[00:00:00] Phil: What's up everyone today. We have the pleasure of sitting down with Stephen Stouffer. Stephen Stouffer, director of automation solutions at tray. ai. Stephen, I wanted to start by thanking you for being the first ever returning guests on the humans of martech. We've never had the same person on the show twice. [00:01:14] So welcome back. Thanks for being here. [00:01:17] Stephen: Thank you. It's, it's an [00:01:18] honor and a privilege. [00:01:20] Phil: We had, uh, Stephen on earlier in the year in episode 112, where we unpacked practical wonders of combining AI tools with iPaaS solutions. At the time you were still working in the agency world, but you've since made the move to in house. And it's very fitting to see you land at Trey, an iPaaS tool that has been innovating a bunch in AI. [00:01:41] Uh, we could probably skip like the full career journey. If folks are curious for that, uh, check out part one of our conversation there, but [00:01:48] ​ [00:01:48] Phil: [00:02:00] [00:03:00] we started chatting again when you were in conference planning mode, right? Like I know you're, you're doing the, the speaking tour right now at a couple of different spots. [00:03:37] Uh, you were just at Anticon. So today's episode is going to be kind of a deep dive in customer journey mapping and how AI is really disrupting this use case. [00:03:46] Customer Journey Mapping Essentials --- [00:03:46] Phil: So for the folks who haven't had the pleasure of taking part in traditional customer journey mapping exercises, um, maybe we can start by unpacking like what this entails for you exactly. [00:03:56] And in your opinion, even before we get into like AI, what is customer [00:04:00] journey mapping, Stephen? [00:04:01] Stephen: so if you sit in marketing, it might mean a little bit different things to different people. But generally, like if you go on your website and you use the three, five, seven rule, right? You got three seconds to get someone's attention. You got five seconds to get them engaged and then seven seconds to get them to do something. [00:04:18] So if you pull up your homepage, look at everything above the fold. That's what the first, you know, you got three seconds to, to get engaged. Someone engaged into that content. Is that doing it? Do you have too much content, not enough content? Is it unclear? And then does your button make sense? Is it, is it visible on the, on the page where they can click it within seven seconds? [00:04:37] And if they click that button, does it take them to, you know, where you want them to go? And is that form. Easily accessible. There are too many questions, not enough questions. Are you getting the right information? So that's a very light version of kind of mapping out that customer journey and just making sure that it makes sense. [00:04:55] Um, you know, but if you're something like a campaign manager or demand gen [00:05:00] manager, you might care about the before steps to even getting to that page, right? The, the paid media ad is the ad copy aligning with the landing page that it's pointing to. So. You know, mapping that out, storyboarding it out and talking about the different profiles and ICPs of the folks who are going to be hitting these different pages and the different experiences that you, you want each kind of role to, to, to, to take with your organization. [00:05:24] So very high level. That's, that's kind of what it [00:05:26] looks like. [00:05:27] Phil: Yeah. Yeah. It's, it's fun exercises. Like the first time I did this was, uh, Before the world went remote. So it was in house at a startup and we had a bunch of different cross departmental folks in the room. And it was like stickies on the board. And people were just like, what is the typical journey that someone takes from not knowing us at all to potentially discovering that we exist, landing on our website. [00:05:51] What do they do after that? Do they like go on to download an ebook at the time? Gated ebooks were like the, the, like hot shit back in the day. [00:06:00] And like, after that chatting with sales, if you're B2B, and then like what happens after that? So we had like customer success, people in their sales, obviously, but like marketing, you know. [00:06:09] Rev rev ops wasn't a thing back then, but now like it can be done like virtually also like a huge fan. I know you are too of like virtual whiteboarding tools, the Miro's and the lucid charts of the world. Like you can, you can get pretty complex nowadays. Like if you're trying to, I like how you coined it as like just this one first thing of what is the web experience on that first web visit versus like, what is the customer journey digitally mapping? [00:06:35] Like all of those touch points. It's like, it gets. Pretty crazy, pretty quickly. [00:06:38] Stephen: And if someone wants a mural board to like showcase the different aspects of this and when they want to map it out, like shoot me a DM and LinkedIn, I actually, I have one built, so I'd be, I'd be happy to share that, but yeah, it's, it just depends on where you sit and what you care about, but even email, right? [00:06:55] Like the subject line is that three seconds, you know, is, is, are you getting their attention within three seconds with [00:07:00] that subject line? And then is, is the body of the email skimmable and as the button makes sense. So yeah, it's a really fun exercise to go through it, no matter if you're in the marketing side. [00:07:08] Sales side or post sales on the customer success side is it's [00:07:12] really important to go through. [00:07:14] Phil: Yeah. So how, [00:07:15] AI’s Role in Automating Personalized Emails --- [00:07:15] Phil: how does AI change the way teams approach this mapping today? Like, I don't know if rev ops teams are more marketing ops teams would tackle this, but where's the AI come into this? [00:07:25] Stephen: Yeah Across the board. I mean, um, I'm running into situations where I just ran a demo, um, at Anticon off in London when it came to triggering personalized emails. So when someone fills out a form, my demo, essentially, they just give me their first name, their last name and their email address. And I could create an entire personalized email to that person written by AI. [00:07:50] Um, a good email. So not just chat GPT, write me an email