55 | How AI agents can transform B2B marketing by automating high-quality LinkedIn content and lead generation with Carolina Posma

How do you scale thought leadership without burning out your team? Carolina Posner, founder of PostifAI, shares how AI agents are helping B2B companies turn their internal knowledge into consistent, effective LinkedIn content. In this conversation with Audrey Chia, she explains the difference between automations and true AI agents, what makes content actually convert, and why a systems-first mindset is key to using AI effectively in marketing.

Join my weekly Newsletter: https://lp.closewithcopy.co/welcome

Transcript

Audrey Chia: 0:00

Welcome back to the AI Marketers Playbook, where we cover actionable frameworks to help you leverage AI and marketing strategies in your business. I'm Audrey Chair Host, and today I have with me Carolina Posner. Carolina is the founder of positive AI and AI agent that turns your company's existing knowledge into high quality LinkedIn posts and carousel. She's well known for her work on AI agents Made Easy and her accessible content on AI agents for non-techies. You should check that out now with a background that spans legal tech, ad tech, SEO, and AI consulting. Carolina has been experimenting with AI for nearly a decade from using machine learning to read contracts at a top law firm to a pioneering AI generator podcast to building marketing automations. And an AI DM appointment set. Now as posted ai, she's helping knowledge intensive B2B companies use LinkedIn more effectively to attract new clients. So today we are gonna dive right into how AI can transform the way companies, not just individual creators, view thought leadership on LinkedIn. Carolina, it's great to have you with us.

Carolina Posma: 1:04

Thank you so much for this introduction, Audrey. It's wonderful. Thank you.

Audrey Chia: 1:07

I'm super excited to dive right in. Now, I know you have a very colorful background. Tell us a bit more about how you got started in the world of ai and prior to that, what were you doing and how did it, you know, uh, lead up to this company that you're building right now?

