Hey everyone, today we diving into an exciting and a highly relevant topic, is implementing AI agents in marketing. And our guest today is a seasoned expert in AI-driven marketing. He first discovered the power of AI through a company called Black Crow. And today he's here to share valuable insights on how agents are shaping the future of marketing. Hello, Frank. Welcome to the show. Thanks for having me. It's good to be here. My pleasure.

Yeah, let's start by getting to know you a little bit better. So can you share your journey into the world of AI and marketing? Like how did your experience with Blackrow shape your perspective on AI?

So yeah, myself, I've been sort of, you know, in sales for well over a decade. And then, yeah, at some point in my journey, so after working for companies like Google or, or similar web, then I got into a company called Blackrow. Blackrow is a, company that's in the machine learning space. And essentially Blackrow allows you to predict in real time, the probability of a user doing something. So for example, buying something on your site, doing a particular action, and then that real time intent.

You can use slides in many different ways, whether you're passing that piece of data, let's say into Google analytics or Google ads account. And then you can essentially bid based on users that are more likely to buy your product versus those that are not likely to buy the product. Now, the way that the machine learning just generally compares is if you would look at sort of the world of marketing or advertising. So a lot of marketing folks have been using what's called rules based, let's say, segmentation or targeting. So let's say any person who was on the cart page on my site.

let's say from e-commerce store, then I should definitely retarget those users. And that's a rule that you set and you can do a bunch of different rules, right? And then the way that that compares versus say machine learning, doesn't matter where the user is on your site. You can predict the intent on any of the pages, which essentially is a lot more powerful and typically drives a seven finger bottom line improvement to your business. If you like a multimillion dollar, let's say e-commerce store or whatever other website, essentially the more traffic you have and the more revenue have you on your side, you should definitely be using sort of machine learning. So yeah, that's where, you know, I spent a fair bit of amount of time in my career, essentially talking to a lot of CMOs, VPs of marketing, especially different businesses, whether that's, you know, these are airlines, OTAs, e-commerce sites, even SaaS companies, you know, when they had a lot of traffic, then targeting the right users based on the intent was very, very powerful. And if you spend like millions of dollars in, in ads with Google or Facebook, et cetera.

then optimizing based on these sort of audiences was super, super powerful. Awesome. I guess you've got an ample amount of experience with AI already. Looking forward to learn more on this topic today with you. Awesome. To set the stage, I would like to ask this question that's been on my mind for many months now. So like, how did we get to the point where AI is becoming, you know, such an integral part of marketing? What's accelerating the need for AI adoption in marketing today?

So I think if you look at macro level trends, so for say, for example, if you look at the CAC inefficiency, so CAC means cost of customer acquisition. If you look at a company's ability to acquire net new dollars for their business is becoming more and more expensive. It's literally kind of sort of going through the roof. And essentially for some companies, it's just not very efficient to be acquiring these dollars altogether based on,

the old school sort of processes that they had in place, whether these are rules based or not, doesn't really matter. But ultimately acquiring that new doors is becoming more more expensive. yes, so then introduction of LLMs, whether that's by opening it, et cetera, allot the whole new world and possibilities ultimately to keep your headcount flat in your organizations, reduce your headcount potentially, and supplementing essentially your current headcount in marketing and sales with some form of LLMs, right? So the ability that you, where you

can sort of programmatically provide some sort of into an engine or context into an engine and then you get some sort of output, right? So this is what a lot of marketing people do. So an example of that will be an ad copy. So ad copies used to be written like heavily, heavily by folks in marketing, fine chain optimizing, and it was a very repetitive work. used to outsource that to an agency and that used to also cost a lot of money. But these days there's a bunch of tools in the AI that use these LLMs off the shelf that you can just plug in and then write beautiful ad copies. even compute beautiful creatives even like, know, which are all done by AI. And nowadays what you're seeing is essentially these sort of fake influences as well. I've seen a lot of companies from Y company to popping up these days. I would say, Hey, listen, you can generate this avatar, which will be essentially a fake person of your business. And that avatar can drive millions and millions of impressions.

And a of people on the web don't even know that these people are not real. That's essentially the world that we are getting into where, um, uh, these sort of, all these sort of AI capabilities unlock humongous efficiency levels for companies, which helps these companies to become essentially super lean and super mean. So what I mean by that is, let's say five years ago, right. When we didn't have those, this sort of wave or LLMs, um, when you look at companies, let's say from San Francisco, even in Europe, companies used to be much bigger than the headcount when they used to start these companies. And generally speaking.

companies used to have a lot more headcount, five years ago. But if you look at a lot of the stats in terms of how much, for example, companies make in terms of revenue per headcount, that increases exponentially actually. And this has to do when you're small, as I said, small companies, just, you know, companies are starting off all the way to big companies. So for example, in bigger companies, a big wave that we're seeing, across companies like Klarna, IBM, et cetera, where they're cutting down heavily on the support stuff, because essentially.

