← Marketing Against The Grain
Loading summary
Transcript5 lines
- [00:00]
A
ChatGPT agent is out. It combines OpenAI's Deep Research Operator mode and the intelligence of ChatGPT to help you do everyday tasks. Now, you probably watched some content already on ChatGPT. You've probably heard a lot of big claims, but is it really any good? Well, we're going to put it through its paces with three core use cases that any marketer or growth operator or someone who's looking to grow their business would want to use ChatGPT agent for. So we can actually give you the unbiased version should you use this or not. All of that and more on this episode of Marketing against Ukraine.
- [00:45]
B
Here'S a quick word from our sponsor. HubSpot helped Tumblr solve a big problem. They needed to move fast to produce trending content, but their marketing team was stuck waiting on engineers to code every single email campaign. Now they use HubSpot's customer platform to email real time to trending content to millions of users in just seconds. The impact? Three times more engagement and double the content creation. Want to move faster like Tumblr? Visit HubSpot.com.
- [01:31]
A
Okay, so ChatGPT agent is out. With every great launch, there is a steady stream of big claims made, especially if you hang about on X or YouTube. And I for once have access to a tool that has just come out. I am in the eu, we do not normally get this privilege. I'm not sure what's happened, maybe they've messed up in some ways, but I have access to the tool and so I've started to play around with it. Now, off the bat, what I would say is I have not felt a lot of noise around this launch. As much noise as I would have thought. Right? There's the obvious YouTube videos that spring up. The X claims that everything is incredible. There's wild examples, but it hasn't been as noisy as I would have thought. It's a pretty cool tool. It combines ChatGPT's deep research, which is a great product, its ability to do tasks via operator it, and then just the intelligence of ChatGPT all in one agent. And we've heard forever that this is the year of agents. I do believe that will come true at some point. A lot of agents out there have really not met the needs of users. We're really still trying to have agents be as good as we want them to be. And so is this a major breakthrough? Is ChatGPT figured out how to create an agent that can really start to automate tasks, not just retrieve information, but actually take action on that information and do things for us. This is the future we have been promised. Now like any good launch these days, the benchmarks come out along with the launch and they always look really good. So chatgpt agent here on humanities last exam where it's able to do expert level questions across different subjects. Wow, looks really, really good. Very, very impressive. This is the one that really I think matters a lot which is its ability to create tasks or do tasks measured against humans. Ability to do that they estimated time for a human to complete tasks. And you can see ChatGPT agent is really starting to show signs that it can do tasks and is comparable to a human. I think one of the stats I heard is it was about better in 50% of the cases than the average human. It's pretty good at financial things. I actually wouldn't mind trying this one out which is investment banking, modeling tasks and then it's pretty good at math. And so when you look at the benchmarks, each new launch does come with a set of benchmarks and those benchmarks look really, really good. And so you would say, wow, this is an incredible tool. I really need to start to use that now what ChatGPT agent is and I'm going to show you quickly some use cases. We're going to demo it here live. It provides the agent with a virtual computer to do all of these tasks. So the agent has access to tools like web Verizon and these different tools and it has the ability to complete tasks within that virtual computer. And the UX experience is actually really, really nice. Okay, so I think this is like one of the more common tasks. A lot of the research task use cases is still very heavily reliant on the deep research which is part of the agents toolset. But this one here is going to allow us to create a competitive matrix from 10 companies competitors. So we're going to give it a domain. In this case we'll just give it HubSpot. And then we're going to research 10 closest competitors. We're going to go to their website, visit all of these different pages, then we're going to build a competitive matrix. Then we're going to highlight any gaps or opportunities not covered by these competitors because we've been on their website and we're going to output this in a structured table and we have some constraints and then we have an output format. So this is a pretty good one. Like you know, in days come past, how long would it take someone to build a competitive matrix? It would have taken your competitive analysis team or your product marketing team some amount of time to be able to do this. And so we are going to switch on the agent mode. Look at me, I have access to agent mode. I'm just so happy that I actually have access to something near enough around the same time as other people around the globe have access to something. I feel like I'm