AI & Marketing: The Future

Andrei and Sam

): [: 00:00:00

And if you'd like to find out more information, please go to marketingmachine. prorelevant. com. Today, I'm interviewing Andre Opresson and Sam Malekarjanan with Agent. ai. And just had a very interesting conversation before we came on and looking forward to hearing more about what their insights are.

ip roles atop companies like [: 00:01:00

ai. And he is currently helping transform customer service through AI driven solutions with Agent. ai. And today we're going to hear more about his insights on the future of AI and how it's revolutionizing sales and marketing. Now, Sam Sam Malakarjanan he is the growth lead at Agent. AI, where he helps drive the development and adoption of AI driven solutions to enhance customer service and sales.

With a deep background in growth strategy, Sam previously led Innovative initiatives at companies like HubSpot Labs and flock. com. He's also a former professor at Harvard university and coauthor of inbound commerce, how to sell better than Amazon. And definitely excited to hear Sam's insights on how AI is going to help us drive sales and drive marketing better for the future of business.

So welcome [: 00:02:00

Sam: It's it's quite a small world, actually. I was at HubSpot in the very early days, roughly 150 ish people when I joined. I didn't have a college degree or anything.

I built a site called HireMeHubSpot. com and ran ads targeting their employees to register for the webinar on why you should hire me. Fast forward a few years, Dharmesh Shah, who's the co founder of HubSpot was running a side project called inbound. org and needed somebody from inside the company to help him.

hey were terrible ideas like [: 00:03:00

org and some of the other stuff that we did. And then several years later, actually, after Andre and I exited our last startup I get an email from Dharmesh saying Hey, I'm working with Andre again on agent. ai. And now here we are. So Dharmesh launched agent. ai at HubSpot's inbound conference in September.

Which, by the way, Andre, next time hopefully I get more notice, I get an email in June saying, Hey, can we have 100, 000 weekly active users by September? It won't be going to work.

Guy (2): Yeah, fantastic. How about you, Andre?

Andrei: Yeah I'm an engineer, classically trained graduate studies in machine learning.

ons often with limited human [: 00:04:00

And then more recently, after we, we both exited our last company, one screen dot a I I started working with our mesh on a fun sort of. Project nights and weekends sort of thing and that escalated quickly. We started to really build out a user base.

I wanted to build out a community and really make it a real project. A real. Company and something that is going to create value in the ecosystem. As we think about AI evolving very quickly, we've all seen what the last two years have looked like. Before that chat, GPT didn't even really exist publicly at least.

g to need to get traction in [: 00:05:00

They're going to look for users, for customers, for revenue and we love marketplaces and that idea of creating as little friction as possible. Creating as low of a barrier to entry and letting those market efficiencies drive out what the future of AI should look like by just letting the best winners went out.

And so that's our, we're here doing that.

Guy (2): Yeah. Very interesting. So now tell us about agent. ai what from what you told me so far, it really sounds very interesting to help both B2B and consumer companies, but tell me a little bit more, tell us a little bit more about that.

Sam: Yeah, so we're attacking the problem from two, two different angles.

most of the other companies, [: 00:06:00

Agent studios for the enterprise, so they're expensive. They're harder to use, require specialized skills, etc. We want everybody to be able to build and customize their own agents. That also creates a new problem, though, which is the discoverability of agents. So if you go to there's an AI for that dot com or something like that, it's like 10, 000, 15, 000 agents already in most of these directories.

And even more if you split them out by the actual tasks that they're capable of doing. And so the marketplace networked aspect of this is Can we make it so that you can use agent. ai and its jobs to know you really well, your job, company, competitors, work preferences, etc. And then based on what you're trying to do, it will help you decide which AI agents to use from the marketplace and then feed them the appropriate context to maximize the effectiveness.

So make it easy to build and customize your own AI agents and then make it easy to use other ones that other people have built.

sting and it's almost sounds [: 00:07:00

Is that kind of how you see it?

Sam: Yeah. Some subset of the listeners probably remember the world before the WYSIWYG CMS, right? Where we actually had to code HTML, CSS, and stuff like that. Build our own websites. Like, when I was in college, it's how I made money was building websites for people who didn't know how to do it.