These companies use a lot of these LLM capabilities to, to, to automate a lot of the parts of business because people have a lot of very simple questions when they get, let's say, go into chat support or essentially in support, generally, if they write an email, cetera, and companies used to use human labor folks from out of university, et cetera, to answer these simple queries. But with, with the introduction of LLMs, you don't need all that labor anymore. And the beautiful thing with, with LLMs, the large language models is the fact that A, you can reduce cost.

you can increase the quality of the output and then you can make it much, faster. Meaning for example, the customer can get the response much, much faster. Right? So there's multiple different advantages of why companies are aggressively implementing these capabilities to ultimately increase their margins, right? Within their business. Typically, what is a threat just ultimately is the session wherever you have huge teams. in an organization that's sales, maybe finance, customer support, et cetera. And these teams do a lot of essentially repetitive work in those teams, then they're more prone to be ultimately what we call augmented, right? Blended with some sort of capabilities or elements to essentially do more with less. Less meaning less of a head count. And what we're gonna see essentially happening as we progress this year and also next year is ultimately redistribution of labor. So essentially people will be doing different jobs with a...

those jobs could be prompt engineers and a bunch of other jobs. We're definitely going to see more of those jobs available on LinkedIn, whatever those jobs are. So we're going to see a lot of that. And the way that I compare this sort of evolution essentially and the sort of transition that we're going through is very similar to how robotics wave came into factories right back in the days. So I'll look at that as there was a hardware revolution, right? So there are a lot of people working in the factories and assembling, let's say, cars, et cetera.

And then suddenly you have a bunch of robots today. We're going through a similar sort of thing, but in the world of software, right? So if people are in front of the computers doing something manually all day long, well, why wouldn't you know, any agent, et cetera, et cetera, and you'll kind of help you with a lot of these tasks. Now this is not to say that, you know, every single person will be replaced because there's still, you know, place for the human and, so, but I'm, I am very big believer in the fact that you will see a world where we will be blending.

humans and AI agents to ultimately drive more output for your organization. Let's say in the world of sales, which is where I'm at with a lot of lower cost per unit produced. So let's say in our case, you know, if we are in the business of helping our customers to generate a pipeline sales pipeline, and that typically means meetings, that will help you to generate more meetings at a lower cost per meeting. Right? So this how we are thinking at the back of our minds. How can we help our customers to do that? And AI agents or LLM is,

part of the puzzle to achieve that, but it's not the only thing that needs to happen. There's also other things that need to happen. Essentially, you need to orchestrate, let's call it the software or the automation, which is the ultimate goal, right? Because essentially, an AI is just a means to an end. But an automation is what everybody kind of striving towards, having essentially be, know, leaner and meaner business. And that can come in sort of many different shapes and forms. So for example, I'll give you a couple of examples. So we have an off the shelf solution. So we have an AI agent called Agent Frank. So you can essentially spit up an agent and literally relatively super fast, get it up and running and help you generate essentially additional meetings fully autonomously. And most of the organizations in B2B, need to typically need to do two things. One is to build a great software or the great technology or great service. And then you need to sell that service, right? So we just help companies to generate that amount for the business a lot more effectively. Now that's one example, right? Off the shelf solution, you find it on the internet, sign up, off you go.

Now there's another type of sort of automation, which a lot of businesses sort of should be introducing to keep the business sort of leaner and meaner. And this is essentially where you look at sort of, know, your internal business process. So for example, what is it that you're doing something internally manually? This could be, for example, collecting invoices, processing invoices, doing reports, et cetera. So whatever you're doing manually within your business, you should sort of stop doing it. Like, because a lot of these things can be automated. So how could that be done, practically speaking?

is let's say you would implement a software for example, make.com is one example of a software or another software which is slightly more advanced, which is N8n. N8n essentially these are like these beautiful sort of workflow tools where you can literally automate a huge, huge chunk of your business. And you can plug in an AI agent and LLM and many other things, right? To automate and ensure that people stop doing repetitive things, right? So.

So think of kind of back of your mind of your business owner, your sales leader, whatever you are, or a marketing person, and you're doing something super repetitively every week or every day you're doing the same thing. That's a point to automation. It's always a question of back environments. Could we automate this task? Is this an end case, what we call? Because we do use NA 10 internally, for example. And for example, so our goal to become sort of leaner and meaner is to produce this year, 1,000 NA 10 flows. That means

We want to make this sort of business like a machine where everything just runs, right? And you don't have to worry about stuff. And if something's broken, then you may get a ping in Slack, right? That's the sort of the business, the sort of the feeling we want to have for every single employee that works here at the Forge. Amazing. That was really amazing. So speaking of use cases, like, do you have any standout examples where AI is making a massive impact? Yeah. And I think there's a lot of, a lot of AI agents that make a huge impact. think let's take one example, not from our space, but something else. I already mentioned customer support. The number one company in customer support today for now is Intercom. And within Intercom, they have this sort of AI agent, which is called Fin. So what Fin does, it looks at, for example, everything that's available in your knowledge base, right? Everything that you wrote up about your company, like the knowledge articles, cetera.