living in a very forward looking country right now. That's probably the only time this will happen. So we switch on agent mode, which is pretty cool. Has some suggestions of what I can do. You can see a lot of them are like everyday tasks. Reserve tables, catch up and team conversations, order pizzas, audit fast fashion versus okay, schedule grocery delivery for tomorrow. So like I need to look at that one. I didn't know you could schedule tasks that would be pretty cool. But first of all, let's, I think let's run this one and see how it gets. I'll cut the video so you don't have to watch the full thing. But I want to show you the ux. We're going to kick this off and you can see it's going to bring up its little virtual computer, which is very, very cool. So it sets up its desktop, turn on its laptop. Sometimes it has to turn it on and off if it's not working the way it wants it to work, just like all of us. Okay, So I really like this. So you can see it has its little virtual computer and you can see that it's starting to actually do these tasks, go into Zendesk. I think this is a really slick UX experience. You have these little options here. You can switch between this mode and see the little screens here or you can switch back to desktop mode. You could take over the browser, which I'm not going to do because I want it to run this. So again, I actually think this is very, very, very cool. Let's let this run and then come back to see what the results are now again, because this is the future we do live in. I'm going to kick off the other two as well at the same time. So we're not going to slow down here. So let's try to use all of our credits in one go and kick off some more things here. And so I'll switch to a different screen. You're watching one of my workers at work. Let's kick off another virtual worker because this really is the future we are going to live in where we're just going to have a bunch of virtual agents doing a bunch of stuff for us. So we have one going there. Okay, so the next one I'm going to show you, which I think is a interesting one. It's still another research one, but actually I think is an interesting one to see. I actually don't know how this one will work because what I'm asking it to do is reverse engineer an ideal customer profile by analyzing real LinkedIn profiles. And so I'm wondering, does it have access to LinkedIn profiles? What I have found and I am building a. A bunch of tools and so I have had to dig deep into this. They all do have access to LinkedIn profiles. When you're doing it through the interface, through this, when you're doing it through the API, they do not inclusive perplexity. So if I want ChatGPT to pull back information around LinkedIn profiles, it usually can't, which makes sense because they probably don't want people building tools that scrape a bunch of LinkedIn content. And so what it's going to do is reverse engineer the customer profile for CMOs in the United States, analyze real LinkedIn profiles. It's going to identify five chief marketing officer profiles working in the US, preferably B2B SAS. It's going to then take all of that data, analyze those things, and then start to create a actual icp. That is a pretty good way, if you are a smaller company and you want to be really scrappy, to create a first version of your ICP is actually pull a bunch of LinkedIn profiles, which are people you would sell your product to, and put them into one of the AI tools and have it build an ideal customer profile. So for people following along, an ideal customer profile is this is the fictional representation of the people we actually want to sell our products to, we want to market our services to. It actually helps me tailor my entire strategy towards a specific person, towards a specific demographic. And so this is a really good way, I think, to create a first version of your ideal customer profile. And we are going to kick this one off. So that one's setting up its computer. My other little worker is working away here. And so why don't we kick off our third. How productive am I going to be? I might actually just, you know, take the rest of the day off. All right. The third one is like close to my heart because I am a fan of a tool that I use a lot called genspark. And genspark is a AI agent that is able to craft really great presentations. Now, one of the things that we saw in the demo for ChatGPT agent that was front and center as a use case was its ability to craft slides. And I want to put it through its paces along with JSpark. And so I'll run this prompt in OpenAI ChatGPT agent and I'll also run it in Genspark and we'll see which of those tools does the best job. Genspark is an AI agent, has a lot of different agents. One of them allows you to create presentations. So I'm going to show you that as well. So let's kick off another little worker. Got my own virtual team. Whatever you think about the results, think about the way that I'm working because this is how we're going to work in the future. This is how we are going to manage work. All right, so you are a competitive presentation specialist. We're going to research, we're going to create a competitive analysis presentation. So again, kind of similar to the other ones because we want to continue to use the core functionality like deep