Latest / Agentic Conversations (formally mlops.community) / The Era of AI Agents in Marketing // Joel Horwitz // #337

Transcript

The Era of AI Agents in Marketing // Joel Horwitz // #337 · Agentic Conversations (formally mlops.community)

• 0:00
  There's a big like product LED growth.
• 0:01
  Everyone wants to be like data dog.
• 0:03
  Everyone wants to be like sneak. Everyone wants to be like Docker
• 0:06
  and like, Oh yeah, we don't need salespeople.
• 0:08
  Like we're just going to put our product out there.
• 0:10
  It's going to be great. It's going to sell itself.
• 0:13
  You know, we're just going to put all of our money and
• 0:14
  resources in a product development and we'll never have
• 0:17
  to deal with sales and marketing people ever again.
• 0:19
  Thank goodness. No, it's just not.
• 0:21
  It's just not real. People buy from people.
• 0:23
  I really believe that. Joel Horowitz, founder of
• 0:33
  Neoteric 3D, which stands for Digital Design and Data.
• 0:40
  I like my coffee black. I like, usually I just like a
• 0:43
  double espresso. That's where I go just like me.
• 0:47
  Short and sweet and sometimes a little bitter.
• 0:49
  And yeah, that's me in a nutshell.
• 0:53
  Funny enough, I reference marketing use cases all the time
• 0:57
  for AI and I've got another buddy that I always say is yeah,
• 1:02
  my friend in marketing is doing this and he's got quite a
• 1:05
  complex agentic workflow set up. You are not that friend.
• 1:11
  But you are also doing agentic workflows, Yes, Far beyond
• 1:18
  having asking Chachi BT to write you a blog post.
• 1:22
  Right. Yeah, that was basically the,
• 1:25
  you know, I think of like Jasper and even before the ChatGPT,
• 1:29
  like that was the big thing was like still is, I would argue.
• 1:33
  And most people don't want to talk about it because they're
• 1:36
  like, they don't want to admit that they're probably still
• 1:38
  doing it, right, creating content with AI and, you know,
• 1:45
  because they're like, oh, it hallucinates and it's not very
• 1:48
  good. And but from what I've seen,
• 1:50
  it's gotten a heck of a lot better, you know, And so now
• 1:54
  people just stop talking about it because they're like, well,
• 1:57
  there's an arbitrage opportunity where if I don't talk about it,
• 2:00
  I can keep charging what I've been charging, you know, for
• 2:02
  creating content. And no one, hopefully no one's
• 2:05
  the wiser. But what's even more exciting
• 2:08
  now that also people aren't talking about, and maybe they do
• 2:11
  a little bit, like I see a few. I just saw a blog post today by
• 2:15
  Adam Schoenfield, who was talking about, yeah, like, you
• 2:20
  know, the, you know, the original kind of like AI
• 2:25
  solutions and automations were more like, they do say
• 2:28
  personalization. And it was all about quantity.
• 2:31
  Like, oh, well, you can, you can generate, you know, thousands,
• 2:35
  if not 10s of thousands of cold emails and cold outreach.
• 2:39
  And, you know, even if you get 1%, it's still, you know, you
• 2:43
  know, some, some good leads that can come in and everyone just
• 2:46
  got spammed. And so that's kind of like what
• 2:49
  the, you know, what, what people you know, view as the, as the
• 2:54
  status quo. Yep, it's, it's nice.
• 2:56
  He actually had a really nice graph.
• 2:57
  I'll try to pull it up later if I can.
• 2:58
  But basically he shows actually, and what I'm seeing is now it's
• 3:03
  more value creation. It's more like quality.
• 3:07
  Like you can create much better in my view, like much better,
• 3:12
  you know, interactions, not just emails, but interactions.
• 3:15
  And I use that word because it's like it's not just content as we
• 3:18
  talk, you know, just as you mentioned, it's across the
• 3:22
  board. And now that's always been the
• 3:24
  marketing, you know, play is like, how do you meet, meet your
• 3:29
  target audience where they are with the right content or the
• 3:32
  right message, right? That that's timely and
• 3:36
  effective. And so nowadays, I think
• 3:39
  there's, I mean, within the last few weeks, I feel like it's like
• 3:42
  always getting better. I would, I would argue, I think
• 3:46
  my hypothesis is that the types of experiences you can create
• 3:51
  for people as a marketer now far outweigh what you could do
• 3:55
  manually, so to speak, right. You know, it's like analog
• 3:59
  digital, right? It's like, sure, like you can
• 4:03
  still, you know, do things analog, but but I think there's
• 4:08
  a, there's a tipping point, not to sound too marketer, but there
• 4:11
  is a there is a inflection point happening where, yeah, like the
• 4:16
  what you can do in terms of like using a gentic AI far surpasses
• 4:23
  anything else that I've ever seen.