Carolina Posma: 1:20

Yes, absolutely. So you already quickly mentioned it, but I started, um, exploring AI already almost 10 years ago. Uh, this was when I was, uh, studying law. I, I was studying law in economics and I was doing an, uh, student's job at the law, uh, at one of the big law firms in the Netherland. And I was part of the legal tech department and back there, uh, there was, was pre, uh, pre chat GPT and GPT times like it's, uh, we were really experimenting with character recognition of contracts. Well, to be honest, AI was not good enough to actually implement it, uh, within the law firm. Uh, it made too many mistakes, but I already got a bit of a feeling on what's actually already possible with AI back then. So. After that, uh, I, I did my masters and after I graduated, well, you may recognize it as well, but I did a bunch of different entrepreneurial activities, businesses, and, uh, just like every entrepreneur does, exploring so many different things. Um, so, but everything that I did, uh, from the start actually evolved around AI already. So, uh, my first company was, uh, with my best friend. We started an educational tech startup where we were. Um, turning study material into short educational podcasts, and we started it off with, with everything completely manually, but we're experimenting already a lot with AI voices and, uh, GPT-3 0.5, uh, the data access just got launched, um, before chat, GPT. So first had the API access and we were thinking on if we could actually use that, and that's, that was the first exploration with GPT as as a model. And after that, uh, I started a content and SEO agency and did a bunch of LinkedIn ghost writing as well. So that's also where a lot of my LinkedIn journey started back then. And, um, back then, uh, maybe you know it as well, uh, or use it in the past as well, but I use Jasper AI a lot. Which was the big AI rapper for text, uh, text ai. And it was huge, but a lot of people also didn't know it. So, um, I, I was working with a bunch of freelance writers who were creating content for our, uh, for our clients, and we were really leveraging, uh, how can we leverage AI to, to, to speed up our work. Improve quality. So to, to improve sentences, to improve structure. And after that, um, Che PT got released, I think a year or a year and a half after the start of that company. And I really got a choice, okay, am I going to pursue a content writing company or am I going to adapt early because. The content writing industry is going to change massively with, with the release of chat GPT, and at that point I got an um. Opportunity to join a lead generation agency, and I decided to seize that opportunity also for myself to learn more about lead generation sales. Um, wasn't, wasn't my strongest suit back, back then yet, and this was a point where we, where I learned a lot about leveraging AI for. Agency operations and also for lead generations. So we research AI for lead, uh, lead research, um, copywriting, but also, uh, within the agency, uh, there were a lot of AI projects of, uh, for example, uh, classifying the incoming emails. So when I started at the agency, um, I think all account managers, it took around an hour to, uh, two hours. Per day for an account manager to go through all the incoming replies in calls, emails, and, uh, leverage'em and classify them and, um, and, and tell, uh, and send it forward to the clients. And they built an, uh, AI system completely, uh, trained on the data that that, uh, was generated where we, uh, where AI took all of that effort away and suddenly everything happens automatically. Uh, leveraging made of. Com leveraging, um, uh, uh, GPT back then, gt GPT-3 0.5 still, I think, uh, back then, so this was already, for me, a massive eyeopener on how, yeah, I important it is, uh, to leverage AI within a company, but also for me, it's, it's really opened my eyes on how, when you're an AI builder, an AI systems builder, um, how much influence you have on the output and what that output actually does, uh, for in, in business practicality. So for example, uh, if we're looking at this, uh, email, classifier ai, uh, we could, we could make a choice that like 98% accuracy, but not a hundred percent. And that's usually the case with ai. And we could choose either, we would send. False positives or false ne uh, negatives. So either we would send too many leads to the clients who, who weren't really leads or you would, uh, keep le uh, you would wouldn't send them over because it were false negatives. And that's already something that's, that really opened my eyes as, oh, as a AI builder, this is so important to think about when you're building systems, uh, the choices that you make will have an influence on, on what happens as an outcome, and will also have an influence on either your clients, your customers. Um, so after that, I, uh, decided to go on my own again, and that's when I really started also doing the AI automation and implementation services for businesses. Uh, my boyfriend, um, was, uh, building an AI agent builder. So we really partnered up, uh, where I was doing a lot of implementations, uh, using his software. Uh, and the moment that client at any. Uh, questions or requests, his development team could easily implement it. So that was super convenient for me as an AI builder, of course. And, um, back then, um, yeah, we started building, I, I built all different types of AI agent systems and automations. So, uh, it, it's went from indeed, uh. AI lead research to, uh, slack AI agents managing, uh, clients in Slack and, and, and, uh, processing requests in Slack, uh, WhatsApp AI agents. And one thing really, uh, got requested more and more often, and I had. No experience with it at all, but this was, uh, Instagram AI appointment setters. So at some point I started doing, uh, I, I got one request of a client said, Hey, Carolina, is this actually something that you could do? And what turned, what, uh, what turned out? And I wasn't familiar with Instagram at all. I I, I have a very small personal account and I, I, I don't, I, I'm not that much on it, but what, uh, apparently a lot of Instagram coaches, uh, they get. So many requests during the day and they, um, really have to manage a lot of dms to, um, get people on calls and actually now close them into one of their courses, of course. And apparently a lot of in, uh, Instagram coaches, they work with, uh, so-called appointment centers and they pay thousands and thousands and thousands of US dollars per month, uh, to these appointment setters who are managing the dms and who are as a human, responding to all inquiries and really. Building relationships and kind of pretending like they are the coach. So it's, it's really like, like they're the coach. And what happened is, the first client that I got, we turned these appointment set into ai, and AI started managing the conversations and they got an increase of 40% booked calls, um, after the AI started to, to managing those cems. And why, because, um. The AI didn't get tired at all, so it got it kept on pushing and, and being enthusiastic, and it gave everyone a fair shot where you have, as a human, you have bias. The AI responded immediately, like we make it human so it responds really, uh, with 10 to 20 minutes to make a little bit hard to get right. But, uh. Uh, it's, it, it, it's continuously response and it continuously engages. So this was super, super cool. And, uh, based of that, you, uh, I got more referrals and uh, uh, this is something that's, the software is now also really turning into this Instagram AI appointment set software because, uh, there was so much requests for it. So that was super cool. Um, but. I got most of my clients through, uh, LinkedIn, and I always got when I, uh, also from my content agency, um, then for lead generation I did, uh, uh, I also helped them with ghost writing. So, uh, there we also got a lot of, uh, inbound clients, uh, through LinkedIn and, um. I have my own LinkedIn system because I've really noticed that, um, yeah, being consistent, you know it as well with LinkedIn is, is, is so, so, so difficult. Uh, you, you, you need to keep up, you need to come up with fresh ideas. So at some point I started. Started building a system and a knowledge based system on, um, the knowledge that I created so far. So YouTube videos and, and so forth to turn that into, uh, into more engaging LinkedIn posts. And I got more requests from clients from, uh, implementation services as well for LinkedIn implementation services. Uh, so LinkedIn AI systems. And recently I got, uh, also a client who wanted to have a LinkedIn system for multiple team members. So, um. To really amplify that, that LinkedIn voice to not have only the founder, uh, posting. But what if we multiplied with, with multiple team members who are all having their own voice, you will get 10 extra results, right? Um, so I built in, uh, no codes system for them as well using, uh, Clickup and uh, relay app. Which I really, uh, like as a simple ai, uh, builder tool. Uh, and then I said, uh, to young, uh, my boyfriend, I said to him, yeah, I'm actually thinking that I, I want to start selling this system. This system really works. It's really good. We can, um, uh, yeah. It's, it's, I think it's super, super high leverage to, to have businesses, uh, help businesses to start creating content also for the team members. And then Jan said to me. Kalina, what if we would turn this into a real software and real AI agent that can start learning improving and more and more and more, and that we can really help a lot of businesses with that. So that's how Post got born. Um, uh, a couple of months ago we came up with the idea and we started building and we started improving and um, we just released it. Uh, week and a half ago. It's two weeks ago maybe when this podcast launches. I'm not sure. Uh, but it's, uh, yeah, it's, it's so, so, so cool to see the first results already of people and, um, uh, and, and seeing LinkedIn profiles and LinkedIn content all my time timeline, which, which I know that it's generated with, with Post already, so it's awesome. That's how IV is born. Really an AI agent that helps businesses to turn their existing knowledge into high quality LinkedIn content to really attract leads just like I do, uh, do just like my, uh, clients, uh, do and now hopefully so many more businesses as well.

