← MarTech Podcast ™ // Marketing + Technology = Business Growth
MarTech Podcast ™ // Marketing + Technology = Business Growth
AI Agents for marketing automation
00:33:01
Loading summary
Transcript52 lines
- [00:00]
Benjamin Shapiro
The Martech Podcast is a proud member of the I Hear Everything Podcast Network. Looking to launch or scale your podcast, I Hear Everything delivers podcast production, growth and monetization solutions that transform your words into profit. Ready to give your brand a voice? Then visit iheareverything.com.
- [00:26]
From advertising to software as a service to data, across all of our programs and classes, we've seen a 55 to 65% open rate.
- [00:36]
Nicholas Holland
Getting brands authentically integrated into content performs better than TV advertising.
- [00:42]
Benjamin Shapiro
Typical lifespan of an article is about 24 to 36 hours. If we're reaching out to the right person with the right message and a clear call to action, then it's just a matter of timing.
- [00:54]
Welcome to the Martech Podcast, a member of the I Hear Everything Podcast Network. In this podcast, you'll hear the stories of world class marketers that used technology to drive business results and achieve career success. Here's the host of the Martech podcast, Benjamin Shapiro.
- [01:15]
15% by 2028, Gartner predicts that 15% of all work decisions will be made autonomously by AI agents. Not human in the loop AI assistants, but decision making agents that take actions without human interaction. We're already making the fundamental shift from AI that helps to AI that does. When will you be managing AI agents instead of human and what does that mean for your marketing team? I'm Benjamin Shapiro and joining me today is Nicholas Holland, the head of AI at HubSpot. And today Nicholas is going to explain how the shift from agentic AI is fundamentally changing marketing automation and what it means for your team's future. Nicholas, welcome to the Martech podcast.
- [01:57]
Nicholas Holland
Hey, thanks for having me Ben. Appreciate it.
- [01:59]
Benjamin Shapiro
Excited to have you here. Excited to talk to you specifically about this brand new world that we're facing. I feel like AI just got here and already we're making this fundamental shift from hey, I can use AI and it makes me smarter to it's just doing stuff for me. We're hearing that 2025 marks the shift from generative AI to agentic AI. What does that actually mean for marketing teams?
- [02:24]
Nicholas Holland
I think the easiest way to kind of start off is defining an agent. At least the way that we look at it here at HubSpot and for like a marketer, the way to think about an agent is something that thinks, decides and acts to do some tasks on your behalf. This could be a little foreign for people who don't do a lot with AI, but for those of us who use it more frequently, you'll start to see now, like when you're working with different things, how it says thinking and then it begins to do something. Agents are really unlocked with that kind of concept. And so you now have a world where teams are starting to evolve. It's early days now, but it'll be very common in the future where you'll see this like hybrid team. You'll see a marketing team that has humans on it and it also decides that it wants to add some agents to the team. What that's going to mean for a lot of marketers is that you'll start to see roles shifting now, which is interesting. You already see that now with content agents, where you might have now a marketer that couldn't do graphic design, couldn't do video editing, couldn't do content writing. Now that marketer doesn't need those other roles now. They have all those capabilities themselves. So we kind of think of this as the rise of what we call the super contributor, where one marketer now might have multiple agents that work for them and now they can do a lot more work. And it's going to be really interesting now because that'll change how teams look, that'll change how marketers behave. It'll change the volume of work that can get done. It's going to be pretty cool.
- [03:43]
Benjamin Shapiro
Super contributor is an interesting way to put it. I think it's a friendly way to say one person can do so much more by building out agents that are doing some of the reasoning and decision making for them. But nested in there is. I don't need someone to come in and evaluate which guests I should have on my podcast because I can have an agent do it. I can have an agent do the research on you and your background so I know what questions to ask you. These were all tasks that I had my team do. We use freelancers and some of them we don't need anymore. And I'm assuming this will continue to evolve and evolve and evolve, which means less work for humans, doesn't it? I mean, how far do we go here?