It uses those articles to help answer questions for the companies super, super fast. Literally within, you know, let's say a few seconds, right? Super awesome. Especially if you have customers, you know, around different parts of the world, different time zones, right? And your employees only work in the West Coast, US, as an example, right? And you don't cover all the time zones. So Fin, for example, could be covering non-working hours, right? Or if you have too much, too many tickets, right? You can enable Fin, et cetera. And the interesting thing about something like Fin is that Intercom will only charge you a dollar or 99 cents only when there is a resolution, when the customer feels like that the agent has actually answered the question, right? And actually incentivizes intercom to do a better job with AI agent, right? Because they want to charge more their customers, right? But also at the same time, what it helps that business, right? That gets a lot of these queries is ultimately reduce the pressure to hire more humans. Because if you then compare, say the cost of hiring a rep in the Bay Area, let's say hiring a rep in the Bay Area for your business and setting a link to come out and answer your question versus using an AI agent.

you will quickly see that the AI agent outperforms in a lot of the cases. Unless, so there's not actually definitely cases where, for example, AI agents are not very, very great, which is for example, if a customer comes to your business and says, hey, I have a bug with your product. Now that's not something that I've seen, for example, AI agents excelling, right? You have a bug in your software, right? Or something is not happening and AI agents are not yet sort of engineers, right? They can quickly go into the software, investigate what's happening, right? Ask the right questions and actually fix those bugs on autopilot. So I haven't seen that.

However, we do see already a world as of sort of, know, six months ago or so where there's companies like Cursor or companies like Loveable, essentially, there's a lot of different companies that probably help to 10x sort of engineers. So essentially you apply a lot of really cool capabilities on top of engineers or on top of essentially product managers that maybe don't know how to code, but essentially these tools help you to code and help you to code very fast. so we're going to see essentially a lot more code being shipped just generally by the current, engineering population. and what we're seeing right now as a, as a wave, I've seen that on LinkedIn yesterday is that number of job ads, for example, of engineers around the world, on indeed.com, which is the number one job site. And you can see, that's actually reducing slightly. there's essentially on, on average, essentially companies hiring fewer and fewer engineers.

And the reason why it sort of goes down is because ultimately the current engineering team that you have can produce more with the tools if you give them those tools, right? And yeah, think of it this way that, you know, if you have a, you know, a really kick-ass engineer, you know, he's a senior and he does a really cool job and then you give him a tool to supercharge him, right? So to maximize his output, that's a good use of your capital instead of hiring more engineers, right? To do more code, right? So that's the world we're gonna...

that we're going into, you have these sort of lean and mean teams and customer success, engineering, and in sales it's the same thing, right? So salespeople are super expensive just generally, like, and so essentially the other reason which, you know, which the other sort of trend that we're seeing why these AI agents or LLMs are coming into place is because also the salaries are rising exponentially due to inflation, for example, right?

That's another kind of pressure that's being applied on the companies because ultimately if I have to pay more to my employees, well, then I have to raise my prices for my software, right? Then which impacts my customers and ultimately that will impact the churn of the business. All so this is another kind of pressure that a lot of businesses are having. And if you just look at inflation rates, whether that's in the US or Europe, et cetera, and the fact that your pricing doesn't go up with inflation in the same sort of trajectory.

then businesses are under pressure to do layoffs, right? Because due to magical and economic impact, and this is why there's that huge demand for AI agents, right? To see whether companies can augment. So this to say, we definitely see a lot of interesting stuff already happening, right? Not like, you know, this is like a crypto wave or something, right? Where, you know, it feels like nothing has really changed with the world of Bitcoin, et cetera. Or at least a lot of people don't feel like a lot of things have changed or not yet, please. But with the AI, you can see already a lot of different places where there's a humongous impact. for example, we're seeing that across our customer base, right? When we have started to use, even LLMs before chat GPT came out. and then also seeing this in other places, you'd sort of transformation of companies, which is quite exciting. So I'm really looking forward to, for example, when Sam Altman just recently announced GPT five, you know, rollout and things like that. And that's it. That will be the final one apparently. So yeah. so we're to definitely see still a lot of change in the next few years to come.