for this person where it's a little generalized and, and, and kind of an AI bot. But when you bring [00:08:00] in an agent into the framework where that agent can use their email and the domain to look up information about the company and then pull their LinkedIn posts and their comments and their job experience, like that email being sent out, um, now suddenly is, is really, really personalized saving marketing. [00:08:18] The headache of creating those variable fields and saving sales. The headache of, you know, creating, um, a whole process of, of research that might take 15, 20 minutes just to craft an email that, that might not actually get opened or the email might not be good. Right. So like that is a real, two really interesting use cases from a marketing and sales perspective of how AI can, can just ease the burden of either purchasing tools on the marketing side, enrichment providers, or sales, just time, uh, [00:08:47] to value. [00:08:48] Phil: Gotcha. [00:08:49] Automating Personalized Outreach with AI Agents --- [00:08:49] Phil: So instead of just using a simple LLM prompt about like, Hey, this is the very basic stuff I know about this person. Let's use an agent, so to speak. I [00:09:00] think like agents, uh, Have like a bit of a bad rep, like folks, like, uh, they, they, they were part of the height cycle really early on and folks kind of like turned down on them. [00:09:09] But when we played around with them on the show and like our, our deep dive on, on AI, uh, tooling, it was still pretty early days, but essentially the way that we were kind of like. Exploring with it is like, instead of like just one prompt and then you have an output, you can give step by step instructions to an agent or even like suggest the goal and then the agent will tell you like what the steps are. [00:09:33] And then there's multiple. Like bots that go in and do the research. They scrape this or they scrape that. They find the URL for this and then they're using an enrichment tool. And so like, is that connected to an iPass tool? Like, is this like where Trey kind of comes into things a little [00:09:49] bit? [00:09:49] Stephen: Yeah, I mean, we built it all within our platform. So, like, that infrastructure is built on the Trey platform, at least internally here and what our customers are doing. But it's exactly what you said, right? So, [00:10:00] when you think about an agent, it's just using the best tool for the job. So, you have access to OpenAI, Anthropic, Google API, LinkedIn API, all of that. [00:10:09] So, that gets plugged into the Trey platform. And then Based off of the instructions, like you mentioned in the context that you give it in the role that you give it, it's like, Oh, this is a marketing email. I have their email address. I'm supposed to be crafted an email on behalf of this, this person. Then let me go look up relevant content. [00:10:28] Like ours does like press releases. It looks up crunch based data. It looks up blog posts by the company, and then it does like LinkedIn profile and, and blogs written by that person, um, to personalize the email. So the email would be something like if it was being written to me, it would be something like, Hey, congratulations on your speaking slot at anti con in London. [00:10:48] Hope you had safe travels. Like. W getting an email like that, that's completely automated and, and written by a, it blows your mind a little bit. Um, like it would take [00:11:00] probably a BDR or um, an ISR to actually write that, probably like 15 minutes to do all that research. So it's, it's really cool once you can kind of get your hands on like a tangible use case versus like, you know, some of them that are out there that are, you are like, okay, that sounds great, but like. [00:11:18] What's the realisticness of me being able to like implement this. So yeah, if someone wants to try it out, I got a, I got a demo. I could, I could share about that, that, that whole personalized email. Um, so yeah, it was, it was fun to, to show that and, and just kind of see people, the minds be blown in the audience a little [00:11:35] bit. [00:11:35] Yeah. [00:11:40] Phil: emails. [00:11:40] Challenges in Implementing AI-Driven Customer Journey Mapping --- [00:11:40] Phil: Now what's like, what do you think are some of the challenges that, that, that teams are facing when they could be implementing, like, I don't know what we call this, like AI driven customer journey or personalizing emails. Like, are you, are you focused on the initial prompts? [00:11:56] Like, is it always going to have, like the human is still figuring out [00:12:00] what the goal is. The example you just walked us through is. More of like the outbound use case, right? Like maybe someone doesn't know about the company or the product that we're selling. And that outbound email is personalized by AI. [00:12:13] Instead of spending 16 minutes writing a good outbound email, that's personalized. AI is hooked up to a bunch of enrichment tools to be able to do that stuff. Like, what are some of the main challenges aside from mapping out? Like, what are the different goals? Where can we use this in the journey? Cause it's not just like outbound based emails. [00:12:32] Right. [00:12:32] Stephen: Yeah. I I'd say a few things. Uh, one, just the technology's ambiguous. Like, I think it's difficult even for. Fairly technical people to wrap their head around how to how to build these types of tools within their organization. Um, to connecting the data to make the relevant prompts, right? Like using an agent architecture, you have to use and pull data from all your different kind of systems, your CRM, you know, are they a customer? [00:12:59] What kind of products do [00:13:00] they have? Which product would make sense for us to recommend based on their existing usage or, or skews that they have? Um, so connecting the data is, is its own thing. It's like everything getting up to writing the prompt, right? I think anyone can write a prompt, but, but feeding it the context for that prompt is, is the tricky bit. [00:13:19] And then I'd say like the third big thing would be just security, like worried about. What is it that we're giving the LLM model? Do I have to be SOC 2 compliant, HIPAA compliant, GDPR compliant? Like a lot of times people don't even really