- [04:22]
Nicholas Holland
Let's take where you are. You just said that you might have it evaluate people on your show. It just depends on a lot of times where the maturity of teams are. Organizations are what we see. Like HubSpot is the champion for kind of small businesses. And when we think about a small business, often what we see are marketing teams who just have more work than they can handle. Overtax. They often have to effectively do the generalist chunky middle. And then to your point, they might hire specialists whenever they have that need but they often have way more work than they can handle. So what this often meant in the past is that for, for even like writing a piece of content, it wasn't as deeply researched as it should be. It didn't go down the path of picking kind of a unique perspective. It was like you had to get it out the door. Because early days of marketing, like in the last 10 years, a lot of it was how do you get more volume out there because you're trying to be found online and all that stuff now it's to the point where you do have time to do that. So the way I think a marketer should be looking at this now is that it's true, some people might be impacted, you might not hire that agency, or there might be some people on your team whose work is redundant. But by and large there's more work, especially for growing companies than what there has been the capability to deliver. Really excellent, top notch work. That's what I think is going to end up being unlocked here. Is that the super contributor, instead of having that like free Shutterstock account or now they can actually have really nice imagery that meets what they couldn't afford to do an on site location shooting for. They can have video that in the past they might have only been able to afford one video a year. They can actually have a series of content, whereas they used to only be able to get one piece of content out the door. So I think that this is what I talk about, like things shifting quite a bit. I think there's just a positive aspect where now these marketing teams just have more capability to do things better on behalf of their company.
- [06:04]
Benjamin Shapiro
Yeah, you're right. I kind of went negative up front. I was like, oh my God, the agents are taking over, the T1000 is being built. Everybody get under the desk. And in reality for marketers now is you can do better, you can do more. And that's great. There is this idea that if marketers are going to be able to be more capable, we're essentially flooding the zone, right? And we're kind of seeing that with content already. Talk to me a little bit about the impact of the marketing practice where all right, one marketer now can have a copywriting agent and a research agent and an SDR agent and all of a sudden we've got this wonderful support system, we've got this agentic team that we always wanted and we're the brain at the top of the ecosystem. That's the idea that we're getting to.
- [06:50]
Nicholas Holland
It is like a symphony Conductor.
- [06:52]
Benjamin Shapiro
Yeah. And there's two things I want to ask you. One is what happens to the function of marketing? And two, the notion of time. So let's start with the function of marketing as everyone can start to do better marketing. What happens?
- [07:07]
Nicholas Holland
People ask all the time. For me, they're like, generative AI lets me create a high volume of content. All this AI slop is going to be on the Internet. And it's true, you can create more slop. But I think of that as even before AI, that's like a solved problem. There's a joke that we sometimes say in the world of marketing that marketers can't have nice things. Like, we always break stuff.
- [07:28]
Benjamin Shapiro
True.
- [07:28]
Nicholas Holland
You find a channel that is not being basically, it's like a channel that you can tap into where your audience is and then you have some marketers that abuse it. The easiest one to think about is like email. We talk all the time about, you should email people who want to hear from you. And then I hear marketers lament, well, no one wants to hear from me.
- [07:45]
Benjamin Shapiro
And you're like, huh, maybe it's not a channel problem.
- [07:49]
Nicholas Holland
That's right. And so a lot of marketers don't want to hear that. And it's always, well, if they just knew what I did, then they would be happy. And that's okay. Everyone's on a different path of that. What I would say is that spam, though, is the outcome of that mindset. And spam has been basically like a blight on the Internet for a decade plus. So when I say it's a solved problem, you already have many smart people out there working diligently to make sure that what shows up in your inbox is something that you want to know. So that's from governments, that's from providers like Gmail and Outlook. And then you get down into there's just best practices, you know, that the best marketers in the world are trying not to basically get kicked off deliverability list. They want to deliver, they want to have a good quality message. So if you're using AI to create slop, you're going to be stopped by government entities, you're going to be stopped by the sending layer, you're going to be basically blocked and ignored by customers. Public opinion, Public opinion. And then the same thing goes with creating content. Pick something like your blog, like it's your website, it's your brand. If you put out a bunch of slop, you've hurt yourself on there. And even if SEO, like, you know, it's Changing quite a bit. But even if you get into SEO, Google is working constantly to surface the right things for one search query. There's infinite amount of content, but Google is trying to surface the one that's most relevant. That's not going to change whenever you move over to LLMs delivering search as well. So I just go down the path of like I think AI slop is not a problem that we should worry about as marketers outside of the fact of if you make slop, what happened to you before AI will happen to you afterwards, which is no one will listen to you. So now if you take that as case law now you have such an optimistic future where all the things that you couldn't do to be super personalized, to be deeply researched, to be relevant to your audience, now you can do those. And I have seen where people are starting to do their old marketing the same way with AI and they have all this extra time, but they're not reinvesting it into better quality. And I do think what's going to happen is that you're going to have a series of people who did their old way now with AI and they're going to see lower results and they won't understand why. So you're asking what's going to change? I think that we're going to see many org level and role level things change. Like there's an interesting debate we have of should marketers be running SDR and BDR type stuff now? Because that's really a one to many relationship. It's one where the marketer is responsible for booking meetings and creating leads a lot of time. But it's often been an SDR is a role that then moves into a sales role. So that's why it's sat over there. So that's where you'll see Oregon role changes. I think you'll see people have to up their level of how do they focus on their customer more now that they have these capabilities. And and in general, I think what we'll also see is that customers will also start to change because I think their expectations are going to go up the first time a marketer delivers like a really beautiful interactive agent on their site. We even talk about this who wants to go to a site and click and click and click and click. If I can just ask a question and get an answer. So I think it's going to be a lot of change, but I actually think it's all well within a marketer's ability to adapt if they stick to some first Principles.
