time, the more meetings you have, the more intelligence that gets put into the system, and the more semantic relationships. So 100% pipeline, pipeline itself is a very generic term, overused term, so you can't just search for pipeline, because you lose the nuance in what pipeline means. Over time, as the system builds on top of itself, it will actually learn what pipeline means to your community. That way, your results are better.

Drew: And as you all are thinking about this, obviously our situation is not unique. You all talk to customers all the time. You're probably doing it with Gong, and we're talking about creating a system, and certainly Gong is building these kinds of things, but there may be custom use cases that you have, as we did, and so a plug-and-play solution wouldn't do it. So help us, Dave, think about for marketers. When does it make sense to, you know, build your own like we did here, versus try to find a system that already does some of this?

Dave: Yeah, so let's use the term flywheel on this one. Okay, we've all seen the flywheels, or probably talked about them this week, at least. When investigating systems like this, think of it as a circle. You have to start something small. What's your initial idea? What's the easy side of that idea? Collecting transcripts, processing transcripts, what's the next step on it? Trying to then get into an intelligent database or sync into HubSpot. So some of that stuff you can build yourself. Some of it's really easy to build. Also, low-code systems and AI nowadays make the entry level very obtainable by anybody. As you iterate and go through it, now you actually know what you want, so you spent the time, you invested it, to learn what the idea is and how it's going to grow with your team. Now you can go out and shop it around. Now you can go talk to the Gongs or the Zooms of the world and see what other features they have that can solve what you're looking for. Even then, if you don't quite know what the end result is, you have options. So going to Gong and saying, "We need to find a partner that can help us further develop this," or internally, if you have engineers, taking the same idea to them to figure out if they have experience in developing something like that as well. So really, the short of it is, by starting small and iterative, your options are actually greater than what would be jumping all in into a proprietary system, for example.

Drew: One of the things that I'm going to say is a myth out there is that anybody could get a Lovable or a no-code platform or something like that, and you could build a system like this, without any code, without any experience. And one of the things that was interesting to me is you just use the term engineer, and you have a lot of technical background. How much of this whole project is AI? 10%? Maybe that's down from where we were, which was 20% when we started about a month ago. So that's an important insight here, too. In order to get the value, the expectation out there is, "Oh, you just use AI and AI will make this code and whammo." But the part that, like, and you had a really interesting point about that, is, use AI when you need AI. But talk a little bit about why this is only 10% AI and yet it's still delivering tremendous value.

Dave: Yeah, so 10% AI is really just coming down to the extraction and the processing of the recaps, the way you want them to be processed. Two examples there. So one example is transcript cleanup. You asked me to clean up the raw Zoom transcripts, which were a little crazy because they have timestamps and other stuff in there, so that processing is just code. Processing has nothing to do with AI. A lot of people will jump to AI and say, "Hey, AI, go and reformat this for me." The risk there is you may not get consistency in the results. AI could sometimes, today, give you the right result. Tomorrow, a model could be upgraded, and now your result changes, so you have to go back and you've got to fix all your prompts and all that stuff. For a transcript cleanup, very easy, just processing with standard coding techniques, and we give you a clean transcript every single time. On the flip side of that is extracting quotes or extracting takeaways or action items. That's where AI comes into play, because it can understand the meaning behind what the words are.

Drew: And I think this would be a good time. I want to share the screen here so people kind of get a sense of what we're talking about. I can do it from here. Share screen. This will be a live demo. Tell people what they're seeing here, Dave, while I find a recent peer huddle.

Dave: Yeah, so we talked about HubSpot and Dropbox as well. This is just a simple web interface. There's three different pages to it. What you're seeing here is a list of all the huddles that we process to extract those insights from. So Huddle 101 right there at the top is the latest one that Drew hosted this morning. I guess it was. I didn't record it, so yeah, and that's why we have the red X right there. It wasn't recorded, so it couldn't be processed, but it's still in your list of huddles that you had, in case we want to do anything with it in the future.