The WYSIWYG CMS made it so that non technical people could be part of the Internet. They could create their own websites. Marketers didn't have to talk to IT. The same thing's happening with agentic AI and AI studios and AI builders. It's the same thing that we want to do. Is can we do for the future of the AI agentic Internet, what the WYSIWYG CMS did for the kind of Internet of content, which is make it accessible to everyone.

hift as I was going to sound [: 00:08:00

Guy (2): Yeah, I definitely see that as well. It's like the spreadsheet too for accountants is the is AI for marketing and definitely a total game shift and I think everybody is seeing that.

How about you, Andre? What's what do you think about that?

Andrei: Yeah, I think I think that the change is going to be even bigger. I think that even the way we think about the work itself and the nature of work, I think, There's all sorts of apocalyptic views into that.

I'm actually a lot more optimistic. I think it's only going to supercharge what we can do on a daily basis. I think every job is going to evolve if you are doing so many manual things, maybe. You're responding to emails all day. You're doing some analysis of some kind. You're logging into the system.

d following up with vendors. [: 00:09:00

Not having to look in five different places and all these different systems that are going to talk to each other. Agentic AI will make it that much easier to connect those as data sources to understand all the reparative tasks that you're doing and then help you unlock, new ways of thinking about that kind of work to optimize the end goal.

And again, I think a lot of the kind of work that most of us do on a daily basis is fairly defined in terms of workflows and repetitive, tasks and context that can really be understood very well by a I connecting with some of these data sources and then give you those insights and those answers much faster.

growth engine for, something [: 00:10:00

And I actually think it's going to help many job families think a lot more long term in terms of, what are we doing and why are we even doing it, and how can we create more value? Some of that change is going to be, yeah, I think I'm comfortable. But part of it, I think is that learning journey and how do we do a better job ourselves and how do we open ourselves to that possibility through that learning, which I think AI really, yeah, I like your

Guy (2): I like your point about the looking at it what is it going to be in 3 years or 5 years and then really designing for it today and using that as your guide.

It's so easy to get stuck in. I got to get this thing done today, but. In reality, what I really want to be able to do is I want to be able to do that thing three years from now and and focusing in on that is, is definitely a mind shift right now, as I've been interviewing both vendors and agencies and and marketers on the brand side the it.

ere. I don't I think they're [: 00:11:00

It's definitely going to be different. What's what do you see there? Andre?

Andrei: Yeah, I still think we're going to have a great amount of human supervision. I think you need that to really be able to make sure we have the right kinds of outcomes. And then I think it's understanding when do we Let what go and what are we comfortable with just like any other, risk management Framework you need to think about how do we balance getting something cheap and fast versus something not as cheap But maybe higher quality right where we need that human Interaction and are we actually creating any value for anyone?

honest sort of analysis and [: 00:12:00

To be able to train AI agents and these large language models, we've already ingested all human knowledge, every conversation, every book, every Reddit, everything has already been ingested as of, last year. And at this point we're creating, particularly relevant to, to marketing, a lot of sites are mainly, or mostly, or somewhat AI generated.

ght? Because AI don't really [: 00:13:00

It's just a probabilistic answer for, a given types of type of question and constraint universe. But as we generate more AI content. It's going to become more challenging to get that sort of unique data out there. And I think those are going to be really important jobs for humans to continue to do.

How do we create more unique content, more art, more music, more writing that we can then feed back as signals into these systems and whatever that art and music and the equivalent sort of writing may need to look like for the various domains.

Guy (2): Yeah, absolutely. And cause I, let's say six months ago or whenever, the first big, push from chat GPT took place.

atever the percentages of AI [: 00:14:00

It's not really unique or necessarily. It is unique and it is certainly creative, but it's not the same kind of unique and creative that you'd have on the human side. But what do you what do you think about that, Sam?

Sam: I think marketers in particular and salespeople we're being very lazy in terms of how we're thinking about AI because it's how do we.

Spend less time sending emails. How do we write more blog articles faster? It's a lot of the stuff that we're already doing, a combination of laziness and fear are leading us to just use AI to replace the tasks that we already do. For me, which far more exciting is that the story of the last kind of 50 years of professional life has been empowerment and independence.

ring one. I'm the like, more [: 00:15:00

I, for the entire time I've known Andre, over a decade. If I needed to do anything more complicated than like CSS, maybe like JavaScript, I had to go bother him. I wanted to build a drunken spelling bee game for my local pub. Andre had to help me. Now I was able to build like a chess coaching AI agent that looks at all the stuff I'm supposed to be doing every week.