- [10:48]
Benjamin Shapiro
I've been working on building our first few agents, and I mentioned before, we have a speaker application ranking agent, which is an internal facing thing that helps us understand who we should prioritize. You scored very well. We also have the guest research. Once we agree to have somebody on the podcast, we do a deeper dive and look at things like what was their previous speaking opportunity, what are they uniquely qualified to talk about? And my experience building those agents is. It's a painful process up front. And I think that's one of the things where marketers get. Because we have this idea of campaign thinking. We want to get a campaign, create it, push it out in the world and see if it works. And if it does, wonderful, and if it doesn't, we'll kill it. But when you're building AI agents, you sort of have to push it out into the world half completed and then get some data and then keep working. It's more of an engineering mindset where you're constantly tweaking and iterating on this agent to give it more rules and guidelines to continue to optimize. And I just found that you basically have to do everything twice that you used to have to do once to build an agent, but then once you've gotten to a good place, you never have to do it again. And it's definitely a mind shift change for marketers. So I'm sure you've built a handful of agents, obviously, and probably the infrastructure for HubSpot as well. Am I the only one that's having this experience? Like, how do we get our mindset around doing the work of creating agents?
- [12:13]
Nicholas Holland
There's a fascinating thing that's happening now with bigger companies talking about should they have intern programs, should they have associate programs where you bring in somebody? And one of the reasons that they bring it in is that they're constantly saying, it's just a lot of work to train these new people out of college. It's a lot of work to kind of spend time, slow down, to give them context and all that stuff. You often see that in small businesses. They make really slow decisions on who to hire next because, my God, it like slows the whole machine down to add a team member. It's no different with AI. The thing is we think about agents as something like a digital teammate. If I were to say to you, hey, you've added somebody to your company, Ben, like, they're on your team and real quick you're going to put them out in the wild doing a job and they will know nothing about your company, they'll know nothing about how you like to do things, they'll know nothing about your process. None of that. And you'd be like, I would never do that. And you say, right. You would train them. Right. And you're like, absolutely. Like, we do a six week shadowing course, we then monitor what they do for a while. That's what you do with an agent. You have to effectively start to give it context. And there's this phrase, for years we used to say data is like a moat. Data is the most important thing. Data, Data. Data was all marketers talked about.
- [13:19]
Benjamin Shapiro
Data's the new oil.
- [13:21]
Nicholas Holland
Yeah. Now what you see is marketers are saturated with data. I have so much data, I don't even know what to do with it. So really it's like we used to say now insights. And now it's shifting to this term now that's becoming quite popular in AI called context. You hire somebody to join your team and you give them context of the company, you give them context of their role, and then the relationship between the manager and them has context month over month over month of the things that you're working on, kind of project or initiatives. And you also expect them to have all the context of a college degree and a background. Growing up in a professional environment. That's all stuff that people don't think about now when they're starting to work with agents. And so it feels heavy. It feels like, wow, why didn't just work out of the box? But when you start to think about him as a teammate and that context is the real key here, that's why it's going to be so profound. And that's kind of like what we're working on here at HubSpot is I want every agent out of the box now to start to know who the company is, who you are, the tasks that they've been assigned to do, background of things that have already worked, the industry, all of that. And so we're working really hard to have all of that context ready. So when you go in and hire the agent, you're really just working on the initiative and all of that is already brought to the table.