understand the rules and regulations and they don't even want to try building something because they're worried about, you know, legal, the legal implications of if they don't do it right. [00:13:45] So like those are the three big things, right? So just wrapping your head around what the heck it is. Getting tapped into the data to kind of start writing that prompt and then, you know, compliance and regulations and the [00:13:57] headaches that follow [00:13:59] Getting Started with AI-Powered Outbound Personalization --- [00:13:59] Phil: [00:14:00] What's a good starting point for someone who's like trying to figure out, let's, let's take the use case of, of outbound there. Like I want to equip my SDR team with Trey plus like a couple of agents and enrichments that makes it way easier for them to do. Personalized outbound email as like, is the marketing ops or the rev ops team, like building that, you know, first initial use case for this, like, what's the first step to mapping this out? [00:14:28] Like how, how do people map out the contexts, all the data required to feed into that tool? Do you start with like a whiteboarding session? Is it like a spreadsheet? You're making a list of like everything that's required to feed, like what, what's the first step to doing that? [00:14:44] Stephen: So just from an implementation perspective, I'd say Gather the requirements and then start simple. Like don't try to build a big, massive agent. Maybe just start with tapping into LinkedIn or just Google, or even just starting with open AI, right? Like the email might [00:15:00] not be super personalized, but like getting to that step is, is, is good, but follow me on LinkedIn DME, I'm happy to do a coffee chats, I do kind of like non salesy coffee chats, uh, and then if you want. [00:15:13] From a trade perspective, just the way we think about it. We, we host free AI agent workshops like every single week. So we'll give you access to our platform for free. You can pop the hood, play around with it, and at least start to wrap your head around the different use cases and actually touch the technology a little bit. [00:15:30] So whether or not you want to become a trade customer, it doesn't matter. It, the, the, the [00:15:34] workshops are for everybody. [00:15:35] Phil: Okay, cool. We'll, uh, we'll make sure to link out to, to that stuff for sure. Um, if like, I'm trying to think of other use cases for this, like we're, we're harping on, on outbound here and maybe some of the folks are just like allergic to outbound, but let's take like lifecycle, for example, like a lifecycle comes in. [00:15:51] To a lot of journey, customer journey, mapping exercises, right? Cause like you don't stop driving revenue after that first acquisition journey [00:16:00] is done. There's a whole like monetization activation, getting people to create those like habit loops, whatever, right? Like that's where your life cycle kind of comes in. [00:16:09] Leveraging AI for Lifecycle Personalization --- [00:16:09] Phil: And let's say that. We're trying to use this same type of AI driven insights for life cycle marketing. Um, the thing that I struggle with, like bringing this from theory to actual hands on keyboard, like we're delivering something is that in life cycle, we're often building this like library of content and we have content that's specific by persona content that's specific by like. [00:16:34] Stage of the journey, where they are in like building something. So how do we feed the content that exists existing today and connect that to AI so that we can obviously use that to personalize the outreach, but like, how does AI know what's going on? Which is the best content to send to that person based on where they are in that life cycle journey. [00:16:57] And is AI changing that content a lot? Like, how [00:17:00] are you prompting that? Like just, I know there's a bunch of questions in there. [00:17:03] Stephen: So I don't know what everyone else is doing, but I'll tell you what I'm doing. So you can de anonymize your web traffic using something like Clearbit reveal or, you know, a handful of other enrichment providers. If you could de anonymize that and link it back to a, a person, a contact, a lead within Salesforce, um, or even using the cookies through, you know, Pardot, Marketo, HubSpot, right there, there, there's all sorts of de anonymizing things you can tap into, assuming you have one of those platforms. [00:17:32] What you can do is you can call out to your CRM from that, having their, their domain at minimum, or maybe even the person themselves. Uh, and you can look at their contact. You can see if there's a, uh, what, what account they're tied to. What open opportunities is the opportunity closed one. If it is closed one, what kind of products and services do they have? [00:17:50] And then you can bring that information back to your website and then you can start to influence what shows up. So, uh, when you think about. Like part, [00:18:00] uh, uh, dynamic content. Like this is like dynamic content on steroids. This is an entire. Custom personalized landing page from your homepage to anything else you want to influence based off of information that you have on that person. [00:18:13] So, you know, life cycle management, let's say they are a customer and they have, you know, your baseline product, your homepage could push them to the next tier up, right? Um, or instead of saying, get a demo on the top right of your, your screen, their customer, it could be reach out to customer support. And to, or, or link it to your account managers, Calendly link, right? [00:18:35] So you're removing those friction points along the customer journey by being able to be intelligent on what you offer up. So, you know, partially getting connected into that data, using something like tray to push and pull the data between your systems and then using AI to be more intelligent on what it is that [00:18:50] you're offering up. [00:18:52] Phil: cool. Yeah. The, the life cycle use cases is fun here because. You don't have to de anonymize a ton of stuff sometimes, like from web visits, [00:19:00] like I get it, but if you're using email as, as the use case, for example, you have that person's email already. So you don't have to skirt privacy regulations with those de anonymization tools, but