It, quizzes me, assesses my strengths. And It's never going to be as good an engineer as Andre is, it's never going to be as good at economics as I am, but it's given all of us this T shaped superpower. So sales reps can do data analytics and design and coders can do economics analysis and not as good as an actual professional, but it's given all of us a much greater range of.

elling bee games. And allows [: 00:16:00

My favorite story for this actually is that my local pub. They were feeding chat GPT. a list of all the ingredients that they wanted to get rid of, having it come up with recipes, come up with names and then putting it on the menu. So this is a group of people who can't change the PDF on their website for people to download their QR code without me helping them.

But they were able to do like artificial intelligence inventory optimization that would make Amazon proud. And they're just like turning over and turning over inventory really fast. I think that for me was, it was another like watershed moment when I saw people Can't log into their Squarespace account and update their website without me doing just some of the most impressive.

They didn't know that's what they were doing, but that's the power of these kind of conversational and language interfaces is just say what you want to do.

Guy (2): Yeah. And it's, it is very impressive. And one thing, I've even started to do is I have an idea of what I want to do. And I'd say to chat GPT, write me the prompt to do what I want to do.

it'll actually write you the [: 00:17:00

This is gonna, and it already has had an enormous impact on sales and marketing. So what do you see sales and marketing looking like in three years?

Sam: I think the immediate inclination is going to be to use AI To be again, mediocre at scale. So LinkedIn is just one giant exact same feed of everybody else, using AI to be thoughtly, very spicy takes and everything else like that.

Outbound emails. I don't even read messages anymore. SEO has been totally trashed between one, the advent of generative search results. The search engines aren't sending traffic to you. And then two just the huge volume of content that people are being created. But my favorite story in terms of how it's going to impact us.

ing disabilities issues. And [: 00:18:00

And he said, yes, he is. He's using it to. To be more cognizant of how he's using my time and the other people that he talks to by having it help him be more concise and clear and organize his thoughts. And I know that, that was a really great story for me is you can either use AI. To be mediocre at scale, because that is the definition of what AI is good at.

It's trained on a bunch of observations. It does things at an average level. Or you can use it, both the free time and the ability to have like a mirror that's not just you and your own perspective. To be better at communicating and more thoughtful about how you're using people's time. I don't want to see another spicy take on LinkedIn.

Please stop posting that.

Andrei: How

Guy (2): about you, Andre?

ith some of this automation, [: 00:19:00

Which today is fairly high, if you wanted to do this stuff at scale, imagine that going to zero which essentially, it's getting roughly 5 to 10 times a year cheaper. Extrapolate that for a number of years and. All those emails that you're getting that are automated, all those sales messages, they're going to improve, but I think we're also going to potentially get more of them if it's just easier to do if they're even more personalized, if you know that the algorithms continue to improve, you can tailor them to, only Sam, what Sam will respond to.

Maybe, the catch line should be about scotch. Great. How will they know that it might just crawl everything. So for every single individual, maybe you have a highly personalized approach for this. Again, I think it opens a lot of very interesting avenues today. That's not really possible. That's you can do it at scale, but it's fairly expensive to do that.

But again, now we're just [: 00:20:00

I real example I was playing around with, we have all the tech at home, Alexa's, Siri, you name it. But I was, I was playing with chat GPT, the new voice that they have and I was playing with that with my three year old son and you can literally ask chat GPT and voice mode to change itself to, a caretaker that is going to tell.

ating a story, interact, and [: 00:21:00

And that was fascinating. And, complete normal sort of normal interaction for, for him, it was extremely normal and he could just have the conversation is oh, and then what happened? Or, let's what about the bunny and, other sort of characters in the story.

And it, it really focused on that and made it. I personalized all in real time and all for again pennies on the dollar, right? And so as we think about, what will even learning look like in classrooms, right? And how do we think about, should we be using AI in those cases?

Do you not, how do we assess people going forward? I think a lot of really interesting questions that we don't have. Good answers for that we need to grapple with just beyond the marketing and sales But I think you know, it starts there because those are revenue drivers.

Those are connectors for most people and it's a microcosm for the rest of society

y, it's fascinating you talk [: 00:22:00

And that is really fascinating. I'm going to have to check out the the voice mode. I haven't done that yet. So I definitely will do that. That'd be a pretty good.

Sam: It's my favorite way to use AI actually, by the way, is as a learning aid. Like a few weeks ago, I couldn't sleep and I spent four hours talking to chat GDT, trying to understand some weird quantum mechanics nuance.