• 4:24
  It's amazing. It really is.
• 4:27
  All right. Well, the obvious next question
• 4:29
  is give me some concrete ways that you're doing this.
• 4:35
  So some concrete ways I think that people are, are leveraging
• 4:40
  this. And to be honest, it's, I know
• 4:42
  we talked about agentic AI, but this is also Harkins back to the
• 4:46
  big data era as well, right, Circa 2000 tens, which I, I
• 4:49
  certainly lived through. And there's a lot of
• 4:51
  similarities, a lot of the types of workbooks or workflows that I
• 4:57
  was building back then using, you know, Hadoop and products
• 5:02
  like data mirror, right, is synonymous.
• 5:05
  In essence, it's like clay is essentially data mirror like 3
• 5:08
  dot O if you ask me, it really is because it still uses that
• 5:12
  same workbook, you know, kind of format which everyone's familiar
• 5:17
  with. So the concrete example is we're
• 5:20
  using clay at a number of our clients to essentially create
• 5:27
  the type of workflow where when we recognize an intense signal,
• 5:33
  whether that comes from like someone who just started a free
• 5:35
  trial or someone who just within a, within an account posted a
• 5:40
  new job opportunity. We can basically get a sense of,
• 5:47
  of that person and that account. And then use agentic AI to not
• 5:55
  only craft an outreach kind of sequence, if you will, not just
• 6:00
  through e-mail and, and phone calls, but like we can create
• 6:04
  like really cool memes, like with that person's actual image,
• 6:09
  right? We can create use like I just, I
• 6:13
  recorded a founder's voice the other day.
• 6:15
  So like I can use his voice now to basically have like a
• 6:18
  personalized message from that founder to a, a prospect, which
• 6:25
  I know, I mean, people are going to see this and hear this.
• 6:27
  Like that's so, you know, sleazy that you do that.
• 6:30
  But, you know, I mean, it's, it's, it works and it's, it's
• 6:34
  quite effective, you know, and so there's.
• 6:38
  Yeah. My big question is why is that?
• 6:41
  Or how is that agentic? Yeah, I think the agentic part
• 6:47
  is where you're using, you know, what you would call like AI
• 6:53
  researchers or AI agents to go out and look for these signals,
• 6:57
  right. And it's quite impressive.
• 6:59
  Like there's, there's a whole, there's a whole industry of, of
• 7:04
  products out there that, you know, charge a lot to get this
• 7:08
  type of information. But if you have an agent that
• 7:12
  you've developed that actually can go out there and essentially
• 7:16
  do the type of operational research that like an SDR would
• 7:20
  do, like a sales development Rep would do.
• 7:22
  So for example, you know, you can basically write an agent to
• 7:28
  go and basically check their LinkedIn profile as you would,
• 7:32
  as anyone would like basically mirror.
• 7:34
  Like I literally like one of the first things I do when I I, I on
• 7:37
  board a new client is I basically sit like shoulder to
• 7:40
  shoulder to them and, and basically and map their kind of
• 7:46
  their work day, right? Like what do you do when you get
• 7:48
  up in the morning? And they're like, well, I go
• 7:51
  obviously check my emails, but then I go and go on LinkedIn and
• 7:55
  I'm looking at who came in, like sign up for trial or form or
• 7:59
  contact us, whatever. So basically, I won't take you
• 8:01
  through the whole spiel, but you know, I sit down and go, OK,
• 8:04
  what are, what is this person doing?
• 8:06
  And then I turn around, I work with our development team and I
• 8:09
  say, OK, how do we recreate that with an AI agent to do exactly
• 8:13
  the same thing so that we can scale them from individual to,
• 8:17
  you know, a number of of agents that can do that for them.
• 8:20
  I mean, that's basically what it is.
• 8:24
  Yeah. And no, I guess where I'm trying
• 8:30
  to see is like you have someone that shows these strong signals
• 8:35
  and I really like that where, oh, there's a job posting or
• 8:39
  there's sign up for a free trial and potentially what the best
• 8:45
  action is, is having an SDR get them on the phone, right, right.