the, so how do you come up with the. [00:19:12] The content to surface though. Cause I feel like that's part of my question is like right now a human is deciding we're going to send out an email on day three after someone converts to one and that 30 email does really good at. Telling someone, what are the other features in the product? And maybe you have a bit of logic in there because it's, it's listening it with, it's connected to the behavioral analytics tool. [00:19:38] And you can see if they've done X, Y, or Zed. So the email on day three is not going to be about X or Y if they've already done those things. And it's really focusing on Z. Like those are rule based like journey orchestration, [00:19:52] Stephen: to be. [00:19:53] Phil: Right now. Yeah. So that's like bleeding into my question. [00:19:55] Smarter Content Choices with AI --- [00:19:55] Phil: Like how, how do you give that information from the context [00:20:00] of the content that you're currently sending out to folks to your AI system so that it does like a better job than like the rule based stuff that folks are doing today. [00:20:09] Stephen: Yeah, I, I feed it a document. So I have a document, it's one place centralized location where sales and marketing can both influence it. And the document essentially outlines the different products, outlines the messaging for each product, and then who, who the best ICP would be for, for that product. Right. [00:20:28] So, uh, using our, our kind of expansion or our, our cross sell upsell kind of as a, as a, uh, use case. I can say, Oh, Hey, this is our mid level tier. The person who, who this is going to be best fit is who's on our base tier. Right. And I just have that in a document. I feed it to AI. So AI has the context of all the different options, how we talk about those different options and who it would be good for. [00:20:52] And then AI determines the best one for the job. So like. The old world would be, we would have this like mind [00:21:00] map if else statements where if it's this stage or if it's this, if it's this, no, you just feed the data of who that person is, where they are in the organization, what their stage is also feed it the document. [00:21:12] And then it picks the best tool for the job or the best product or service offering for the job, and then levels that up in the email level up on the landing page. So you're, you're not kind of gone are the days where you have to kind of map all this out on a whiteboard [00:21:25] with sticky notes. [00:21:27] Using Historical Data to Improve AI-Driven Messaging --- [00:21:27] Phil: Is the next phase of that also feeding it historical data on what's worked and what hasn't worked in the past? [00:21:34] Stephen: Yeah, that's, that's one piece I haven't done yet, um, which I'm really excited to do. So, so, One thing that I'm thinking about right now as a, as a use case is actually our outreach sequences. So like looking at the data, the open rate, the click through rates, and then telling that to AI, giving that to AI before it writes the email. [00:21:52] Where it can actually go look at specific emails and say, okay, this one email from this rep is performing really, really good. And then replicating that [00:22:00] success. So I haven't dipped my toes into that, but you're a hundred percent right. Like that is where we're going. Is. Feeding the context of the different offerings that you have. [00:22:09] Feeding the information about the person, the company, the opportunities, the products, um, and then also looking at historical performance. So like from a landing page perspective, Google analytics data, right? Which pages are performing the best? What are those layouts? What are those offerings? What color are the buttons? [00:22:25] And then [00:22:25] replicating it. [00:22:27] Phil: and I think like even taking this a step further, like I think the step after that is like, obviously the historical data helps with the potential for something to resonate and perform better, but then you can also use AI to have real time. Incrementality and experimentation within everything that you're doing. [00:22:47] So you're always splitting out cohorts and having a control group. And it's consistently learning on, you know, even though that I'm looking at the document and Stephen and the sales team says that this is the right [00:23:00] message for the, this perfect expansion target. And you're sending that out. But then AI is sending out a different version of that to someone else and all the different versions all have control groups and learning the lift that certain message has. [00:23:14] And eventually, like, you have a big volume of stuff and there's like, incremental dollars. Associated to messages and you always know what's more optimal. So I feel like for some folks listening, it's just like, this is futuristic shit, but we're actually really close to this. And I chatted with some folks last year that, um, more on the B2C world, like had some of this stuff built in already, and especially like enterprise teams. [00:23:38] Like we had Paul Wilson on the show, like former Slack Salesforce and myself too, like I had a short stint at wordpress. com, the enterprise team. The company and like, there was a ton of data scientists that were working on this shit, right? Like, [00:23:54] Making AI Personalization Easy for Every Marketing Team --- [00:23:54] Phil: I think that you're an incredibly technical marketer thinking like way differently than most [00:24:00] technical marketers. [00:24:00] There's a lot of data scientists that, you know, have been doing this for 20, 30 years, right? Like machine learning shit. Is very simple to them. And at WordPress, we had like an internal customer data platform and every event that we tracked, our data science team built a UI for all the marketers to use, and we could pick any of those events that were tracked and create a propensity model on that event to see the likelihood that someone would convert or would. [00:24:27] Churn or would convert from paid or from free to paid. And so every time it came to do a campaign, we could do it manually and say, like, instead of saying, I want to send this email out to the last 30 people that started a free trial. We could say, I want to send this email out to. The last people that started