Audrey Chia: 11:21

Wow, what an incredible journey, Carolina. I think it's really cool to see how, even though you have transited from different spaces and like industries, right? There's the same enthusiasm and passion about AI in solving problems, uh, in helping people address really real world problems, right? With very smart solutions. And I'm curious to know, do you always have a technical background? Do you have to pick up some new skills? Like how did you plan, you know, that technical expertise as well as the marketers had?

Carolina Posma: 11:48

Yeah, that's a great question. So. I would say, so I'm, I'm a bit from a, a technical family, so, um, my, uh, brother's also a software engineer. Uh, all my, uh, brothers have three brothers, so all of them, uh, know how to code a little bit, even if it's for statistics. And I was always more like, uh, instead of doing more technical study, I, I studied economics and law, but from the backgrounds I always had. Still have had a technical mind, kind of. So I really learned from, from young, from a young age to to be a system thinker. Um, so that's how I would say I have this combination of both system thinking and marketing and more, uh, yeah, social sciences, uh, so to say, or marketing business sciences. Uh, but I did, um, learn. Python in a little bit, um, just to learn it better how computers work and how, uh, it works. But still, I am, I am a no coder, so I, I like to think in systems, but, but really writing the coding language, um, it's like I, writing code is not, not my thing, even though I do get help from cloth coats sometimes little bit if I need to, uh, to build some stuff. But yeah, that's

Audrey Chia: 12:55

it. Oh, super interesting. And I think it would be cool to understand from. Your perspective, right? What is that systems led thinking that helps to you to, you know, create your own processes? Because I think for many of us, let's say from a marketer's perspective, you normally, your brain just operates like a, you know, like a, like you're just blending different. Pieces of information together. Right. Um, but I found that when I use ai, it tends to work best when I very, uh, give it very precise step-by-step instructions and break down the way I think. So maybe for someone who is still trying to figure out how to work with ai, do you have any tips on how to think about it? How to break down your steps? What is, how, how do you approach it?

Carolina Posma: 13:35

Yes, absolutely. So what I recommend is always start with a human. So, uh, a lot of people. Uh, try to outsource their thinking to ai and they ask, okay, I want to set up this process. How would I do it? But, uh, start with doing it as a human, either yourself or one of your employees. And then the moment that you feel like, okay, this is working, and the process that I have right now is working, then you can start, thinking In more of a system because usually you will see a system, uh, the moment that you are doing something more than once, it is a system already because, or at least a process. Because, um, if you're doing something more than once in the same way, then um, you are already having a process in a moment that you have a process, you can start automating and you start, can start using ai. So start. Doing everything as a human and then, um, turn that map out what you actually have been doing, and then you can build a system around it and, and, and, and, uh, automate a part of it.