Drew: I'm going to look at a peer event. We had a peer huddle on events this week. I'm going to open up the meeting. And what do we see across the top here?

Dave: Yeah, so these are the different outputs from the processing. So there's attendee recaps that I mentioned before, the main meeting recap, which is the insight extraction across the entire meeting, the cleanup. We have a different type of meeting summary, so a recap and a summary. The summary is just more of a raw summary than what your specific recap is. And then we get into the other metadata-type things that you will expand, or you may expand to over time, so that comprehensive metadata that can sync into HubSpot, the outrageous quotes that you could use for socials or whatever you want to do with that. Ultimately, the tabs are just the process and results.

Drew: And just on the outrageous quote, so many of you know I write every Saturday what I call my rants, editorials on LinkedIn, and they always start with a quote. Two weeks ago, this system was up. I said, "Oh well, what am I in the mood for as a quote?" And sure enough, I went to the outrageous quote, and it came up with a couple that I thought were really good thought starters. So again, I didn't expect this system to be able to save me time. What I used to have to do is go back through, say, the last four meeting transcripts, and try to find a quote. This one just surfaced it up, but it's surfacing up many. I'm still in charge. I get to choose which one, and if I don't find one, I go back to another meeting. Meeting recaps are something that come into your inbox every single Friday. Those used to take me a good hour and a half to go through a really dirty transcript and clean up the transcript, clean up the quotes, and then try to figure out, what are the insights, what's interesting? What this system provides usually gives about 12. I take 12 of those. I go, "Oh, I think eight of them are really important to our community." So again, human is in charge. But what used to take an hour and a half to two hours now takes about 15 minutes. And again, it's a repetitive task. I did it every week. Nobody yet has complained that the value has declined of those recaps. Our open rate has not dropped. If it did, I'd start to worry. And then every six weeks or so, or four weeks or so, I go back and do it manually just to see what would happen and how much better it is.

Dave: There's some hidden features here that I haven't even brought up with you. So besides the manual recap, if you were to chat with the system on the right side here, which allows you to chat with the entire meeting transcript itself, you could take your manual recap and paste it in here and ask it to compare to its generated recap. You can learn the differences. So then you can adjust the prompts, which are basically the Drew prompts. This is how you want the recap to happen, by using AI to improve upon itself. That's the takeaway.

Drew: Interesting, and I can't remember, so we're sort of sharing all here. Did we give you a bunch of examples? We did give you a bunch of examples of recaps so it learned on human input.

Dave: That's exactly how we started. We reverse-engineered it from prior examples.

Drew: Right. So this was the problem that we wanted to solve. And by the way, the meeting recap shows up within, I don't know, a few minutes after the end of the meeting. So it's kind of like, wow. And again, these used to go through three different places, and then I'd have to go find it somewhere. And now everything is all here, but I want to get to the sort of unlock of real extreme value. And that's really this box right here. That's part one of where we see the value coming from, which is, how do we match people better? Is there anything else that is worth, while we're here, to show? Like, what does comprehensive metadata really mean, and what does that mean for the future of the searchability of this data?

Dave: Yeah. So when we talk about intelligence, that word itself doesn't mean much as to what you're going to do with it, so the metadata is the extraction of that intelligence in schema. So what's the tech stack? What are business challenges? What are goals? What is their history? What is their timeline of an attendee? Are they transitioning away from Marketo to something else? Has that been completed? And that builds upon itself. So say you have another peer huddle, same thing: peer huddle for events next month, same attendees join and they talk about what has changed with them. This system processes those changes and stores that all within the record of those attendees. So that's that temporal intelligence that changes over time to just further enrich your entire community. 

Drew: Great. Okay, let's go off of this. I think people get a sense of what it looks like. 