research, its ability to craft that information into things and take action. And so in this we are going to actually have it craft a competitive deck on HubSpot. Again, one of the use cases here that if this worked really well is in the future you would have one of these created for target accounts for sales. Right. So a contact would come in, we would qualify that sales would create a deal and you would automatically kick off an agent to create a deck around that target account. So these use cases can all be integrated into your go to market in some way. And so we're going to kick off this one here, which I'm asking it to do an executive summary, which is two to three sentences and then some slides. Just gets a little messy. Because we're a public company, I'm not going to be pulling an internal data to show you here. So I usually use synthetic data, but for this I just want to use external data. So these are the slides here. I did give it a structure and so we will turn on our agent mode. Okay, like let's kick this off. Okay, so we're going to give it the same prompt. Usually what I do with jenspark, to be honest with you, is not prompted in this way. I have an O3 assistant that I have trained on how to create great presentations. And so I would Normally have the O3 assistant create prompts for me. I didn't do it in this case. So this is looser than I would normally prompt genspark. So I'm interested to see how well it does. And let's kick off Genspark okay, so we have four different agents running. One of the things you realize when you start to do this at scale is you really do need like a dashboard for all of your agents. Right. I want to be able to see how all of my agents are progressing at once. How many agents have actually kicked off, because I start to lose track. I think this one is finished, actually. This was the one that was building the competitive matrix.
- [11:56]
B
Hey everyone, we'll be right back to the show. But first let me tell you about a podcast that I love. It's a podcast called I Digress. It is hosted by my friend Troy Sandage. It's brought to you by the amazing HubSpot Podcast Network, the audio destination for anyone in business. And Idygress is great. It's got 30 minute episodes and the podcast is all about helping you eliminate complexity, complications and confusions in your business. It's really heavy in frameworks and strategies to help you scale and sustain your success. And recently Troy had a really great episode. It was called Branding Leveraging Style to change the narrative, boost, influence and profits. And that was with his guest toy, Sweeney. It's really good. You should check it out and you can listen to that or all the episodes of I Digress anywhere you get your podcast.
- [12:52]
A
Let's first dive into the one where it took the LinkedIn profiles. You can actually, I think, go through its work. This is what I wanted to do. I can see here the actual chief marketing officers and how it's pulling those LinkedIn profiles. I really wanted to see if it was getting them from the LinkedIn profiles versus getting them from sources, which I see it pulling here. So it pulled the CMOs that we wanted to pull. So we have some CMOs here, gives you some patterns which are pretty cool. Then it gives me an ideal customer profile. So it talks about company size. Now we could have specified company size, although I don't think it could have got the company size. I think from LinkedIn. We probably should have pushed it on that. Talks about the industries, talks about key responsibilities, pain points, common goals, buy in triggers, which is pretty cool. Preferred tools, their decision making power. CMOs have significant budget for marketing software and services, but often collaborate. Yes. Message and tone, that resonates. And so it gives you some message and tone. Then how do you position your product for these CMOs? So I think that's a good one. This one, instead of actually getting it from LinkedIn profiles which are public, it has used other sources. I think you can play around with this use case, but pretty Cool, right? You would be able to reverse engineer real profiles if we figured out how to get the data from LinkedIn. Or in this case you can figure out how to reverse engineer real profiles from publicly available information about these folks. The other one I think is cool is that they went to visit all of your target competitors. And so we have the different competitors. It went to see Salesforce, Zoho, ActiveCampaign, Mailchimp, Adobe, all of these different competitors. It visited their website. So it's pulled out a ton of information here. Features, pricing, positioning taken from the website. Salesforce grow faster and work smarter all in one CRM with AI agents and unified data. This is super cool actually. It's pulled the position and convert more. Build last relationships and grow your business resiliently with the magic of contextual AI and thoughtful ui. It's actually pretty cool to actually just see all of the different ways that people are using AI in their strap line. So that is awesome. Then it goes who are the target audience? It's got me the target audience for all of these competitors and then the differentiators. So what has it got for HubSpot? Broad ecosystem, extensive free tier, strong content