And what I want to kind of say to everybody is that you have to think about your sort of like employability, right? So you have to think about, okay, with all these changes that are happening in the world of EIL levels, cetera, how do you ensure that you can just can't be replaced? Right? So that typically means, you know, know, know, things that have to do with, you know, kind of on the creative side, you know, let's call it content being on the podcast, et cetera, you know, all of these things will still remain, right? But there's definitely a lot of the jobs. If you're doing something like super repetitive.

then you should know that the agent is coming after your job essentially, right? And it will eat up. And ultimately that means there'll be fewer jobs in that particular function and there'll be more competition for the same jobs. And ultimately the companies will only hire the best people for those jobs. And I think that's fine because ultimately when I think about running my business, so we scaled, let's say from three co-founders last year to right now 37 employees in literally 12 months. I have a very hard time to find really good people for this business.

And so what that means is, yes, there's a lot of people in the market, but they're not really good for our business for different kinds of reasons. Right. So this is why I'm also forced to, okay, if I can't find really good people for my business, I'm also forced at the same time to, build these AI agents, right. To supplement my good people that I currently have. But yeah, so you know, kind of in my mind and, and, know, again, if, if somebody's a business owner, et cetera, or if you have a, you running a large team, if you feel like your head count is going through the roof a bit, right. Because you're growing very, very nicely, which is what's happening in our case.

Like you should kind of think about the same time. Wow, we're growing so fast and the headcount is massively increasing. You have to ask kind of back of your mind, are you kind of automating any of the things that you're doing internally, right? Are you implementing AI agents? Are you implementing any of these automation tools? Because if you're not, trust me, your competitors are. They're thinking about this. They're thinking about improving profitability levels, et cetera. They're thinking about how to make sure that you're leading the mean. And the VCs today, so investors are excited and particularly about companies that grow in revenue like massively, but the headcount stays low. Right? So think of that as well. Right? So how will the market perceive your right efficiencies, et cetera. And ultimately the perception today is it's not anymore a growth at all costs. Right? Like it used to be so two, three years ago, but it's about efficient growth and like efficient growth in simple times could also mean like flat head, come with revenue, countries keeps going. And then every employee in the business should be incentivized to think in that particular way.

Awesome. So we spoke a lot about automation now. I also like want to touch upon how AI is influencing the strategy strategies in different departments. So are they just about efficiency or do they also help drive revenue growth? It's both arguably. Now in some places, you know, there's some functions, let's say customer support, right? These are not typically revenue generating functions or engineering is not a revenue generating function.

However, when you look, speak to your CFO, know, sometimes you may do speak to your CFO, whoever owns the budget for the company or financials, typically they look at the largest cost centers in the company. So means where do we spend the most of our money? Right. And can that be optimized? Now, when you're in marketing and sales, it's all about, you know, hiring the best people so that we can have the greatest output, whether that's MQLs, know, leads, demos, whatever that is, right? Hiring the best people.

because they'll bring the best output, right? So, however, I mentioned for hiring good people is really, really difficult, but let's say I will hire one really great individual and I can't, can't really find other lookalike great individuals, right? So you build like a bigger team. yes, at the same time, the strategy by default means our hiring strategies like humans plus agenda capabilities. And I think that's the mindset that we're in right now, implementing the, the kind of the dual strategy of ensuring that

You know, like maybe the other way to look at it is like, let's say we have a pain in the business, right? And we think that the pain should be solved in the business by hiring somebody to solve that pain. Right? This is how companies used to think. Now, the way that we think about solving pains, it's first of all, we think, okay, can AI agents solve this, you know, as a first thing back in our minds, then if not, okay, if an AI agent can't solve this because there's no company, let's say that popped out out of YC that solve this for now. Cool. No problem. Can some sort of automation solve this? again, n8n, make.com, can we sort of automate this somehow? So it's kind of mimics an AI agent in one way or the other. Okay, not an option, no problem. Now, as I said, hiring humans is very difficult. Then you start thinking about human options. And the next human option that comes to mind is outsourcing to a freelancer, an agency, et cetera. Because freelancers, et cetera, they specialize in these particular areas to solve that pain.

But to hire an individual sometimes to solve a particular pain is like super like impossible essentially. So I'll give you an example in our case. So we had a pain that we needed to produce content on our site, which would be really well SEO optimized for our niche. Right. That was the pain that I had. And I was thinking, do I hire a person? Can AI agent build content? No AI agent can't. Can I do it via NA 10? No, I can't. Then the next step is like, all right. So it's like, you know, the human question. Okay.

outsourcing or hiring a person in-house. I scoured my whole LinkedIn, I could not find a person that could be really good for this particular problem. And then yes, an agency typically is the third option I would say. find a specialist, a freelancer or agency that could really do that work. And then the last resort is the human really, that you hire into your business. So this is how I think about things these days. And I think this is how most companies should be thinking about if you want to be lean and mean.