And no human teacher. would possibly have spent that much time listening to me ask stupid questions that I'm not really qualified to be asking anyways. But then eventually after four hours, it's oh, I see what you understand wrong. The word force is used here in one way, the word you'd force is used here in another way.

f Bezos quote, we don't make [: 00:23:00

That's going to be a really interesting application of AI. Is how can you use it to teach your customers and your prospects better?

Guy (2): Yeah. And that gets into kind of like my next question, which is the competitive edge and how a I can provide a competitive edge. One of the things that I've seen with banking, for example, we've had a couple of clients with banking.

Some of the clients have turned off a I totally within the organization. Because of the regulatory and the risk and whatever. Okay, fair enough, there is risk and, and stuff like that. But some of the in other banks that are competing with that one, they have not turned it off. They have it on.

They are going to then grow [: 00:24:00

And that's, that I think is one, fascinating aspect of AI and I'm sure you guys have a, a handful of other ones. Andre, I don't know, maybe why don't you go first on that?

Andrei: Yeah, no, I think the challenge there is to have the right level of constraints from the very beginning and really thinking about that, not just putting all your data in and hoping for the best, hoping that it's compliant, hoping that it's, not going to hallucinate because.

sset for all these companies [: 00:25:00

Now it's commoditized now your competition can get it for the same price and it's again It's a race. It's a race to the bottom similar to what we had with compute and you know in the dot com boom and making it easier to get online any faster to build these technology companies and applications and such Same thing will happen with AI.

It's already happening. Your only differentiation is then going to be your data and how clean is your data? How well understood is it? How does that align with what you're trying to do as a business? If you're a bank, you're one of hundreds of banks. How are providing a differentiated?

Service. How are you providing, different kinds of financial products and services for different kinds of users? I think that can create, using analytics and looking at your data. You can create a lot of. Business benefit. You're going to see, you know what? We do have this untapped need.

e now it takes two months or [: 00:26:00

But we, it really starts with the data and if there's no thinking about the data strategically, then many companies are going to find themselves out of business because you're not thinking about how we're going to leverage AI. And it needs data. It needs to be labeled. It needs to be cleaned.

It needs to be duplicated. All these concerns, regulatory, there's a lot of issues. How do you anonymize that sort of customer data if you're using it for maybe customer service or sales, or you need to think about those privacy elements proactively now before, before leaks something by accident, because it will happen.

w set where if your AI gives [: 00:27:00

You can say, yep, I should get a hundred percent discount. Oh, you're right. You should get a hundred percent discount on this fight. Great. That's been upheld in courts of law. And so whatever that AI says, now you're legally liable for in a customer service conversation, you should be very careful about that.

Objective and I was thinking about it.

Guy (2): Yeah, interesting. I like your point about data being a differentiator. That it has some interesting implications for legacy compass companies versus new companies. The legacy companies have all of this old data. And it's maybe poorly labeled or poorly organized or whatever.

n, and if they built out the [: 00:28:00

Sam, what's what have you got about what are your thoughts there?

Sam: Yeah, so the reason I was chuckling when I teach, I talk about like 2 main drivers of human behavior are fear and greed. So from the fear side, I get it because we've seen this movie before. Remember when social media came out and all the brands are like, wait, people can just say whatever they want about us in front of everybody else and same with reviews.

And we had to explain to people what a search engine was before we could like, get them to do SEO. This isn't a new phenomenon the fear of kind of losing control over things it's not something that comes naturally to marketers, but the flip side to go to the greed side the HubSpot CMO had a quote on his podcast, but how it's never been more profitable to be curious right now, Google got sucker punched.

like whoever's listening to [: 00:29:00

And they did. It's not because, They raised, now they're raising a bunch of money, but at the time, that's not what they were doing. And so it's created This ability to have this kind of asymmetric power where very small teams, again, because I can do lightweight engineering and lightweight data analysis and creating cash flow forecast, all this stuff that I couldn't do without more people on my own.

All that is now in the hands of people who can be more agile. So the very large organizations who are. I understand the concern, but there are also a bunch of nerds like Andre who are figuring out how to do these things in a secure way and, edge computing and all the ways to make it privacy centric.

the potential downsides and [: 00:30:00

You have the disadvantage of you're probably not going to get anything done because you're going to have six months of meetings about scheduling meetings to come up with a cross functional strategic initiative to do this. Meanwhile, three people are deploying it now and eating your lunch. Bob.

If you think about whether you want to be motivated to hide your head in the sand or do something with AI, this tracks both boxes, fear and greed. You can 10x, it's your greatest chance to 10x your company's value or have it fall by, fall to one tenth of what it is today.