Ad Break: This show is brought to you by CMO Huddles, the only marketing community dedicated to B2B greatness and that donates 1% of revenue to the Global Penguin Society. Why? Well, it turns out that B2B CMOs and penguins have a lot in common. Both are highly curious and remarkable problem solvers. Both prevail in harsh environments by working together with peers, and both are remarkably mediagenic. And just as a group of penguins is called a huddle, our community of over 300 B2B marketing leaders huddle together to gain confidence, colleagues, and coverage. If you're a B2B CMO, why not dive into CMO Huddles by registering for our free starter program on cmohuddles.com? Hope to see you in a Huddle soon.

Drew: And I think this is important too. I mean, this is so much more than efficiency, right?

Dave: Yeah, it's amplification. Because in the past, if you were only able to do five of these per week, you can scale not only yourself, but the insights that go to your community at the same time. So if you could only do five in the past, now you can do 20, and it's just amplifying all that potential. It's unlocking more potential.

Drew: Yeah, I mean, there's just no doubt I'm a beneficiary of being able to get this information faster. Elie is now able to do much more. Melissa does a lot of the one-on-one matches, and now, instead of just going to me and saying, "Hey, who should we talk to?" there's an opportunity. Now recognize that we've been doing this for five years. There's five years' worth of data, and this has maybe four months. So this will get smarter over time, and I think that's really important in a system like this. You know, it's not all in there yet.

Dave: Yeah, it's definitely going to grow, just as an employee would with you.

Drew: Okay. There is—you talk about this term "semantic search," and I know we sort of covered it, but I do want to make sure that we sort of clarify it because I think that's important. What is it? Why does that matter?

Dave: So to generalize, there are two different ways to search for something on the internet. You go to Google and you type it in—you type in your keywords or your phrases—and you're matching based on those words or phrases as they are within a result somewhere. Semantic search takes it more than that. It takes the meaning and the intent of what your question is, and it matches it to the meaning and intent of other words or phrases that are similar. So, Salesforce example: some people will ask, "Who has Salesforce experience?" Some people will ask, "Who has SFDC experience?" Those two keywords are completely different, but semantically they are the same. So now your results come back depending on who says what and how they say it. They match better now.

Drew: Got it. I think basically that's better search for trying to do the matches here. Someone asked a question about the outrageous quotes, and just so you know, those are associated with every meeting. So I know that if I was going to the peer huddle on budgeting and I wanted to pull some outrageous quotes, or at least look for one—and I only need one a week, or at least for now—they're all tagged. Yvette, thank you for asking that question. And we've talked about the recaps. I mean, when we record a bonus huddle like this, it'll go in here. It'll have a transcript, and that's the sort of beginning starting point, because these become podcasts. So that eases that process a little bit for us as well. There's another benefit in that that we hadn't anticipated, but it will be there. And importantly, like for Transition Team Huddles, which we do twice a month, we record those. Sometimes we have experts that come in. Sometimes—often—we have folks who just found a job and they share their experience. And again, what's great about that is we can go back to those transcripts now, and we can pull those, and we could take the last, say, eight times we had CMOs who found jobs and sort of, let's find the commonalities. What did they do consistently over time, right? And it's all there for us. Before, it would have been so hard to find this information. And I'm pretty sure that if I said to you, "Hey, Dave, we want to build that functionality into this," you could probably figure out a way, right?

Dave: Yeah, you actually said that to me a few weeks ago with the outrageous quotes. I didn't plan this, right? You said, "Hey, wouldn't it be cool if we could pull quotes out?" And I said, "Sure, give me five minutes." That's all it took, because we built this foundation that will expand with you over time. So where this would have come into play in the past, like manually to do something like a new recap or pull quotes from all your huddles, you'd have to go through and find every single one of them. Now, all we do is add a new analysis processor and run it, and it's done across everything.

Drew: So a lot of folks are nervous about AI replacing human judgment. In this system, it feels more like AI is amplifying human value. What are your thoughts on that?

Dave: My approach is always amplifying the human value or the human potential. You know, there's a lot of work we do as marketers that repeats, and AI, really automation—anything we can do to ease that repetitiveness—allows us to unlock and move into higher-value activities.