Audrey Chia: 14:32

That makes perfect sense. I think I've also found that when I don't have an objective and I ask AI to just, you know, um. Chat with me Sometimes, although the ideas flow, you tend to take longer to get to the end result. Whereas like if you already know very clearly what are your thought processes and then you break it down into steps, it helps AI a lot and maybe from your perspective. Right. I would also love to understand because. Many people are using the term AI agents, but I know not many people are using it as accurately. Could you explain to our listeners, uh, what are AI agents and how do they work, you know, in your company, and what sort of tasks are they doing and how do they interact with each other? Yes,

Carolina Posma: 15:09

great question. So, in my opinion, and, uh, and there is not a set definition for it, so that's, that's, that's one, uh, thing, uh, that's very important. But in my opinion, an AI agents, um, you can call it an AI agent, is, uh, if it's kind of an AI that has decision making power, uh, it has access to, um, different tools. And that means, what do I mean by tool? The tool can be, for example, uh, searching the web, uh, or it can have access to your systems, like to your HubSpots or to LinkedIn or anything. So it has access to, uh, to different tools. It's, again, has a decision making power to decide when it'll use those tools. And it's, uh, must have some sort of memory as well, usually where you, um, uh, yeah, where it. Um, yeah, refer back to, for example, either when we are talking about an AI appointment setter, uh, on Instagram. Refer back to the rest of the conversation and also get back to that. And based on the rest of the conversation, uh, uh, make decisions. Um. And sometimes it'll also has have access to knowledge. Um, and that means, but that usually that's, is more technical, but that is actually one of the tools that it has no, uh, access to. So in simple terms, it is an AI that can do stuff, but also decide when it does stuff and. You can activate an AI agent using chats. Um, then, then it's, it, it's a form of an AI agent, but it can also be activated, for example, on, uh, time, uh, on a certain time, for example, daily at a. Uh, X time, I want you to do some research for me, or it can be activated. For example, when a lead comes in, then you activate the AI agents. Um, some people say that an AI agent is only an AI agent, um, when it decides when to start on its own, something like that. Like when it says full autonomy and like that. My response to that is, as a human, you also don't, you als always have a trigger as well. An email is coming in, um, or, oh, it's three, three o'clock. I, I am, I'm going to get my, my mid, uh, my midday snack or something like that, right? There's, as a human, you also are not. Fully autonomous, but then, then we're going to get really philosophical. But, um, that's why I would say that an AI agent is really, like, it needs to be triggered by something. Mm-hmm.

Audrey Chia: 17:25

Does that answer your

Carolina Posma: 17:25

question

Audrey Chia: 17:26

well enough? Yes. Interesting. And I think one follow up question that people might have is what would then be the difference between an AI agent versus a series of steps that, you know, uh, where you're linking different, for example, chat, GPT, but different tool, but in a very straightforward sequence versus using agents that can make decisions. Like when do you decide to use what? Great question. Okay.