around the last 30 days that are more likely to convert to paid if I give them a 30 percent discount. [00:24:51] And so there's like all the uplift modeling that goes into that. So I don't know. I feel like this whole world of journey orchestration and using propensity models, like [00:25:00] even takes this a step further in AI. Personalizing and like adding experimentation on top of all the messages. Like shit's getting wild. [00:25:08] Stephen: It is. And the job of, uh, me and like Trey and our organization is to, to make it where someone can do this easily and they can wrap their heads around. Cause that's the problem right now is like, not everyone has a whole bunch of data engineers, uh, in their organization to, to, to deploy this. And, and if you want to keep up with your competitors, you're going to have to do it. [00:25:32] But like. You know, everyone's trying to do more with less right now. Budgets are, budgets are tight. So like the job of me and like our, my team here is to make this easy, accessible and, and, and something that everybody can do. So, so like, that's the challenge right now. It's like, yes, we can do all these amazing things today if you know how to do it. [00:25:51] Uh, but for most people it's just out of reach. Um, so we got to fix that. [00:25:56] Phil: Yeah. Yeah. For all startups or any company that doesn't have like a massive [00:26:00] data science team, like it is out of reach and it's not as simple as saying like, Just one tool or like one data engineer is going to fix this for you. Like, I feel like there's, and maybe one of the thesis or like missions of the show is bridging that gap between the marketers and the data teams, that even if there isn't a big data team, like there is someone oftentimes supporting the whole organization and not just marketing, [00:26:23] ​ [00:26:23] Phil: [00:27:00] [00:28:00] but I guess like something that I've, I've gone down with a couple of recent guests is this idea of data literacy for marketers, like understanding. [00:28:21] Agents and AI and I pass, like, that's like one part of the story. [00:28:25] Building Data Literacy for the Next Generation of Marketers --- [00:28:25] Phil: The other part is. The data foundations. And I feel like a lot of marketers were robbed of some of this stuff, like the statistical, like foundational stuff. Um, what advice would you have for marketers that are listening? They're just like, I want to be Stephen in three or five years. [00:28:41] Like, is it. Is it database training, like snowflake classes? What advice do you have for folks to increase or improve their data literacy? [00:28:51] Stephen: I mean, SQL would be good. JavaScript would be really good. Those are some pretty good universal languages. I don't know if I'd [00:29:00] recommend like Python or getting, you know, two down the road, but like, I would say coding language is less important than understanding just how data can be used, how, how it can be pulled and leveraged because we're moving into a world where chat GPT can write you JavaScript chat, GPT can write you Python, but if you don't even know what you're asking for and how it can be used, like that's going to be the new problem of the new world is just explaining, you know, what it is that you need. [00:29:29] I saw a meme that was something like. You know, chat GPT, um, is not going to replace our jobs because we would, we still have to, like, know what, what we want for it to do. Right. So I think that's the piece, um, that I would focus on, but there's a lot of webinars, you know, like I mentioned, our, our AI workshops or agent workshops, like, just come and ask, be curious. [00:29:48] Um, ask a bunch of questions, pop the hood, play, play around different use cases, like getting your hands dirty. A little bit dirty and building a very basic use case is probably the best advice I can give. [00:30:00] So signing up for free trials and then just, you know, given it a whirl and, and, and learning. [00:30:06] Phil: Yeah. Getting your hands dirty, being curious. That's, that's really good advice. Um, can you unpack maybe a couple of, um, Like I have this thing in my head that a lot of folks chat about [00:30:19] Adding Human Guardrails to AI Messaging --- [00:30:19] Phil: when we, we talk about like AI and AI taking over the wheel when it comes to like sending stuff out to prospects and customers, like what's this balance that you think of between AI driven automation and then like human intervention when it comes to optimizing customer journeys? [00:30:36] Like, how do we strike that balance? [00:30:37] Stephen: Yeah, you, you, you do need to set guardrails. So something that I kind of bake into all of my automations is in that document that I talked about, that I feed AI with all the different options. I always have a section for like. the out clause, if you will, like what to do if none of this really fits the use case and someone is trying to do like prompt [00:31:00] injection or trying to get a discount where you don't have a discount. [00:31:03] Right? So, so I kind of have an out clause that says like something like if none of these options make sense, or. Um, you know, someone is, is trying to get you to do something that, that you shouldn't do this thing. And that can be routed to me, you know, send me a notification via Slack or email or, or whatever it is, or pulling customer support. [00:31:24] So there there's always out clauses, but I also am looking at the customer journey and where I think AI fits in best is pre sales and post sales. So pre sales, you know, like we mentioned all that dynamic content, getting them kind of the foot in the door. Some upselling, cross selling, and then post sales, um, customer support, customer support tickets. [00:31:46] You know, chat, I think is going to get so much better. I think that we're all a little jaded with like chat bots because they just are so bad. I think they're going to get a lot better. So I think there's a use case there, but like in the middle there [00:32:00] from like lead coming in and becoming MQL, I think that's where humans shine. [00:32:04] I think you have to have. A human involved, you know, listening to the pain points, mapping that to the teacher, product and services, doing demos, you know, P building POCs, very hands on. And