marketing tools. Let's pull out another competitor to see what they think is differentiation. Let's go pipedrive High visual intuitive interface AI powered suggestions, affordable entry, large integration marketplace. So I think that is actually really cool. I could play around with this forever actually to pull out some more information around competitors in different ways. Executive summary Giving me a good executive summary here. Underserved features and market gaps Integrated customer service. Many competitors concentrate in sales and marketing automation. Fewer from HubSpot Service Hub provide robust customer service tracks. So that's pretty cool. Affordable omnichannel marketing vertical Specific solutions, strategic opportunities leverage HubSpot's unified platform, expand the AI capabilities. Provide a complete front office suite. So it's done a pretty good job here. Let's get into our last one which is still ongoing, which is the big one, the presentation against jenspark. But we should go over to see what genspark is doing. Okay. Genspark. Look at that. Genspark go. What a great tool. I'm sorry But I think OpenAI is going to struggle to get anywhere near as good as this. Kudos to genspark. I feel like I need like the kind of clock counter dn dn din din. You might not get that in the usa. I think that's like from a show that's in the UK. But come on ChatGPT agent. This was one of your Primary use cases we have like genspark is just like, look at this, this is so good. Like my prompt wasn't even that good for genspark. It's not as good as my normal one. And this is kind of blowing my brain. How good? I use jenspark all the time and I'm just kind of. My brain is a little melting here how good this actually is. It's kind of like genspark for presentation is a little bit like replit or lovable for code. It's just figured out how to do it and no other agents have really figured out how to do it. Like if you're in a marketer in a big enough company, you're doing like presentations all the time. What a time save this is, right? Who wants to spend their time doing lots of presentations versus actually doing the craft. But it's a great way to communicate things and so they are needed but like the other thing just to do. So we're clear, genspark pulled all of this itself. I just asked it to go and research these things. I didn't give it that information. I'm mostly just doing an advert for genspark now Please, jenspark, I need to get the founder on. If anyone watches this, has a contact with the founder, please let me know. Okay, so it's obviously pulled out externally, but I would have to like validate this. But it's like pretty close of what we did. All right, we'll check back here. I do have to be somewhere in 2026 and so you know, if it's going to take another six, seven friggin months to finish this, you may have a new presenter finish off the end of the video. All right, so after a good 45 minutes it is back with a presentation. Look, if you had given us this months ago before we have tools like genspark, this would seem pretty cool. It's giving me the financial, it's giving me competitor feature comparison against AI and automation pricing ecosystem SMB. So it's given me some like feature comparisons here to other companies. Give me a little SWOT analysis here which is kind of cool. Give me some market gaps and opportunities, some strategic recommendations. So for 45 minutes I don't think you're saving that much time compared to JanSpark which was like less than five minutes. It still has a ways to go I think for some tasks for sure. But I think one of the things I would take from this episode so we went through three use cases. It's pretty good. Again, a lot of the goodness is coming from its deep research tool, which is just a great product. I have not put it through its paces as much on how to create action. We did that once to create a deck, which is a very common use case for marketers and knowledge workers. It took 45 minutes to be able to do that. That's a long, long time, comparable to tools like genspark. I still think these tools have a ways to go, they are making progress. And what I would take away from this video is just the way I started to work, right? I had four virtual AI agents completing tasks in parallel for me. And so at some point I probably could have hundreds of these things working for me. And that is just a very different type of day that I'm going to be living where I can just kick off all of these agents and they're able to do tasks. And I think what I would look at here is how much better ChatGPT agent is than operator. Like, what is the difference between those two things? And then do I expect this to get much, much better over time? And I do. And so I do think it's worth playing around with it, starting to figure out where it can be useful. And then over time, assuming that it's going to get much better, or you're going to have comparable tools that are much more specialized in certain areas, like a genspark for presentations, where you're going to have these virtual agents be able to do things for you. So that's the ChatGPT agent. I think it's cool. It has a ways to go in some ways to do like a lot of things for you, but definitely worth digging in and starting to see where you can use it in your day today. Sam.