Drew: Let's talk about if someone wanted to build a version of this for their organization, how would they start? And are there any traps for them to avoid?

Dave: Yeah, starting now—just start curious. You know, take a step back and look at the systems you have in place or the data that you have in place and just ask the question, "What can we do differently?" You know, we all have phone calls. We all have recordings with customers or sales calls or whatever. So just think about that. What can you do to amplify that? Getting started, the same thing is curious. You know, Replit or Lovable or these little app builders right now that are AI-driven—you can take your idea and you can chat with it back and forth and have it prototype something out for you just to see if it's matching what you're thinking. Those are very easy starting points. Within a matter of minutes, you can get something, and then you can see if it's actually going to work or can evolve the way you're thinking it should.

Drew: I love it. And it's funny—we had Carilou Dietrich, who worked with Lovable in their first eight months, which was pretty amazing. First eight months—it's quite a story. I have yet to try it myself, but it is definitely something. But, you know, I'm so happy with this thing. Let's just revel in this. But talk about tech stack for a moment. I know that there's a lot going on behind the scenes. What is this thing built on?

Dave: Yeah, so this is just a simple private server, private cloud for you, that is Python-based with a SQL database. That's it. No other tech involved in there, and just connects to Gemini for the AI processing. 

Drew: You did some API integration, right?

Dave: So it connects to Gemini for the API, Dropbox, HubSpot as well. And that's APIs—we've all been doing that for decades now at this point. So it's all a tried-and-true process. 

Drew: And Zoom, 

Dave: That's right. 

Drew: But I'm pretty sure very few CMOs on this call would actually know how to connect an API.

Dave: You'd be surprised.

Drew: I want to be surprised. 

Dave: Yeah. So in the instance of how I set it up, I set it up as a developer. But take like the Replit and Lovable example. Those systems make it easy to connect to your Zoom account or to your other app accounts, and it uses the API, so same thing. It does the exact same thing. So that's the nature of where we are right now with technology. The entry level is very, very low. All you need is that curiosity to get started.

Drew: I love that. But so when do you need to call in the engineers?

Dave: When your prototype is ready, or before it's ready, really. When you're in this situation where you're saying, "Oh, this is a great product. This is a great idea," give someone a call, show them your idea, brainstorm on how to actually launch your idea. So Replit—you can launch it—but they're not necessarily ready for prime-time launches yet, especially around data security. And then the other reason to call someone for help is to help break apart your idea. So nitpick it a little bit, play devil's advocate, play the "what if" game. And the combination of the two—like you're ready, let's have someone else also say you're ready. You know, that's really all it takes right now, just a second opinion.

Drew: It's interesting because this came up a lot. Marketers have gotten very used to doing tech on their own with, you know, whether it's a rev ops or a marketing ops team, and sort of on their own little IT. And suddenly, that word is coming up more in marketer conversations, particularly when it comes to compliance and security. And if you're a global company, you know the various things that you need to comply with. But it used to be the place where projects go to die if you were a marketer.

Dave: Yeah, from my perspective, where projects start or new projects start, let's put it that way. But yeah, there's so many different aspects of operations nowadays. It really comes down to who has the time to get started. So yeah, historically it's been, you know, where things go to never get talked about again because they're always too busy to do anything. But now, with marketers and that barrier to entry being lower, you don't have to rely on IT to get started. And that's the whole point of where we're at now with AI and automations—just getting started.

Drew: Yeah, I think that's really important. This also came up with the Super Huddle quite a bit, which is it feels, even though some people don't like this notion of go-to-market engineer or the language of it, it feels like every CMO needs to have someone on their team. I've been calling it AI ops who can do more than write prompts.

Dave: Yes, very important to understand what that next level is beyond prompting. And really, there's levels of prompting itself. There's bad prompting and good prompting as well, but it's consistency that you want to look out for. You know, when you create a prompt, can you consistently get the output that you want and need from that over time?