- [14:31]
Benjamin Shapiro
I think that's really important. And honestly, that's probably where most of the work is for marketers. Here's the pro tip. Start with a manifesto. All of the context, this larger body of work, no matter what you're doing, whether it's your campaign, strategy, your company, you have to give all this context. But if you can basically create a formatted Context upfront. It's a really useful tool to hand LLMs to be like, hey, here is everything you need to know about my company. Our marketing strategies, our goals, our tone of voice, the Personas, everything we've done, what's been successful, what hasn't. If you could feed all that context up front to an LLM and then you ask it to start taking some agentic tasks, it's going to do a better job. That context and that manifesto upfront really matters ton better.
- [15:17]
Nicholas Holland
Yeah. And so there's a lot of AI forward thinking now. People that literally to your point will have special docs, special PDFs, they have a rhythm of how they bring context. All these LLMs, I think actually that's going to fade away. I think you'll see the LLMs start to basically keep up with some context on their own. You're seeing that already with memories. But I think again for me at HubSpot, where we are focused exclusively on like the go to market functions, marketing, sales and service, there's an enormous amount of context that should just be ready to go when you start using HubSpot, period. What's funny is like right off the bat when you sign up for HubSpot, we already start building out all the company context. So how you said you like researched me, for example, imagine now like I signed up for HubSpot, I research everything about the company, I research everything about the user and then I store those as context that the user can then go in there and edit and collaborate on. Same thing goes. I understand their role, I understand the industry they're in now I start to look at the products that they sell. All of that context, you don't have to have that massive PDF, we just have that baked in. And I think if anyone's never seen an LLM respond pre context and post context, that's one of your action items from this particular call. You've got to go see, it's night and day.
- [16:24]
Benjamin Shapiro
Yeah, you have to give it an understanding of who you are, what you're trying to accomplish, all of the well context. But also there's this notion of kind of micro tasking, not asking for too much upfront. One of the things that I've experienced building AI agents is that if you go and you're like, hey agent, rank this speaker for me. It's going to come up with some arbitrary numbers. If you go through and you say first evaluate the speaker and find some information now understand what the company does now understand whether they're a good fit for Our speaking Persona. Okay, Take those three bits of information and then create some sort of a ranking. It does a much better job. How do you think about the exercise of prompting and how do you ask for the right amount of information without asking for too much?
- [17:16]
Nicholas Holland
So for everybody here, you might hear the term prompt engineering. And last year that was like really, really key. And the LLMs have all evolved so that specifically the concept of chain of thought, which is a fancy way of asking the LLM to like take the problem, break it out into a plan, and then prompt against of those.
- [17:34]
Benjamin Shapiro
It's reasoning, right? It's anything past the O1 model, you got it.
- [17:37]
Nicholas Holland
That's like where reasoning's original concepts came from. So now instead of them effectively making you ask for a chain of thought or saying please break it out into steps, that's all baked in now and is able to decide how much reasoning to put into it. So what I think will end up happening, Ben, is that when you say, go rank this user, it might get it wrong. Not because you didn't prompt it all the way there, it's because you didn't give it enough context. And so this is where like these ideas come together very in a wild manner. Hey, go rank this user. It gets it wrong because it doesn't have enough context. But whenever you say in that thread, hey, go research it now, that's one level of context. And then it's, hey, then go see what things they speak about. Now it's another thing to basically another piece of context. And then it's like, hey, what are the things that they should speak about? For us, that's another piece of context. And then it's rank it now it has kind of three other pieces of context to now do that. That's why a lot of people now are talking about what they call context engineering. Because had you uploaded that PDF or had you made an agent that already had something along the lines of like, this is what a good speaker looks like, the reasoning systems now will go out, they'll research, they'll look up things that they've done. They'll then look at some people call it like an eval. And then it'll basically see how it stacks up and then it can do it. So there's a lot of stuff under the hood that's happening here, but the context engineering is kind of the takeaway that a lot of people are starting to think through. This thing is now smart enough with reasoning. Now the question is, how do I give it context appropriately and tell it what Good. Looks like if you have those two things, you're off to the races.