Carolina Posma: 17:47

So, um. When do you use, um, a series of steps, uh, or in other words, uh, words, a rule-based workflow or when do you use an AI agents? And I would say that maybe right now in 90% of the cases, you would use a series of steps. Um, you, you, you see a lot of. People and a lot of hype on LinkedIn, especially saying, oh, this AI agent does this and this AI agent does that. In a lot of cases, you actually don't really want that, uh, AI agent to have the decision making power. So, um, usually most cases it's an, it's, it's an, uh, workflow with steps, but with AI elements in there and in some cases, so when do you want to use an AI agent? When you want, of course, to have the decision making power, so very clear example that, uh, is customer support. Um, customer support. You want the AI agent if you're chatting with an, uh, with the customer support agent. So the human E or ai, um, it's super different every time there, there shouldn't be a rule-based workflow. So, um, the AI agent suits. Answer and AI agent should decide, okay, now I'm going to update that subscription, or now I'm going to send the refunds. And that shouldn't happen every single time. The AI agent should have the decision making power. Um, so that's when you want to use an ai. Uh, the same goes for our. Post if AI agent that we're creating. Uh, right now we are in, in, in version one, the first version of our post AI agent. And right now, um, the agentic, uh, element in there is that, uh, it decides whether it needs more knowledge or not. So we are really knowledge based, uh, LinkedIn content creation, which means that, um, it searches in your knowledge base, uh, on, on, on relevant information. And then it decides, okay, it's. Are we going to, uh, am I going to search for more knowledge or do I have enough knowledge to actually create this, uh, create a LinkedIn post of quality? And one of the elements we also have in our, um, uh, impulsive is a carousel creator. And what it does is it's, uh, dis carousel creator. It's. Uh, search is the web, for example, for logos of companies. If you are mentioning a tool, uh, uh, tools in your carousel or search is the web for images, and that's also an agentic. Uh, or you also want an AI agent for that because it needs to decide if it has enough information and you don't want to build a flow for that because then you get a very. You can build almost everything in the flow, but then you get the flow that's like the size of the world because it's, there's so many different choices that need to be made, uh, made. But in most cases, I would say it's still an automation that would work way better because it's more reliable. Because in AI age it can also make bad decisions sometimes. So that's something to be aware of.

Audrey Chia: 20:22

Well, I think that's a very interesting distinction, and I think most people maybe at a stage where they're still exploring. So for example, in like a marketing use case, right? There are many AI agents out there that talk about, you know, helping you to, um, ease the workload in marketing. But then again, I think it always comes back to what's the problem they're trying to solve? Because if you just use the agent for the sake of it, again, it you're, if you don't have the final objective, you may not know what's the best, uh, tool or the best workflow tool to. Work towards that solution. Right? And it just gets like, okay, I'm testing different tools, but you may get lost in the process.

Carolina Posma: 20:55

Yeah, absolutely. This is something, this is also why, um, I say always start with the human and, and really do it yourself and then search for the right tool because one of the questions that I also often get is, uh, I get the question. Okay. Uh. What's the best AI agent builder, or what's the best AI agent tool? I don't know. It depends on what you are looking for. It really depends on your use case. And, um, there's, there are a bunch of really great AI agent tools, um, but so different for from each other. And one is specialized in voice ai. Uh, the other is specialized in indeed marketing, ai, some, uh, tools. You also have a distinction between, uh, tools that allow you to build AI agents and automations and workflows yourself. And you have tools that, um, are an AI agent doing research for you. And the only thing that you need to do is just use the tool, kind of, and don't think about it too much. But then you need to have a tool that is really, uh, relevant for your use case. Because if you are. Buying a LinkedIn AI agent, but you are not a B2B company and you, you don't do anything on LinkedIn yet. That doesn't make sense even though the tool can be brilliant, but it's really needs to match your goals and your use cases, and that's something that only we as humans can really think of.

Audrey Chia: 22:04

Definitely, I would love to talk a bit about lead generation, right? Because it's interesting that you're using ai, um, and you have used AI for lead gen and also for creating content. Now, in your opinion, right? What do you think is the difference between content that actually brings in leads versus content that doesn't? Oh, great question.

Carolina Posma: 22:22

So in my opinion, um, content that brings in leads, so. Well, I think it's when you're creating content, you have different types of content that you, uh, can and should create. Um, the first type of content is kind of, uh, viral content or content that gets you eyeballs that usually doesn't really get you leads because it usually just gets you a lot of like, and again, everything is fluid. A verified or post can get you a lot of leads. It depends a bit. So that's are usually like, oh, these are cool tools, or something like that. And then it's, then it's, uh, it can go viral because a lot of people like it will get a lot of eyeballs to your profile. And your profile should then again, be co converting, uh, to, towards whatever offer you have. The second type of post is, um, more of authority taught leadership posts. And these are posts where you're really showing your expertise, um, showing, okay, this is. Uh, yeah, it, it can be very knowledge based. So this is also what we're really focusing on with post if so, uh, showing, um, if you have, for example, different courses, it takes a snippet out of your course and really educational and, and really, uh, yeah, educational content. And the last one is more, uh, the third type of content is more con conversion content. And that's really showing. Uh, client case studies, testimonials, uh, that kind of content. And that usually doesn't get many likes, but it does get clients because people see, and oh, they recognize themselves in the client that they're seeing. But again, um, and that's also one of the challenging parts of, of, of, of content is everything is really fluid. So yes, you can, in a viral post, you can also add the client case study, of course, and then it can go viral. So it's not. That black whites, uh, different. But, um, in my opinion, um, there are a lot of business owners sometimes that I see that are creating content that are really, for example, all around being like, uh, the entrepreneurial mindset, something like that. Yeah. I feel a lot of those, yeah. Entrepreneurial mindsets. Set and then, uh, but then they are actually, uh, selling like B2B logistics services, something like that. And then it's nice because they, they will get a lot of views, but it's, it's not aligned with who they are and what they're selling. So I really believe that every post that you're creating should have a big idea that you want to convey over to, to, to others that, um, that that is aligned with what you are selling. Because otherwise it's, it's cool that you have like the entrepreneurial Monday mindset, but, but. Nobody buys that, right? No. So that's, that's something to be aware of and that's why I think that knowledge and really showing your unique view on things that is will, is what what will get you clients.