Drew: If we were to sort of look at the biggest lessons learned, you know, one for me is this started out in my mind as an AI project, right? We were going to solve this with workflow. And it feels like this was, yes, it was about AI and it played a really important role, but there was a lot more to it than that. Were there some other big lessons learned for you in the process of building this tool?

Dave: Yeah. So I've built a lot of tools similar to this over the past couple of years, and the one consistent takeaway is always what I mentioned before: articulate your challenges up front, try to get the idea out and put it on paper. And that makes success criteria and makes any of these projects easier. But yeah, every project is slightly different still. Every way to approach a project is different. Sometimes you can start with Vibe CTO, sometimes you have a code base to pick up on. Sometimes it's migrating from one CRM to another and trying to automate pieces of that. That's what I love about what we do as marketers, as agencies and all that—every day is a little different, and it's a little bit chaos and a little bit structure at the same time.

Drew: And I do, I'm going to emphasize this point again at the risk of repeating myself, which is we started out—this was an efficiency play, like how do we get rid of repetitive steps? It became a lot more business value once we realized, oh, we're gathering all our information, it's going to be in one source, and we are going to be able to serve our customers better on the simplest level because we can do one-on-one matches. But I think there's a lot more to that. I think the knowledge that we're gathering is sort of building up this collective intelligence of the community that, if we really start to think about it, you know, it's probably a GPT in and of itself, right?

Dave: Yeah, could be. There's really no limit as to what you could do next with an intelligent system like this, you know, creating custom GPTs or exposing a Slack bot to your community to allow the community to ask questions too. This is the starting point of what a new future could be for what we all do when it comes to community engagement.

Drew: And I do want to emphasize there's the—it's funny that we parse the data in two different ways. There's the anonymized data which goes into certain aspects, you know, whether it's a recap or some other things, so that is in one. And then there's the things that people have expertise in that we're particularly interested in tagging, right? That's the main thing so that we can do the matches. But it does get into sort of privacy, data security and all of those things. Just talk a little bit about the security and what we have here and why we won't open up as a GPT just yet.

Dave: Yeah, so security is the number one most important thing when it comes down to processing our data or our customers' data, and just taking steps to make that database—just focus on the database. For example, your database is in a private system that only you have access to. Okay, every database should be that way. Nothing special there. So we use the term AI and automations, but really what it comes down to is we have an infrastructure that is built on the past 20 years of infrastructure. We didn't do anything new. We didn't let AI code something new and risk exposure of the data. It's all there. And then the AI processing, because we do some preprocessing with it, we're removing certain things that could be considered private or industry-specific, things that we don't want to process with at all anyway, so we filter out some stuff from there.

Drew: So I want to open this up to the audience. Oh, Ryan, you have a question?

Ryan: So Dave, just tell me about your business. How much of your work is actually doing this type of thing for clients right now?

Dave: For clients specifically, it's still lower than what we would like it to be as an agency. A lot of what we're doing is internal, just trying to get those internal tools and foundational learnings right for all our employees. First, clients are in a similar situation where they're trying to understand how AI or how new automation can help them. Every conversation I've had was brainstorming activities, not action yet. So we're getting there, but there's still a lot of internal work to be done.

Ryan: Okay. And how do you charge for this service?

Dave: That's a great question. It depends on what's going on. So we—at Level Agency, at its core, we're a performance marketing agency. So, right, we use AI and automation to accelerate that entire customer lifecycle journey. Think like turning attention into conversion and then conversion into loyalty and retention, things like that. So we're using AI and automation systems to help with performance marketing. So that's where it helps us internally because we use that every day to better report or better adjust media plans for our customers. So how do we charge for that? Right now, it's bundled into our retainers. It's like the evolution of evolving as an agency, okay? Way I look at it is, if we don't evolve, we're going to be obsolete.

Ryan: Yeah, you stand out by offering this.

Drew: Cool. Kathie, come on. What did this inspire in your mind? And I know you're just getting your business started, but given what you heard?