- [19:08]
Benjamin Shapiro
Yeah, it was an interesting exercise going through and doing speaker rankings and guest research. And I've found that you can't give too much context up front. Our Martech Manifesto is a 20 page document and when I try to feed it into an LLM, it's just got too much and it can't figure out what to parse out and what to prioritize. So I was like, here's the document. Figure out our tone of voice, figure out our priorities. Like figure out these four bits of information and always prioritize them. Now go evaluate the speaker. Now go evaluate the company. Now go evaluate all of this and then compile it all at the bottom and come up with the score. And it did a much better job because I am not just giving it a ton of context, but I'm also telling it what the important context is. And this goes back to this notion of a super contributor. I feel like I'm positioning myself to be a super contributor, which sounds very egotistical, but if I'm creating AI agents, anybody can. And now all of a sudden I have the capability to do these really fascinating, deeply well researched activities and we're agentic over time. Let's go back into what the impact is into marketers. I as a marketer can do way more and it can be way more in depth than I did before. So am I going to be working with agencies anymore? Am I just going to be the only marketer in my organization? What's the future look like there?
- [20:23]
Nicholas Holland
No, you will, you'll work with agencies. Still, it may look different. What I think is happening is AI search is kind of rewriting discovery. So with Google AI overviews now you have LLMs like perplexity and others that can answer that. Like there is this new way that people are being discovered and there will be techniques around that that will require some specialization for a while now, maybe over a longer arc, you might not be there, but there's little techniques. Now. How you shape and handle content on Reddit, for example, has an outsized impact of how you show up in the LLMs. Many marketers I know, no matter if they have the most well equipped AI capability in the world with chat, GPT, et cetera. Like Reddit is kind of a dark art on how to do it, how to effectively spend time on it, how to be authentic. And almost everybody can spot when you have marketing slop on Reddit and they'll.
- [21:13]
Benjamin Shapiro
Tell you about it.
- [21:14]
Nicholas Holland
Oh man, and you can get low reputation bounce off. So that's an example where an agency would be really good and they would be using AI, but they would know how to use AI in the right context. I think that what you also have is a world where like AI probably took a lot of your top of the funnel. Now there's a lot of optimizations you can do in the middle of the funnel. So once again, when you get into a world where you talked about setting up the agent, like there's still art in setting up the right agents. There's still art in orchestrating a bunch of people. The easiest way I look at is if you had unlimited budget and you could put tons of people on your team. Not everybody is still a great manager. Not everybody knows how to deploy human resources. Not everybody is able to shape kind of a multi human campaign. So there's still going to be room for agencies. But if Your agency charged $300 to kick out a three paragraph blog post, that could be a real problem. So I do think that just like I talked about how the benchmark is moving up for the marketer, the quality bar is moving up. Agencies that did certain types of roles, they're going to have to adapt. But outside of that, I do think there will definitely be agencies.
- [22:16]
Benjamin Shapiro
It seems like there's this split between leaders and laggards or people that understand how to build prompt engineering context engineering building agentic agents. They're going to be much more powerful resources. And then there are the people that are fighting it and trying to do the way marketing used to work and they're really at risk. So let's try to help them. For the laggards, what are the steps to move from? Hey, I occasionally use ChatGPT to like figure out how to write an email to building agentic agents and workflows.
- [22:47]
Nicholas Holland
I talk about it internally at HubSpot. There is this AI mindset that I really encourage people to adopt. This is also of my own staff and of hubspotters. First, you need to have curiosity. There is a world where I don't know if you've seen this pre AI, there was this kind of interesting sentiment growing in the world of marketing that everything is saturated, everything's already been done. Marketers have sucked all the bone marrow out for every single channel. Like it's very hard to stand out. And now with AI all the rules have been kind of reset and so there's a ton of room for creativity, a ton of room for doing. Did you see some of the commercials that have come out being written by VO3.
- [23:26]
Benjamin Shapiro
Yeah, I've seen a little bit of it. Tell us about it.