Awesome. And you also mentioned about how companies should adopt AI agents. are there multiple types of AI agents? What types do you see in your everyday? And can you share some exciting examples from that? Sure.

So, so one thing I mentioned, these are, there are these sort of let's call it off the shelf solutions. So one example of that is let's say our own agent, right? So I'm gonna give you an example of that. So what our agent does, this is especially, this is important for companies that want to reduce their CAC that, or maybe that they're lazy when it comes to sales. Like they say, Frank, I don't want to do any, any sort of prospecting pipeline building. They're just lazy and it's fine. You know, there's a lot of people that lazy. that's why we have in the business where we help sort of these lazy people. The other people that we help are that are not competent in sales. They just don't know how to build pipeline. That's fine. So you can kind of learn everything yourself, how to do it and become a great salesperson, do prospecting, whatever, all that stuff. but if you're not great at that, maybe you want to use this, you know, an off the shelf solution, you know, an AI agent that could do it for you and it caters for, for most of the kind of use cases. So what is the main use case? You know, a typical B2B company. Well, you need to build pipeline. You need to get these meetings. You need to close these deals, right. And then.

This is how a company typically grows, whether you're a B2B SaaS company or some other B2B company. Now, what the agent does, which essentially mimics a sales development rep or business development rep is we set an ICP within the software against that ICP. We can either source the leads against the ICP across multiple databases in the world. Or there's a second option that a customer can just literally pick a list from the CRM and we will just process that.

then what the agent does, what a typical rep has to do is really craft a great email. And if somebody doesn't know what a great email is, just go to salesports.ai and there is a section which says experience agent frank. You can literally just pop in your email, pop in your LinkedIn and you'll see an example of that email. But essentially a unique email is an email that's relevant, personalized to you. And we were essentially trying to match what is it that you're selling. So we call that the seller data or why the heck you're reaching out to get it with the buyer data. What do we know about that?

potential buyer based on publicly available data. So for example, we, that could be the website of the prospect. That could be the LinkedIn URL where we literally access in real time. We, we, fetch everything that we know about you as relevant to what is it that you're offering. We could check for example, what's in the CRM, you know, have we, did we have any conversation in the past with you, cetera? So then actually that is the context that goes into, into the computation. So the seller data and the buyer data, we, we marry these two data sets and we compute on top of that in any language, whether you want that email to be delivered.

in English, in French, German, et cetera, it's up to you. I always say, you know, do not send an email in English to a Frenchman, it will backfire. But if you do send emails in French to a Frenchman, guess what? A lot of replies coming back. So localization is like super also important to drive performance. And essentially these agenda capabilities make it possible. Imagine you would send an email to a person in a language that you don't speak. but in the IG can do that flawlessly for you. So it's not just one email that we send. We send a probably called campaign or sequence, which is three uniquely crafted emails to that individual. Essentially every interaction is unique. And then yeah, if a prospect ever replies back, then the agent actually steps in and within a few minutes replies back. So the reason why speed is so, important is because that user is currently in the Gmail mailbox, you know, answer to you and then say, wow, I've already got my reply back, et cetera.

But there's a bit of a ping pong sometimes kind of back and forth before you actually book that meeting. So speed is key if you like, you want to increase your meeting booking rates, just generally. This applies for inbound traffic or inbound revenue and also for the outbound revenue, same thing. And AI agents can make that possible because a typical human, so let's say you're on lunch right now, you get an email and then you only reply after an hour or two. Now that kills conversion rate. And there's a lot of statistical data to show out there that if you reply to your prospect,

after more than two hours, then the probability of the conversion essentially deal happening reduces significantly the more time passes. So this is why it's super important. Typically, there should be what I would say a two hour SLA on every reply from prospect if you're looking to maximize your conversion rate. Now in the IG can do that fully autonomously. And then all you do when in that conversation when it happens back and forth is to set a goal.

So the typical goal is a bunch of different goals that you can set for an agent. Let's say goal one, which is a tip, like a, know, a thing that I already mentioned is the macro conversion that we're going after, which is let's say meeting booked, right? So you're saying the agent, Hey agent, we want to book a meeting with every person that we're interacting. So you can hook up to your calendar link, your HubSpot link, or you can tell the agent, Hey agent, please. I would like to get their Cal link and then book into their link or their calendar. And the reason why this is important is because there's definitely a power dynamics, a lot of sellers out there.

say, hey, here's my link, I would really love to chat with you. But the customer is the cane. You should really ask for their time, their availability, their link. And this is one of the biggest things that from my experience, sellers fail at. They always send their link. No, always ask your customer, hey, do you have a link that I could use to book some time in with you? Again, an agent can do that. And then the other goal, so a lot of companies, let's say POG companies, don't want to have any meetings whatsoever, but they still want to use an AI agent to... to drive the revenue for the business. So instead of the meetings, a lot of them, they do with us is you can essentially implement what's called a URL within the AI agent, where the AI agent will say, hey, if you wanna check this out, hey, here's the signup page, for example. Essentially a particular section on the website where an agent will try and drive that lead to, and then I guess your site takes over and then your site converts. So that's an example where of...

off the shelf solution can do entry and sort of, you know, pipeline generation for a business. Now the, the other type of AI agents is where, as I said, you can do something like an eight N. You know, so you can implement the full automation. So I'll give you another example. So let's say the other way, how we do it leads for our business is when somebody sends me a cold email and that cold email lands in spam.