Kathie: You know, I myself am thinking about what are the workflows that I can build as a company of one that might be relevant. So it's a little bit of a different space. I guess I would be curious, from your perspective, Drew, on why you selected Level.

Drew: They've been a partner of ours for quite a while, and when we were talking about a project we could do together, they showed me their tool. I said, oh, I want that, but I want it differently. And so it was clear that this was not going to be from scratch, but it was going to be something new based on an iteration of something they had started. So it was a lot easier to get going since they already had a base for it.

Kathie: So you did a customization off of their offering.

Dave: Ours right now is Zoom processing, but then we send recaps and takeaways into Slack for our team visibility. So Drew's was an extension on that for additional processing that in time we'll probably do internally as well.

Drew: And we didn't do things like that. I was thinking about it because there are other calls where, you know, we could be pulling out next steps. And I know Zoom does that. And Kathie, we talked about they do a terrible job at it, so none of those tools. But it would be interesting to see if we could get to the point where it was actually good at that and then could even automate it further. Okay, Bindu Chellappan, what's your question?

Bindu: How do you think about which processes to automate? And, you know, coming from a big company with a lot of people, it might impact a lot of people's work, right, with what you might be automating. So how do you think about that?

Dave: Impacting work for people is always a forefront concern, but it's unlocking more higher-value activities, as I mentioned before, for those people. So what do we do first? It's very easy to say pick off the easy stuff, the low-hanging fruit—automate email labeling, for example, or automate customer support tickets and processing it to the right person from there. So there's automations there that people do, but time is an issue. Time back to your customer, for example, that's a great way that you can automate to speed things up, to make your customer experience better. I don't focus myself on what can we automate that can reduce our employee count. I don't believe we should ever be looking at it where we're at right now in that regard because there's so much all of us can do if we had more time. And that's what it comes down to.

Drew: And I think, Bindu, just to build on what Dave said, in terms of ours, we started looking just for what are the things that we do that have the most repetitive steps every single month and touch the most people. Now it happens that most of those steps were Elie, our assistant, but still, a lot of them involved me later on down the line, having to go find the transcript, going to look for pulling something out. But I think if you look at repetitive tasks as a first thing, and obviously repetitive tasks that there's no business value, right? And we wanted to take repetitive tasks so that we could add business value, and that was the thing. Yvette Stoltman, you have a question?

Yvette: I do, and this might be for you, Drew, but maybe Dave. So we have on our radar a project to do something similar with our transcripts from sales calls so that marketing can have more insight into those conversations, understand pain points better. We really rarely get that insight. But I'm curious, did you already have a database of your transcripts that you were building when you started to do this project, or was it just like, oh, you know, we need to go gather all of this?

Drew: We had a collection of transcripts. I wouldn't call it a database. They were organized. It's called Dropbox.

Yvette: You had a starting point where everything was already, right?

Drew: Yeah. You have a database, you need data. Yeah, okay.

Yvette: I was just curious because we're going to be starting with, like, how do we get people to start centralizing the storage of their transcripts so we can start to have access?

Dave: And for us internally at Level, the starting point before I even thought of a system to create around it, the starting point was turning on recording in Zoom.

Yvette: Yeah, that's going to be another big starting point for them. Some people record, some don't. It's not always consistent thing.

Dave: So we created the consistency, and then the ideas came from that consistency.

Yvette: Right, right. Okay. All right. Thank you.

Drew: Okay. Steve Loewy.

Steve: Hello. Dave, have you had some common requests that come where you can just use common agents or resellable pieces, or is it always custom work that you find yourself doing for clients?

Dave: For clients, it's closer to the custom side than the agent side or the low-code side because a lot of times the customers will end up doing it themselves because it is so easy. From a consultant side of things, we work with our clients to actually help them understand what's possible as well.

Steve: So how does that work? Is it—I know sometimes it's difficult with training employees to do this. Are you actually doing training sessions and coaching them? Or how do you coach team members up on AI to really level them up?