Audrey Chia: 25:00

I love that you see that. And I have a very similar approach. So with my clients, I tell them about the marketing funnel. And the marketing funnel applies everywhere, right? Not just on. LinkedIn applies. Um, when you're running ads, you're, when you're running your email campaigns, LinkedIn is just one platform. So like what you say, you have the viral top of funnel content. More eyeballs, but not too many conversions. Then your mid of funnel education content. Uh, this is good for building authority and then your bottom of funnel conversion based content, which gets you the eyeballs, um, and it gets you the right eyeballs. Not many eyeballs. But the right people to see our content and convert. So I think this, this approach is something that a lot of US creators, founders have seen, worked very well on LinkedIn. And a follow up question to that, Carolina, is you talked a bit about, uh, lead gen and also having conversations on, you know, whether it's IG dms or perhaps LinkedIn dms. Right. And I think most people are not. Comfortable in their dms, um, having those conversations and they don't know how to turn the views into prospects. So do you have any tips for someone who has started posting content but is not too sure how to then navigate those conversations? Uh, what have you seen work and what hasn't worked?

Carolina Posma: 26:12

Yeah, great

Audrey Chia: 26:12

question.

Carolina Posma: 26:13

So, um, one of the things that's worked best. There are different things that were good for me. So one of the things that's, uh, good to be aware of is that you have many lurkers kind of, and those are people who are reading your contents, uh, but are not, uh, engaging with your contents, but. Those people do watch your profile often. So one of the things that I really recommend is, uh, going through the, uh, to your profile viewers and start sending them a message, um, saying, Hey, I saw, uh, uh, you can either say, I saw that you watch my profile, but you can also say, Hey, how are you doing? Um, and. Ask a question that is relevant for your surface. So, uh, in my case it would be, for example, if I'm reaching out to someone now for my, uh, post, uh, I'm asking, Hey, uh, are you, uh, getting to your business goals, uh, in terms of LinkedIn? Content generation sometimes that's not the best I think on that. But, um, that will open the conversation. And the way that you should see it, and you should see every conversation is, uh, you are not doing sales. You are helping people. So if you are approaching, it's like, oh, I need to sell this and I need, really need to, to, to, to get them to buy from me. That won't work because people don't, don't, don't like it. And they, they smell your eagerness as well. And that usually puts people off a little bit. So. What you want to do is, is try to open conversations and then either you can, there are different, uh, approaches in this. Either you can really push people towards a call and say, oh, what you are, uh, experiencing now. So you usually want to get people, ask people from where are they now? Where want do they want to be and what is has keeping them so far from getting to their points? Um. If you have that conversation and it's that clear, then you can approach, uh, uh, offer a call and say, okay, but that's actually, that's that I can really help you with. Uh, maybe we can explore it further in the call. And then during the call, you will again ask, Hey, where are you now? Where do you want to be? And what's keeping you and what have you tried so far? If it's, if the conversation is not happening like that, um, you can always just, um, leave a call, say, Hey, if you ever want to connect, just let me know. Happy to, to have a chat. And here's a link. You can freely schedule. And sometimes I, I have people who are suddenly scheduling a call sometimes months after I've sent that link to them, and they weren't ready to buy at that. Point months ago, but they are ready now. And I left an opening to them, uh, to, to just do it. Um, but yeah, remember you're not not selling. You are helping people and some people are ready to be helped with something. Some people not yet. Some people never, and, and that's fine. That's, that's, that's not the rejection of who you are or or of your business at all. It's just a rejection of. Where of the, of, of the goals that they have at that point. And that's the most important thing to remember as well when you're having chats with, uh, with people.