Guy (2): Yeah, absolutely.

And it's so funny that you mentioned the, the big gorilla. When you're a small company, you're so afraid of what they're going to do. And those guys are sitting there trying to schedule a meeting.

Sam: Exactly. Yeah. Take advantage of that. There's never been a greater opportunity to sucker punch the gorillas since maybe the dawn of the internet.

you didn't have to have the [: 00:31:00

We all know that, feeding classified documents to an online cloud system is probably not a great idea. But if you have good people, you've got good training, everybody's got good incentives that are aligned. It's very similar to how we do cyber security training in general, right?

Don't click the email that's pretends to be from the CEO that says please download this file and fill it out with our credit card information, like use good judgment. And then other than that, unleash your team, let them go do stuff.

Guy (2): Yeah. Yeah. Man, oh, man, that's those are some great examples, and I wish I could keep going.

y're, here they are, they're [: 00:32:00

Sam: I know there tends to be like, when you're a hammer, everything looks like a nail kind of phenomenon, but this does very much remind me of the early days of my career, right? So I was college dropout with no real experience, built hiremehubspot. com. To get hired at HubSpot, ended up teaching at Harvard.

I'm a faculty chair at the school I dropped out of. University of South Florida, Tampa. And all of that's because nobody was teaching the stuff, the internet marketing, mobile marketing, social media. Nobody was teaching at the university level at the time. And so it didn't matter that I didn't follow that normal path because I could go faster than everybody else.

And I wasn't bound by knowing the right way. So I could find arguably more interesting ways. Still very much the same thing. Andre and I don't know. The future, and I stopped trying, frankly, to even predict relatively near term outcomes, and nobody else does your professors. Don't the CEO of the company where you're getting your internship at.

[: 00:33:00

Use it, use AI, do all this stuff, like leapfrog ahead, and then and all the people with all their money and experience and everything else like that will be able to come up against you.

Guy (2): Wow, I really like that shake the ladder, I really like the that commercial, I'm going to claw my way up to middle management.

Sam: Yeah, no one wants to do

Guy (2): that. That's right, Andre what's your what are your thoughts there?

Andrei: Yeah my, my advisor, and I can give the same advice recent computer science graduates, right? I think the world is changing so much that you have to lean and you have to use all these tools.

ng new. It's confusing. It's [: 00:34:00

It's a very powerful tool, but also have to keep in mind that your competition is likely use every single day. You're going to have people who may not be able to move as fast as you maybe, haven't had that experience and now can move a lot faster. Maybe they figured out the right.

The right AI or AI to use. I use just because of nature, what I do, I have AI agents doing a whole bunch of work for me every single day from research to analysis. I even have an AI agent that screenshots my whole computer screen every three seconds and summarizes it so that I can invoice faster.

I don't do invoices. My AI does it for me, right? All those little things add up. And it can really drive a level of efficiency that, or really self reflection am I wasting my time on the internet too much? Again, some of that can be dangerous and if you're not ready for it, but I think, it's helpful.

It helps, it's [: 00:35:00

Sam: Yeah, just to illustrate very clearly that we do understand why people are nervous about this.

I didn't know that's how Andre does his invoices. And as the person in charge of tracking all of the cash flow and expenditures for agent. ai, I had a tiny heart attack about 14 seconds ago, but it was immediately overcome by the fact of do I really care? Whether Andre's like invoice tracking is plus or minus 10 percent fidelity because AI, induce mistakes versus the fact that he doesn't have to worry about that and can spend his time focusing on higher tech stuff.

If you're listening to this and you're worried about that, I get it, but the pros are going to outweigh the cons.

ols because it does give you [: 00:36:00

Guy (2): person, the most valuable person in the company, what can I do to to save some, even a couple of minutes every day, it adds up and and really comes out to be an incredibly important.

Anyway, thanks so much. I really appreciate it. Sam Andre. So where can they reach you to find out more information on on your company and on what you guys are doing?

Sam: So the downside to having the name Malakarjan, it's hard to pronounce, the upside is Googling anything even close to it, you will generally find me.

But we did just launch a community at agent. ai, so if you go to community. agent. ai We're trying to get everybody there asking questions, sharing ideas, because again, we're not pretending we know all the answers here. And both Andre and I and Dharmesh and the rest of our team are all very active there.

on the blog and then go from [: 00:37:00

Thanks so much and really appreciate it and We'll we'll be talking soon. I hope Thank you so much for having us. Absolutely.