Dave: Yeah, so we used to do a little bit of coaching over the summer. We had a consultancy spinoff that we've brought back internally right now. The aspect of that was excitement. So work with the executive team, create excitement around AI and how it could help their company, and then the training and the adoption for the employees from there. So it starts off with the foundational building blocks of what AI is so we can explain and show what's possible without scaring employees.

Drew: Steven, just one quick follow-up. If you are looking for training, we have some great trainers that are friends of ours—that's all they do. So happy to provide an introduction to those if you're looking for it. But it is essential if you really want to build sort of—I hate the term AI-first—but an AI-enabled team. Training is going to be really important. Okay. Joseph Chong, you have a question?

Joseph: Yeah, thanks, Dave and Drew, for walking us through this. I just was interested in a couple of the mechanics because the way you set it up obviously impacts how you can use it. When there's a summary of a meeting or a call, what's the container in HubSpot? Is it attached to the contact or a meeting event or an opportunity, or all the above? Because I'm just wondering what you attach it to and why. And then I have a follow-up question after that.

Dave: It could be all the above. We're currently attaching it to the contact record notes field, and that allows HubSpot AI to read the notes and also do extraction from HubSpot search as well.

Joseph: Thank you. And then you mentioned you use Gemini. I was just wondering, you know, why Gemini as opposed to Zoom's AI Companion that also summarizes, or ChatGPT or Anthropic?

Dave: So Gemini, in my experience at least, is better at getting structured, consistent responses, and that's what we needed for that metadata extraction. We started with Claude. We use ChatGPT also. What I found on both of those, they're more geared towards creative writing than anything else. You can tune them down a little bit, but we didn't want them to make stuff up. We wanted them to just extract what was already said in a transcript.

Joseph: Okay, thank you. 

Drew: Yeah, and it's funny because we use—I mean, I use Claude, and I have a Drew GPT on ChatGPT that's great for creative writing and being able to do it in my voice and so forth. But it's just the facts, ma'am, if that's what we're looking for. It feels like—and there were some budget reasons as well. You know, you already have the license, so again, there was no point in spending extra money to get a license for something that—

Dave: That's a great point because Gemini's free tiers are pretty amazing. Even if you're a Google Cloud customer, you still get a free tier. So all the processing we did for you, Drew, hasn't cost us a single penny.

Drew: Loving that. And that's important in the scheme of things, right? It's how you manage the budgets because we didn't know what the business value of this project was going to be. So it wasn't like we necessarily wanted to spend a lot of money upfront on licenses and other things to do it. We wanted to really just take advantage of things we were already doing but radically streamline it. So I don't know what our estimation on hours saved over the course of a year is. That's exciting to me because I'm already seeing the impact of those hours on really more value-added things, which is the important part of any business, right? You have tedious tasks that suck up employee time, and if you want to get to that 30% growth goal next year, you're not going to do it by just making things more efficient. You're going to do it by taking that efficient time and being able to apply it to things that actually have more business value for your customers. So that's, I think, a good place to sort of wrap up this conversation. So, Dave, where can people find you?

Dave: Yeah, so I'm at Level Agency. Dave dot Brong at level dot agency, or on LinkedIn. I'm the only Dave Brong that I'm aware of.

Drew: If you're a B2B CMO and you want to hear more conversations like this one, find out if you qualify to join our community of sharing, caring, and daring CMOs at cmohuddles.com.

Show Credits

Renegade Marketers Unite is written and directed by Drew Neisser. Hey, that's me! This show is produced by Melissa Caffrey, Laura Parkyn, and Ishar Cuevas. The music is by the amazing Burns Twins and the intro Voice Over is Linda Cornelius. To find the transcripts of all episodes, suggest future guests, or learn more about B2B branding, CMO Huddles, or my CMO coaching service, check out renegade.com. I'm your host, Drew Neisser. And until next time, keep those Renegade thinking caps on and strong!