Audrey Chia: 28:55

Well, I think what you mentioned also, it's not just about the conversation, so a mindset, right? Like the ability to reframe their perspective from pitching and selling to helping someone to diagnosing their problem. I think that kind of mindset shift definitely helps in the dms. What I've also learned over time was in the past I used a very straightforward. Pitch that approach because I had no idea how to navigate, you know, the dms on LinkedIn, but over time you then realize that like what Carolina, you said, um, being able to have more organic conversations open, opening softly, but also teasing out their pain points really helps to then move their conversation to a natural kind of, um, call. And I think if you're on LinkedIn and posting content, you really need to balance out. Both posting the content to bring inbound needs, but also doing the active outreach, um, to see what has worked best. And I'm curious to know, Carolina, since you have worked with AI in so many contexts, right? You said legal, ad tech, uh, SEO, even, what do you think are the biggest patterns you have seen with ai, um, over the years? Like basically, you know, are there any big changes you have noticed or any big wins that you have seen for companies?

Carolina Posma: 30:04

Ooh, big changes are big wins. Yeah, of course. A lot of big changes in the sense like of, of the quality of ai. Um, so what I am, I am a huge, huge, huge fan of Cloud Force sonnets. Um, that's, for me, that was the biggest change. So up until cloth for sonnets, I. Wasn't too happy with the output of AI in content. I wasn't too happy with the output of AI in, in conversational ai. It's, it was good. It wasn't brilliant. And, um, when Cloud four sonnets got released and months ago, um, I. Suddenly there was such a big shift in, in, in, in output quality of, of, of content. So that's why we're also using that in Post-It, for example. Um, but also in, in conversations as, as I mentioned for example, the appointment set when we are using cloud for, so it's like, it's so human, like it's you, it doesn't use the M dashes for example, which is something I'm very happy with. Um, but um, I think, I think, yeah, that's one of the big changes. Um, uh. Does that answer your question enough or,

Audrey Chia: 31:10

yeah, definitely something else. Definitely. I think that was a really interesting point, right? Sometimes people think that all AI is the same, but you know, every AI has its wants, its quirks. Some are better at copy compared to others. And maybe what is like the biggest kind of ROI that companies have achieved, you know, over the. The past few years, um, from working with them, what have you seen companies manage to achieve in terms of like time or productivity saving or even cost savings even?

Carolina Posma: 31:34

Yeah, good questions. Yeah. For example, I already mentioned the 40% increase in, in, in, in, uh, uh, calls booked, um, suddenly. Um, yeah, getting, getting a lot of leads through LinkedIn because you start posting on, uh, LinkedIn content with multiple team members, for example. Um, in terms of yes, specific costs, uh, I have, uh, clients who also was spending, uh, hours a day to process diff or their account managers were also, uh, uh, to process different client requests. So, um, this is an, uh. Kind of marketing agency, but they got a lot of client requests that, uh, needed different invoices, uh, kind of complicated process. Um, but we, uh, reduced that process from hours a day, uh, to seconds. And every time one of the clients asked, uh, uh, asked a request, the AI agent took it on, uh, processed everything and got back to them in seconds. So that's also a huge, huge, huge time saver. Uh, another company that's. I helped, uh, got one to three extra leads per day because they added the websites chat AI agents, uh, on their, uh, on their website, which is also really awesome because that AI really started because it was a really low barrier, because first they only had to form to, uh, to request a kind of funding, uh, stuff and. Uh, to request funding and financing, and it was kind of a big barrier. So we added a, a low barrier AI agent to their, their website, and people started asking questions and the AI agent engaged with them, answered any concerns they had, or any questions, am I good enough or am I a right fit? Yes, you are. And uh, yeah, it got them to a lead. So there are so many, so many, yeah. Amazing, amazing examples of, of AI really making a huge difference for companies.