So the way that I have set up my, it's called spam box as I have cooked up my spam folder essentially to NA10. And whenever I get a cold email and it lands in spam, then I pass that email and the context into, the agent through NA10. Then we add, when the allies is this cold email, is this email a cold email or not? plan is spam. If it is a cold email and somebody's trying to pitch me back, then, you know, one of the other things that we have, you know, for every business is to improve their ability. So, right. So then what we do is we fire away back.

an email to that individual saying, Hey, it looks like your email line is spam. And it looks like, one of the reasons why it landed in spam, because you will send a template to me. You did not personalize, cetera, which is essentially the reason why a lot of companies get these spam reports. So, and that's it. So then we will fire an email back and guess what? you can get a lot of replies and essentially, book people into, into, into demos. So that's another example, essentially where you're in a clever way using, using something about your business, but

because we offer sales technology, then we can also convert these individuals that land in spam then into agent frank sales. And I'll give you an example of what happened really in our business. So we had this one employee, he's currently our head of YouTube. His name is Simon or Sam, sometimes he changes his name, sometimes not sure why. But essentially he sent me an email a long, time ago. I called the email trying to pitch me the idea of becoming famous on TikTok as a brand. And he landed in a spam and he was using competitive software. It's like, Hey, listen, you line in a spam. so I was doing that manually, right? back then, before all the agents came in. So you land in a spam and, and, and then, yeah, he became a customer and then he became an employee, you know, later on. so that's an example of, you know, trying to think of clever ways of like, okay, you've done this. This has resulted in some form of success when you've done this manually. And then you have to think back and like, could I.

apply in the AI agent because I get a lot of cold emails that land in spam. Literally like every day, maybe 10 to 100 emails in my spam and they land. In my case, I think about, damn, these are leads, but I have no really, not much time to go and go out and chase after these leads. it's a lot of manual work, right? When there's so many, so many emails landing in spam, it's like, hey, listen, you then you start, have to start to think, hey, I could provide that context into some sort of AI agent.

and analyze that context and kind of fire back a really personalized email back to an individual to actually help my business grow. So yeah, I welcome a lot of cold emails that land in spam because you'll be getting an email from me. Awesome. That's very clever. Okay. I want to jump back to one of the points that you mentioned in your intro. So you briefly mentioned about the difference between machine learning and a generative AI. I want to dig more into that. So I know that

Those two are different topics, but could you break down the key differences and how each is used in marketing today? Yes. So machine learning exists for a very long time. Apparently I read it exists since 19th. That's a very long time ago. machine learning as a concept was mentioned for the first time ever, it has massively evolved over time. Now, the reason why machine learning as a concept exists, at least, know, when I'm doing things in marketing or in sales,

You always know that a human doesn't matter how good you are, me, you, whatever, all the individuals. We have, you know, we always do this thing called human error. There's always human error with the work that we do. Right. And then of course, the more senior you get, the less human error you do. And that's great. guess what? Imagine if you could predict what is the best course of action for your business. Right. Are you not really sure? So you're going to use your domain expertise. You're going to look at some Excel data, what happened in the past, and you're going to try and optimize as a human, you're going to say, okay, I'm going to do this in this way. And that's the rules, you know, the setting in your kind of strategy workflow. Now, imagine if you could predict what you should do for your business every day. So I'll give you an example of that. Let's say with our software, for example, we have the ability to run the software as a human, right? So that's sales was there or you can, and or essentially you could use these AI agents, right? To, build pipeline. And so we can run a prediction, even internally when a customer signs up, is the customer better off?

to be on the human path, we call, or the AI agent path, or doing both to maximize the results for that customer. So you run essentially a prediction and you try and figure out what is the best possible course of action based on machine learning models. So you need a lot of data for that to do this generally. So then one of the downsides in terms of machine learning, you typically need a lot of data and it has to be labeled, et cetera, et cetera, so that you can do a lot of interesting things with that. So it's essentially predicting outcomes and doing the right course of action. And so you can do that.

in marketing, in advertising in particular. So naturally Google uses, for example, machine learning models when you, for example, set, let's say, a CPA in Google Ads, right? So Google tries to predict, for example, what should be the corresponding bid to get to that CPA, right? So that's real life examples of how machine learning has been used or has been used for decades, decades and decades. And then naturally large language models came in very recently. This is why the whole world...

you know, when Bazinga on chat GPT and you could put in an input and then you get an output, right? So because, you provide some sort of context input and then you get something back, which comes back in the form of text. Then the next thing that was, which was a huge wave last year, voice, right? You could speak to an AI agent. Now, if somebody's wondering, you know, how can I speak to one of these cool AI voice agents? There's one cool one, is currently my top favorite.