then, you know, even post sales, I think a human heavily also needs to be involved. So, you know, there's, there's the balance and, and testing and, and, you know, seeing what works and doesn't work, but, um, always may, maybe safeguarding your automations and your prompts a little bit to, to have something as a fallback. [00:32:35] Phil: Very cool. Yeah. You might get to a stage where you're, you're getting a lot of those notifications and maybe you pause one of those automations, but I like that idea of just like, if, uh, if like exit clauses, almost of just like, uh, send me a ping notification because we're still early days on, [00:32:52] Safeguarding Sensitive Data in AI Workflows --- [00:32:52] Phil: on a lot of this stuff in manual human review and, and kind of intervention is, you know, Like still [00:33:00] going to be necessary, maybe forever for industries that are highly regulated. [00:33:04] Like think of like the health tech [00:33:06] sector, the financial sector, where just this idea of letting AI decide the messaging, especially if you're bringing in like confidential information and PII inside of messaging, like [00:33:20] she [00:33:20] Stephen: And I wouldn't [00:33:21] like you got to be very careful with that. So like something that we have, we have Merlin guardian in like the tray platform that tokenizes that sensitive information before sending it to something like an LLM. So like, things like that, like, like, Those are like the foundational points that I think people need to wrap their head around. [00:33:39] Um, so yeah, attend one of our workshops. I think it'd be, it'd be very helpful to folks to kind of. know, see tangible use cases of like how they could do this in a meaningful way where they're, you know, you're HIPAA compliant, SOC two compliant, GDPR compliant. Like there's a, there's a [00:33:53] lot of things you have to worry about. [00:33:55] Phil: yeah, yeah, [00:33:56] Use Case Example: Automating LinkedIn Lead Gen Forms --- [00:33:56] Phil: maybe walk us through like a, give us a teaser of like those, those [00:34:00] use cases, like maybe one like AI power journey mapping that had measurable impact on revenue growth. [00:34:07] Stephen: Sure. Yeah. So, um, this was actually at my previous company, we actually had the trade platform. Um, and I had a customer that did LinkedIn Legion forms. So they, they had like, you know, the, the, the. Uh, LinkedIn ads that popped up and then you put in your information and then it, um, got sent to the tray platform. [00:34:28] So, so one issue was before we built that it was export import, right? So they exported the CSV once a week, um, sometimes like every other week. So it was taking seven to 14 days to get that information into Salesforce and then to sales. So if you're in marketing or sales, you know, like time matters, right? [00:34:47] So like, From when you submitted a form to exporting and importing that into Salesforce, doing all the cleaning of the data and then importing it. Like 14 days later, you might be like, who are you? Like, w what did I sign up [00:35:00] for? Right. So we, we automated all of that. Um, and the reason why it was so manuals, because for those of you LinkedIn Legion forms, the state and country values are open text fields. [00:35:12] They're not pick lists. So someone. You know, can type in whatever they want. It can be outside the region that you support. So you have to clean all of that up. So we actually just had AI cleaned it up. We said, Hey, here are all the values that we accept. And for those that don't fit this value, your exit clause is like, just put an unknown, right? [00:35:30] So, so you have that fallback value. Um, and then we turned around a 14 day process and it took 20 milliseconds essentially from, from lead submitted to getting it into Salesforce. So, uh, We saw an uptick. Oof, you're, you're stretching my memory here. Um, I think we saw an uptick to like 40 to 50 percent of conversion rates from, uh, MQL to, uh, sales qualified in a, in a stage one opportunity being created. [00:35:57] So like, you know, one simple automation [00:36:00] with three steps in it. Intake, clean it up with AI, send it to Salesforce, create the lead or contact. Had a 50, 60 percent conversion rate impact. So like, you know, you don't have to start complicated. It could be just a simple, you know, bleed intake from, you know, LinkedIn being [00:36:17] your use case. [00:36:19] Phil: Yeah. The impact on conversion rate, also the impact on improving the life of that revenue operations or marketing ops person that was doing the CSV imports. I feel like a lot of folks listening to [00:36:30] Stephen: Yeah, well, and the customer and then like the potential customer, like, can you imagine the frustration of just like wanting to sign up for a demo and then not getting it for 14, like, I think we've all had those situations, right? So, um, yeah, just across the board, it just makes everyone's life. And then from a, a data integrity perspective, like it's, it's cleaning the data into your standardized formats of, you know, state codes or full country or, or state names, like just to be able to do that on the fly. [00:36:56] It's, [00:36:57] it makes a huge impact. [00:36:59] Phil: Okay. Is there [00:37:00] another one that comes to mind? [00:37:02] Stephen: Oh, that's a good one. Uh, I've seen, this is more recent, but I've seen an uptick in conversion rates on emails using personalization. So I mentioned like the demo I did in, um, anti con, like I did that during my presentation and I've had a flood of people who have responded to that. Uh, so like even just me personally, uh, like I've seen a big uptick, uh, in, in just conversion rates of emails using AI. [00:37:30] So like, Um, that's a big one that I've, I've run into today and it was like a conversion rate of, um, like it was like 5 percent and went from 5 percent to like 15 percent of, of email, like people responding back to emails and, and, um, like setting up a coffee chat with me. So, uh, it's been exciting to see, but my time on my calendar has become a little less available. [00:37:58] So it's a blessing and [00:37:59] a curse, I [00:38:00] guess. [00:38:01] Find Time for Skill Building --- [00:38:01] Phil: What do you think is good advice for like RevOps, marketing ops teams who