- [23:28]
Nicholas Holland
It's amazing. So VO3 is a custom model from Google under the Gemini family. And Google has created a model that can do eight second videos but with audio, so background. It has characters that are talking, it's perfectly synced with the lips. So people now are able to make just anything they can think of. So there's amazing commercials that have already been made, completely made by AI where it's like scene after scene after scene. And that one where the technology is there, but you can't account for taste, you still have to have tastemakers that are involved there. So curiosity is something that I think that wasn't as high because everybody thought, I know how to do that. Yes, I'm familiar with that technique. Now. Curiosity should reemerge. Everybody should be trying to learn again. The second thing is I think adaptability. What worked yesterday may not work today. What got you here, may not get you there. All those phrases you hear, innovator's dilemma, who moved my cheese? All of that, that is happening now. So you've got curiosity, adaptability. And then I think that right now there should be a healthy degree of paranoia. We're in a world right now where so many things are changing that effectively there is risk out there involved. For companies that don't become a leader, they say a laggard, there is risk out there for people who think AI is a trend or security is the most important thing. I don't know. I don't know if that's true right now. It depends on the competitive landscape. So that paranoia is interesting because that is something like healthy paranoia will really be that kind of impetus for the org to understand that at the end of the day, if you aren't innovating on behalf of your customers, don't worry, somebody else will. And I think that that's the intersection of how you should start that AI journey from there. There's a lot of things that you do from just starting to learn how to interact with the frontier models. That's kind of a V1.
- [25:14]
Benjamin Shapiro
Frontier models being the GPTs and clouds of the world.
- [25:17]
Nicholas Holland
That's right. Chad, GPT, Claude, Gemini. And then I think once you get some of your personal productivity underway, then that's where I think you start to move into how do you start helping out your team? So for HubSpot, we treat it super simple. We say at your company, get your data game on point. So that means like get your structured data Like CRM and unstructured data together. Unstructured like your calls, emails, all that stuff. Second, give every employee their own AI assistant. So for go to market employees that we serve, we have the breeze assistant. So it's your kind of AI chief of staff for marketer sales and services. If you don't want to do that, you can go get a ChatGPT or a Gemini or an Anthropic, for example. Claude. So now you've got your data game on point. All the employees have their own AI assistant. And then we believe in a world where there'll be these hybrid teams with both humans and agents. So you start to do like what you were doing, Ben, grind out training your first agent. And is it messy? It's like managing your first employee, of course. But you start to get the agents on the team and we have out of the box agents that are ready to go. And then also you can start to build custom ones. And then the last thing you do is you then get your whole company bought into this concept of being an AI leader. So that means that you choose software that is AI native or AI first, and you begin to get the whole company thinking about how would you approach these problems. And when you do that, those four steps, simple doesn't mean easy, but that's where things are. Get your data game, get an assistant, get the agent, and then get the company using software that's AI native, you're good to go.
- [26:42]
Benjamin Shapiro
It sounds terrifying. I'm not going to lie. And I imagine that if I wasn't relatively well versed in artificial intelligence the last year that I've been using it just like everybody else, I would be really intimidated if I'm just like, yeah, I've been fighting this AI thing and I don't believe it and I don't want to do it. You're going to lose your job if you fight it, right? Eventually somebody will be able to do what you do more efficiently. But I have good news. You don't have to learn a new technology. It sounds crazy, but you're using this really advanced technology that is built for you to be able to just type and talk to it. You don't have to learn a new language, you don't have to learn to code. You just have to work with it. You just have to practice. And that to me is the experience. Having built what I think are relatively sophisticated agents, how did I do it? I asked the agent how I should build the agent.
- [27:29]
Nicholas Holland
Totally.
- [27:29]
Benjamin Shapiro
I asked ChatGPT how I should put the prompt into Zapier to make it do the thing I want. And meanwhile, it's like, oh, you need a webhook, and here's how you set one of those up. And I'm like, I'm not that technical. I don't know what a webhook is. And it's like, just put these fields in these boxes. I was like, I know how to build a webhook now.
- [27:46]
Nicholas Holland
Yay.
- [27:46]
Benjamin Shapiro
It takes your ceiling off.
- [27:48]
Nicholas Holland
It does. You're a super contributor. Monday, you didn't know how to build webhooks. Tuesday, you don't need a developer to go build webhooks. Now, what would you call a developer for something that's harder than that, but now you have a wider set of capabilities.
- [28:00]
Benjamin Shapiro
All right, the last question I have for you is a little bit about the autonomous campaign set up. Now, the laggards are using AI. We're all on the same page, we're all rowing the same direction. Everybody's got the skills. Let's do some autonomous stuff. Walk me through autonomous campaign management. How do you get it set up and what does the practice look like now and what it'll look like in the future?