It's called Bordy.ai. Bordy.ai, just go on the website. And what Bordy does is Bordy will give you a call and will say, it will say, what are you doing? What are you working on right now? What is your biggest challenge? And then Bordy will try and connect you with another individual somewhere in the world to help you with your challenge. So essentially that's the only thing that Bordy does. Essentially, there's a lot of different kinds of use cases for voice. So on the voice front, for example, one of the biggest use cases on the voice front is when applying voice agents for call centers, Let's say customers calling up that particular, I know you have to speak about, let's say your utility bills, right? And there's something wrong with it. Instead of waiting in the queue for like God knows an hour, et cetera, to speak to an agent, a voice agent could pick up literally within seconds and have a conversation with you, process your request and essentially make changes in the system based on the conversation. And that leads to, know,

better service standards just generally within the organization, right? Being able to hit your SLA. So being able to respond very fast to your customers. And you can do that in multiple languages, whether that's in English, German, cetera. Right? So, so naturally this is where you're not predicting, you know, anything, right? We are, you're just processing, inputs into ultimately outputs, right? So ultimately machine learning and generative AI is all about inputs and outputs. So the better the inputs, the better the outputs, right? It's very simple. Like in, in that sort of sense.

It's just that machine learning predicts something, Or it helps you with decisioning essentially. Where AdjournmentCVI gives you some sort of text or voice. And then this year, what we're going to see, which is a massive development, is definitely going to be in the video space. Ultimately, you'll be looking at different videos, podcasts potentially like this one. You'll be thinking to yourself, is this guy, Frank, even real? I'm not sure whether he's real. Is this an AI agent right now that's telling me stuff? Or it's actually a real human.

And we're literally going into this, you know, into this year where you will be questioning the stuff that, what, that you're going to see on the internet. So I'll give you an example of, of, what I'm saying or what I'm also tested. said tested, example, deep fakes sales, deep fakes. What does that mean is I would give you my, my voice. would record my voice. I would record my face and then I would tell the machine or the agent, Hey, listen, I'm trying to reach out to this person. Can you please.

produce like a super cool 30 second video where I'm gonna mention some stuff about them on LinkedIn and then I'm kind of tie that in with what I'm selling, right? And you could produce these videos and there's also a lot of solutions that popping out there but that's just one example. The other example I already mentioned is you can build these sort of fake influencers as well. They get millions and millions of impressions and can build your brand essentially which is super powerful in B2C arguably, less so maybe in B2B. But definitely the videos that are like in my eyes the next wave where we're gonna see videos on the web which are fully indistinguishable from real videos that we're used to. And in my eye, for me personally, that's exciting. A lot of other people are freaking out, but you know, I'm typically the person that embraces, you know, the new wave of tech. You know, I find it pretty cool, but I also know a lot of people that say my mom, they will be like freaking out, you know, about all these things on the web. And yeah, so, but naturally that also comes with it's sort of, let's call it dangerous, just generally, right? When you're going to the sort of new world. And so I also expect...

that there will be more regulations in the US and Europe once there will be, let's call it, naughty AI agents, you know, doing something that they shouldn't be doing, right? So then this is where the governments will be more likely to step in. But right now, it's pretty okay in the US, in Europe, it's not okay. I think in Europe, we were really seeing the regulations kicking in before even people are using the technology.

So we'll see also what other parts of the world are going to be doing it. Naturally, in China, we're seeing the large language models like DeepSeq, which can reduce cost of LLMs by 95%, which is massive. If you were able to reduce the cost by so much using something like a DeepSeq or maybe similar models from China or similar parts of the world, that actually unlocks even more possibilities because one of the problems with LLMs or YI agents is some of them were actually a bit too expensive. You're looking to run them on like thousands or millions of tasks.