are trying to future proof their tech stack to stay ahead of all of these trends and changes? Like I'm putting myself in the shoes of the audience and I've got like a. Mile long list of shit on my backlog. And as much as their stuff is exciting and I want to get my hands dirty. [00:38:23] And like you said, like, be curious and attend those workshops. Like I've got a day job, Stephen. And like a bunch of shit is like drowning me in work. Like how, what's your tips for, or like your, your ideal advice for, for companies trying to like future proof their tech stack and staying ahead of stuff. [00:38:39] Stephen: Yeah. Well, subscribe to Humans of Martek. Uh, stay up to date on all the trends. There you go, your free little, uh, little, uh, yeah, there you go. Uh, I would say, like, be mindful of where you spend your time during the day. And just make note of it, right? Like, are you spending, like, Three hours a day going through emails. [00:38:59] Like what are [00:39:00] those manual tasks that you're doing? And just try to automate to make your life easier, because if you can do that and free yourself up, then you can spend, you know, another 45 minutes to an hour doing research and attending workshops and going to webinars. Uh, but more importantly, you're going to be finding like the tangible things to make your life easier today for yourself and your organization while also learning. [00:39:21] So just try to automate and, um, optimize. what you're doing, where you are, and don't try to go, um, you know, across the organization or in other departments, like, you know, solve problems that you have today. Um, and then you're gonna learn, and then once you learn, you're gonna kind of expand your knowledge a little bit, and then you can kind of, your problems are gonna shift, and they're gonna be a little bit different, and then you get to solve [00:39:44] those. [00:39:45] Phil: Very cool. I feel like you touched on this one already a little bit, but [00:39:49] Getting Started with AI in Marketing Operations --- [00:39:49] Phil: what advice would you give to marketing ops folks who are like really early in that? Let's explore AI journey. Like they're listening to this. And they're just [00:40:00] like, guys, I can't even get my team to use Chad, GPT, like, let alone thinking about how to implement AI and like our customer journey mapping exercises. [00:40:10] Like, what are the first steps to like convincing senior management folks to care about this? Like a lot of folks are doing 2025 planning by the time this episode is going to be dropping. Like, what are some of the first steps that people should be doing to explore like the early use cases? [00:40:28] Stephen: Yeah, that's, that's tough. Um, I know that, In our organization, we have a dedicated champion. Um, another thing that I've seen some of our customers, so it's like, you know, you need to have at least one person, right? One person that you kind of task with pushing the bounds while everyone else is busy, right? [00:40:45] And then what happens is another person in another team realizes what that person is doing. They kind of want to do a little bit more, and then you kind of organically are attracting the right people that should be working on it in your organization. Another thing that I've seen that is really interesting [00:41:00] that some of our customers here at Trey are doing is they're doing hackathons. [00:41:03] They're doing AI hackathons, right, where they're dedicating a single day, where it could be just one team or one department. Or across your entire organization, where you have different department teams or cross collaboration, mix up of different people from different teams, and it's just for fun. It's like, Hey, here's open AI. [00:41:19] Here's anthropic. Here's the different tools that we have. Let's just build something just for fun internally to see what we can do and see if we can measure the impact. And then you just kind of let everyone go wild with maybe some guardrails of like what the problems that they should be solving. But we've seen a lot of fun and excitement been. [00:41:37] From executives. Once they can see how it can apply to their organization, that they're much more likely to put budget and adopt it. And then if you're a bigger company, you know, board members and whatnot, having them see kind of like the way you're thinking about future proofing the business and stuff, it can make a measurable impact all the way up from, from [00:41:57] top to bottom. [00:41:59] Phil: Very cool. Great [00:42:00] advice. Um, yeah, I jumped with a couple of last ones here. Uh, this has been super interesting. What, uh, [00:42:06] A Shift in How We Approach Paid Media and SEO --- [00:42:06] Phil: what advice do you have for folks that like, maybe you're looking at all the different disciplines under. Marketing operations. Maybe we'll keep it to marketing operations and just like marketing. [00:42:17] Um, like what are some of the areas that you think are going to become less important in five, 10 years because of these advancements that we chatted about today and what are some of the disciplines or specializations that you, you think folks listening, like, Hey. Pause work on that stuff. Like you said, like, look at what you're, where you're spending your time today. [00:42:38] Like pause work there and like double down in this area over here. Like you kind of talked about like data literacy there, but, um, what, is there anything that comes to mind? Like this is less exciting. This is an area that's way more exciting for you. [00:42:53] Stephen: Yeah. You're. I don't know if I'm going to be making any friends here now. Uh, I, [00:43:00] I would say that there's, there, there is within paid media, PPC, um, you know, building search search queries and stuff like kind of hacking the internet. So your pages kind of come up first. I honestly think that's all going to change the way the internet works is changing. [00:43:19] You go to bing. com. It looked a lot different today than it, than it did. Three years ago, four years ago. So, so from a marketing perspective, I think worry less about filling keywords within landing pages because a lot of that stuff is, I think, just going to be dynamic and, and then the way the algorithms work today and, and going in the