- [28:20]
Nicholas Holland
We're in a middle zone right now. So here at HubSpot, a campaign for us is you're going to do a collection of marketing activities towards a goal. And what's cool about, again, I tell all marketers, revisit first principles. Like, you're trying to help your customer learn, you're trying to help your business grow with leads. Like, those first principles will guide you. Then AI just becomes like a way to get it done. So a campaign is this collection of things towards a common goal. Well, what we found is that even today with where agentic stuff is today, we go back to the things we had just talked about. You have to get your context in one spot. So the most early part of that is like a PDF file or a word doc or something like that. But with HubSpot, this kind of stuff is like, already taken care of for you. But it has to know something like your role. It has to know your business, your industry, what thing is the campaign going to be about? You then have to have details on who's the target audience. You have to have details around your brand voice. Like this is all stuff that you just said when you were doing your own stuff. So you have to have all of that somewhere. And so when you have that context, then you begin to now move into, okay, what's the goal of the campaign? And so the agents like we have, for example, take all that context, they take the goal and then the next thing they do is they basically create a scaffolding of a bunch of different content pieces for you to look at. So if it's a multi channel, it could be blog post, it could be emails, it could be YouTube scripts. We even see people now where they'll want to take a video. One of partners I really like, he was like, I take the master format, he takes a YouTube video, he puts it in. We have a tool called Content Remix. He puts it in there and then Content Remix shatters that out into a bunch of short form clips, pulls out the script, it makes like a audio version of it. So the deal is now you have this scaffolding of all of the different content and then you go in there and each of those now is like a human agent back and forth where you're tweaking it, getting it ready to go, you're getting it onto a content calendar and then once that's published, once it's out there, you start to get into more specialized agents. I would tell every marketer, there's a very few are all the way at the end of this journey. But like specialized agents now you're taking that content and you're promoting it on paid channels and there's agents that manage that and manage your budget on that. Now you're basically starting to create social posts and you want to have agents that are monitoring it to see how is it resonating, are people interacting with it. So now you'll start to see that like you started with all the context, you had a goal, you then used it to scaffold out content, you collaborated on each piece of the content and now you have agents that are kind of watching that towards the end of that stream. And all of that data is now going to go back into helping you make decisions on what to do next. So it's a little bit hard with ChatGPTs and Geminis, but in a system like HubSpot, you want something like your Breeze assistant, you want it to pop up and it does and let you know, like, how's it going? What are things you should change, what are kind of things that are not going well. That's kind of what the agentic campaign looks like in the future and that's going to be pretty cool.
- [31:06]
Benjamin Shapiro
What I'm hearing from you is there's an agentic strategist, there's an agentic creative, there's an agentic ops agent, there is an agentic media buyer Then there is an agentic data scientist that's evaluating the performance. And each step you're basically handing off one piece of work from agent to agent to agent and hopefully having it optimize the campaign performance over time.
- [31:29]
Nicholas Holland
That's right. And we don't have this yet, but one of the things I'm really thinking about here at HubSpot is I think that there will be an emergent role. So it's this super contributor. But I think one kind of like sub bullet of the super contributor is that you will be an agent manager. There's a lot of interesting things that go into that, like how do you collaborate, how do you orchestrate? And then I think there will be an agent trainer. Sometimes they will be the same, not always. By the way, I'm seeing worlds now where for the first time you'll have a sales leader who is like the agent manager, getting a prospecting agent assigned to all the reps. But then the reps are tasked with updating training material for their own agent. So new roles, new techniques. But when you step back to, like, first principles, it's like you have a new worker on your team. Somebody's gotta train them. Somebody's basically signing the bill and paying for it.
- [32:13]
Benjamin Shapiro
It's funny how we can do so much more, but the activities are relatively the same. All we have to do is sit here and evaluate and still do training. We're just not necessarily training the people. All right, that wraps up this episode of the MarTech podcast. Thanks to Nicholas Holland, the head of AI at HubSpot, for joining us. If you'd like to contact Nicholas, you could find a link to his LinkedIn profile in our show notes or on martechpod. Or you can learn about HubSpot and how they're building AI for marketers by going to HubSpot.com AI and if you haven't subscribed yet and you want a daily stream of marketing and technology knowledge in your podcast feed, hit the subscribe button in your podcast app or on YouTube and we'll be back in your feed next week. All right, that's it for today, but until next time, my advice is to just focus on keeping your customers happy.
- [33:01]
Nicholas Holland
Foreign.
- [33:23]
Benjamin Shapiro
Thanks for listening to the Martech podcast and I hear everything. Production. Looking to launch or scale a podcast like this one for your brand? Then visit iheareverything. Com.