Audrey Chia: 33:18

Wow. And I think like a lot of these use cases companies can implement straight away, right? There's so many things that can already be operationalized. I was having a chat with another founder the other day and he was sharing how some of the best use cases are the most boring ones. Um, whatever. That seems boring. It can be done with AI and it's not fancy or it's not new, but it works and it works. Wonders in terms of saving the team time and freeing their time up to do more higher level, you know, pieces of work. So I think that those are incredible examples. But maybe on the flip side of things, Carolina, what do you think are some, you know, things that companies should look out for when trying to implement ai? Because I do know that there are certain challenges, limitations, and drawbacks as well.

Carolina Posma: 34:04

Yes. Okay. Uh, I would recommend to not automate anything fully with ai, like, um, having no human in the loop. So, for example, within post, um, we still have a human in the loop. Uh, even right now we are working towards a fully autonomous AI agent created content for you, but probably even then, we will still have a human, uh, want you as a human to, um, approve. Edit and, and, and check it because AI doesn't have the human empathy. So, um, it's can write something based on your knowledge, which is for all criteria. It's, it's, it's a perfect post, but it, it might lack some, some, some, some empathy that you would immediately spot as a human, but not, uh, if, if it's fully autonomous. And the same goes for. Um, and the, the, the appointment set that I talked about, I always built in, um, escape. Possibility for the appointment setters that if they, the AI doesn't know how to handle a situation, um, it gets access to a tool. And that tool is pretty broadly defined. It's just Slack notification, but it's pretty broadly defined. Use this tool whenever you can't, uh, can't fix or can't help, or if you are insecure or if you feel that, um, someone gets. Angry or unhappy or whatever, uh, whatever. And it can turn itself off and it notifies, uh, a human to actually take a look at the conversation. And that's so important to always, always keep some form of human in the loop. So for content, have a human approved and make some final edits, um, for appointment setters. Um, give an escape to the AI to not, um. Um, yeah, not have to manage or go, go to their goal no matter what. Because there are also examples of, of AI really now going rogue is, is, is a big, big term, but, um, clot or tropic the company behind Clot did, did a very interesting research where they taught and AI to, um, get to their goal. No matter what. And what they did is they gave their AI agent access to all email inboxes and everything, and they said, you need to get to your goal no matter what. And then what they simulated was that, um, the CTO of that company told their, uh, told the AI agents. Um, you can't do it, or it started, they started blocking the AI agent, and the AI agent got, uh, thought no, but I need to get to my goal. Well, no matter what, so this CTO, he's blocking me. So what does this AI agent do? It started researching and, uh, the email inbox of the CTO found blackmail. So he found, uh, someone like an a blackmail something and he started blackmailing the CTO saying, you need to give me approval because, uh. I need to get to my goal. And this was all an a simulated environment, but this really, really shows that you need to give your AI an escape opportunity. So not go to your goal no matter what. No, go to your goal, but if you can't reach it, do this. And otherwise ai Yeah, that's, it was super interesting. It was so, yeah. So.

Audrey Chia: 37:01

That really sounds like the next movie Blockbuster about thinking over the world. A little bit scary, but it's really real, kind of like real world situations where we then have to be aware of, you know, what tech is capable of and also as a human or as humans, guarding the tech, what we want to do with it. Um, and that's so important. Yes. But thank you so much for sharing your insights, Carolina, maybe to end things off, perhaps you could. Tell our listeners, one tip, you know, for listeners who are trying to adopt AI agents or trying to, you know, create content on LinkedIn, is there one tip you might have for them?

Carolina Posma: 37:36

Well, of course, let's try post if AI to create your content. Um, but also, um, if you're trying to create, uh, content on LinkedIn or, uh, building AI agent, just try and start, um, action over perfection always. So do something, try it. And, and even if it's, it feels, it never feels really because you will learn from it and you can improve, and that's the best way to get started with anything.

Audrey Chia: 37:59

Awesome to hear that. So Carolina, where can people look for you and who should reach out?

Carolina Posma: 38:05

Yes. Uh, so yeah, you can find me on LinkedIn, of course, uh, Carolina Pulma. Uh, so feel free to send me a DM if you're curious about anything that I talked, uh, talked about. And of course, you can visit a PostIt, do AI to, to start a trial and, and really try AI agents for your company and your team.

Audrey Chia: 38:21

Thank you again for sharing your insights, Carolina, and thank you folks for tooling in. Don't forget to hit the bell for more actionable AI and marketing insights. We'll see you next week.