future, it's going to change. [00:43:42] So like. We kind of need to stop the idea of trying to hack our way to the first page and think in terms of just building the best content and solving customer pain points. And then, uh, before you know it, the lead sources that are going to be coming in are going to be from a I sources that are not going to be Google and being it's [00:44:00] going to be chachi BT and thropic, right? [00:44:02] So, um, yeah, Don't try to game the system and just do what's best for your customers and just, and, and make it as easy as possible for them to do business with you. And then I think you're going to be rewarded in like the future that's being [00:44:13] built right now. [00:44:15] Phil: Do you, do you still heavily use Google search as much as you did like five years ago? [00:44:20] Stephen: I don't [00:44:21] know. [00:44:23] Phil: Yeah, I, uh, I'm a heavy perplexity, uh, user, especially for like, so like I'm doing the podcast full time now, but I still keep my hands dirty with a couple of, uh, clients on the side, like doing a bit of consulting. And, um, I find that, like, I hope they're not listening to this, this stage late, late stage in the conversation, but I utilize perplexity all the time. [00:44:45] So like one of the services I offer, like they're connected to my Slack and. They can ask me questions asynchronously throughout the day. And I answer a sync. I will oftentimes just copy paste that question and perplexity and see if there's like [00:45:00] a quick, fast answer already perplexity is way better than I find so far, like then chat, JPT at connecting like all the docs. [00:45:08] With some of those tools, like a lot of it is like questions. Um, and so like it surfaces, the answer is by looking at like developer docs on those tools and finds like other community sourced articles. But yeah, huge fan of perplexity. It's like really changed the game on like how I find answers online. [00:45:25] Like I don't. Don't use Google search that often anymore. [00:45:29] Stephen: I, I, I've been using chat. GPT now for a long time from. Co development answering questions to coming up with like LinkedIn post ideas. Like it's, it's wild how much my behavior has changed. And like the last two years, two and a half years, when compared to like the last decade of being in operations, like it's, it's, it's, it's wild to think how much like the day to day work of just how we [00:45:59] operate has [00:46:00] been changed. [00:46:01] Phil: Yeah, I, I need to play around with, uh, with clode more, uh, and anthropics tool, like, uh, everyone talks about it and I only played around with a free version and I was comparing that to like the, the, the, the pro version or a Chajupiti 4. 0 and Chajupiti 4. 0 is still like miles ahead of clode. So I need to like. [00:46:19] Uh, pay for a couple of months and just like put it to the test, like the, the, the pro version of it. [00:46:24] Generate Content Ideas from your LinkedIn Profile with AI --- [00:46:24] Phil: But like when you said like, you're using it for LinkedIn post ideas, like that's anthropic, not JGPT or still JGPT. [00:46:32] Stephen: A little bit of both. Um, I actually feed it through, through, through both, but what I, what I've done is actually exported my entire profile from LinkedIn. You can request your data if you request it and then upload it into like a knowledge base for chat, CPT, or would your AI tool of choice, uh, it can recommend. [00:46:52] Like topics a lot more personalized than you having to kind of like hack it. So I don't know if you've done that before, but it's pretty, it's pretty interesting. Yeah. Request your [00:47:00] data from LinkedIn, give it to an AI model, and you've essentially like replicated like your tone of voice topics and industries that you care about. [00:47:07] And then it can recommend, uh, it's pretty good. It will recommend some [00:47:11] pretty good topics and stuff. [00:47:13] Phil: very cool. So like the export is like a, a JSON file format that you [00:47:17] Stephen: Yep. Yep. It'll, it'll give you, uh, like connections that you have. So like you have kind of who your ICP is, right? Your followers. And then you also have like all of your posts, your comments, like just a history of kind of how you've interacted with LinkedIn and given in that as context is, is [00:47:33] pretty interesting. [00:47:35] Phil: I appreciate that tip. Um, well, yeah, [00:47:38] Embracing AI and Building Use Cases --- [00:47:38] Phil: any parting thoughts, like any other tips or things that are top of mind right now that you think folks that are still listening clearly are like excited about AI stuff and, and marketing, anything else you want to drop? [00:47:48] Stephen: Yeah, like just get out there, uh, build out some fun use cases. And if you're, if you're not really sure where to start connect with me, I'd be happy to do a coffee chat. Um, show you kind of like what I've been up to and what I've been [00:48:00] working on. And if you're a Trey customer, I have templates that I can export and import into your org. [00:48:04] So I can just give you what I've been working on. Um, but yeah, it's, it's fun times to be, uh, working with AI. Hey. It's the future, so the quicker that you can kind of wrap your head around it and start bringing value to an organization is, you know, you're going to be future proofing your role within the organization, which I think all of us are kind of feeling a little bit of angst. [00:48:26] So, uh, you know, figure out AI, at least a couple base use cases, and [00:48:31] go start from there. [00:48:33] Phil: Awesome. We'll, uh, we'll make sure to link out to the workshops that you do at Trey. Um, I'm sure there's a ton of other. Yeah. Useful resources and content on a site, but yeah, we'll link that to your LinkedIn. I think you're a super fun follow on top of just like the, the, the AI stuff. I think that, uh, there's a lot of comedy in the comic relief, I guess, from the marketing ops stuff. [00:48:53] So [00:48:53] Stephen: Yeah, I'm trying to not take myself a little bit too seriously. But thanks for having me again. It's an [00:49:00] honor again to be here [00:49:02] and chat with you again. [00:49:05] Phil: Yeah, man. Appreciate your time. We'll chat again soon. [00:49:07] Stephen: Alright, thanks.