But if DeepSeek drops the cost by 95 % or something like this, damn, it unlocks a humongous possibility. I can go super micro granular on everything that has to happen in the business. And I'll be more essentially likely to use something like DeepSeek. Now this is not to say you should use DeepSeek. A lot of people are kind of very skeptical about DeepSeek because a lot of data goes into China and the service, et cetera. So yeah, everybody has the different perception, but I do know a lot of companies, let's say Perplexity, for example, implemented DeepSeek, but there's a lot of bunch of companies.

that did implement DeepSeek while other ones wanna, not companies, but actually governments, there are some governments that wanna ban DeepSeek because they're based in China. we're also gonna see these sort geopolitical tensions essentially on LLMs and AI agents. So, but still at the end of the day, it's quite interesting, there's still so many billions and billions and trillions of dollars being plowed into sort of LLMs and everybody wants to become the LLM region, et cetera. Europe is going off that US and everybody's kind of putting a lot of money at it. for the end user, right? In B2B or B2C means better experience or ultimately, right? So you're being able to communicate with these agents to, to, to make your life easier. Ultimately, whatever is that you're doing, whether that's shopping or anything else really, right? That's quite exciting. Just generally that we're going to see this sort of transformation, but then you also have to be taken into account that there's still billions of individuals on the planet that have not used Chagypti, you know? So you also have to kind of

know, think back in your mind, Hey, yes, there's a lot of companies or individuals or even this podcast. We're talking about LM, charge PD, many other things, but still a lot, a lot of parts of the world have not even yet tested charge of PT. So the market is still humongous to go after and kind of seeing that transformation bit by bit in different parts of the world. Awesome. This session is actually opening up my mind.

I don't know about my listeners, but I'm learning a lot. So to wrap up, I wanted to ask what's next for AI agents, but I guess you already covered that part, but do you have anything else to add to that question? Yeah. So I think sort of last year or so the trend was we were seeing these sort of AI agents popping up, right? And then typically you think about the sort of standalone agent. But the next wave after these sort of standalone agents, we're to see what's called swarm of agents.

So we're gonna see essentially multiple AI agents working with each other to drive even bigger outcome for the organizations. So let's say in our case, the way we think about the swarm of AI agents is so we currently have this one agent called Agent Frank. Agent Frank is an agent and this is what he does. But then we wanna also build other agents which are very, very different to Agent Frank. And they're gonna do different things in a different way, et cetera. It's just like you're assembling your SDR team.

Right? So there you have Agent Frank, have Agent Sally, you know, Agent Sanjana, whatever, you know, you've got all these agents and working together as a team to arrive at your big outcome, right? So, you know, it's just like humans working with each other, trying to work together to get to that outcome, because each of these agents will do things differently. So they will have different way of thinking about how to solve the problem. They can communicate with each other, et cetera. We're also gonna see things like agents buying from agents. So we're gonna see definitely this interesting world where ultimately that we're moving into, where you're gonna have these agent teams ultimately working with each other. But also the other kind of world that we're stepping into. So for example, let's say we are building AI agents to reach out to companies, right? To see if they will be interested in the pain that we're solving for them, right? Then what these companies do, what these potential buyers do, they will also use an agent to process these emails.

So we're also going to see a lot more sort of on one company using an agent to reach out to another company. And then that other company using their own agent to re to have a conversation kind of back and forth and to qualify this, whether this is a, whether this is a good opportunity or not. and that's typically ties to the company sort of data around, okay, does that company have that sort of pain to be solved? Yes, no. Right. So that's the context that you can provide, into the agent on the bias and to see if there's something in here because yeah, a lot of buyers are overwhelmed by.

by more email messages, all links in WhatsApp, you name it, calling, et cetera. But there's still something in there, right? So when you get an email, one of those hundred thousand emails is probably of value to that organization. And because the buyers are overwhelmed, they just don't know which one, which one of them is the one. So we're also going to see them, they allow, for example, agents processing your emails that come into your mailbox and try to figure out, across all these hundreds of emails that we get, if it's a called email, does this, you know, solve my problem, my company?

If yes, I may want to talk, I may want to look at that email, right? And the agent will do that. If not, I don't know, block the user, whatever it is that you want to do, right? If it is not relevant, you know, so we're going to see that sort of also processing. So what I'm trying to say is that we're going to see the AI agents talking to AI agents from different companies and essentially users using their own AI agents to communicate with each other due to let's call it inefficiency that happens in a world of communication between two humans, right? This is why you're to be using AI agents to have, let's say, the initial qualification.

Awesome. This has been such an insightful discussion for me. know AI and marketing is evolving so quickly and I'm sure our audience has gained a lot, lot of, valuable takeaways from this conversation. So before we wrap up, can you tell us like, where can our viewers connect with you and learn more about you and Salesforce? Yeah. So the best way to connect with me is always on LinkedIn. If LinkedIn doesn't work for you, I'm also on X, but typically I'm most active on LinkedIn. Otherwise, you can definitely find me. also in chat section on somewhere on the website, et cetera, et cetera. But yeah, LinkedIn is typically the best way. Awesome. Amazing. Thanks again, Frank. This was a fantastic conversation. I'm sure we'll be talking more about AI and marketing as it continues to evolve. Looking forward to catching up in the future. Bye bye. Thanks, Adana. Have a good one. Cheers.