• 8:50
  OK. But that's easier said than
• 8:52
  done, right? And so you want to entice them
• 8:57
  in other ways and you don't want to have that SDR have to spend
• 9:03
  10 or 15 minutes creating a meme that may or may not get read.
• 9:09
  And so how can you create these workflows that are leveraging
• 9:14
  your creativity of going out there and getting prospects to
• 9:20
  answer, but that also are not having the workforce have that
• 9:28
  like time intense activity? I mean, I think that's what's
• 9:32
  exciting about this, at least as far as I can tell, is that it's,
• 9:36
  I think there's a lot of like people who might make the
• 9:39
  argument, well, if you can do this with AI agents and you
• 9:42
  don't really need, you know, a sales development Rep or what
• 9:45
  have you. But that's not what I'm finding
• 9:46
  at all because it's not like these things are, you know, copy
• 9:51
  and paste, you know, every SDR has their own kind of flavor of
• 9:56
  how they, you know, style really, of how they do things.
• 10:00
  Sorry for the background noise. I'm, I'm on campus.
• 10:03
  But they have their own, you know, style of, of doing these
• 10:07
  things. And so that's really what you're
• 10:10
  what you're doing. And so yeah, like 1 when you,
• 10:14
  when you basically look at a account.
• 10:17
  And I definitely do think at the account level, right?
• 10:20
  Because you know, Beyoncé like start-ups where the founder is
• 10:24
  likely the person that's purchasing, you know, piece of,
• 10:27
  of software, right? When you get in the larger
• 10:30
  companies, say over 100 million in ARR, it's usually multiple
• 10:35
  people who make a buying decision, right?
• 10:36
  You have to go through multiple kind of buy buyer, you know,
• 10:41
  decision making. And so I think it really is at
• 10:44
  the account level. And the reality is there just
• 10:47
  isn't enough time in the day for them to truly understand like
• 10:51
  what's in what's in that account, you know, going through
• 10:54
  and reading their, you know, if they're a public company, you
• 10:57
  know, their, their filings, right?
• 10:59
  Or if they're a private company going and searching things like
• 11:02
  Crunchbase or, you know, pitch book to see kind of like other
• 11:06
  information, like understanding their competitive.
• 11:09
  Like there's a ton of research right, that goes into it that I
• 11:11
  think a lot of the agentic AI stuff solves for, which is
• 11:14
  really exciting. But like you said, there's also
• 11:16
  the creative part of creating not just like a a meme, but also
• 11:21
  we create entire websites for them, right?
• 11:24
  So, you know, today, like you said, like, yeah, you're
• 11:28
  depending on, you know, what the relationship is with that
• 11:32
  prospect. You know, getting them to it,
• 11:36
  you know, getting them on a call or asking them for a meeting is
• 11:40
  probably way, you know, premature, right?
• 11:44
  Realistically, what you're doing is you're trying to surface in
• 11:47
  trying to educate them to understand how your product or
• 11:51
  service meets their specific needs.
• 11:54
  Because the biggest mistake that I see people make in marketing
• 11:58
  at least is, you know, assuming that there's there, it's like a
• 12:02
  blank slate, like there's nothing there already.
• 12:05
  And often times there's a lot there already.
• 12:06
  Like people have built their entire, you know, development
• 12:09
  environments or entire data, you know, environments.
• 12:12
  And so you have to find a way how you can reasonably, you
• 12:16
  know, enter, you know, integrate yourself into that.
• 12:18
  And so, you know, we built some incredible personalized
• 12:23
  websites. We've used Mutiny as a vendor
• 12:26
  for that. I did, there's a whole blog that
• 12:28
  was written about how we did that at a client called single
• 12:30
  store, which was exciting. We're, we're doing that also
• 12:35
  creating customized like LinkedIn ads for specific
• 12:40
  accounts, a specific messaging based on, on what we find.
• 12:43
  And it works. I mean, I've, it's been, it's
• 12:45
  been crazy because I've actually seen target accounts like thumbs
• 12:51
  up and like LinkedIn ads. I've never seen that before in
• 12:54
  my, in my time where like the account itself is like giving us
• 12:57
  a thumbs up for the ad because it's so, you know, on, on topic,
• 13:01
  right? It's still on target with like
• 13:03
  the messaging. So anyways, it's, it's exciting.
• 13:06
  It's an exciting time to be a marketer, I would say.
• 13:08
  So let's run through a few more of these use cases.
• 13:11
  I think they're fascinating. I've seen folks on the marketing
• 13:15
  side who are buying Google Adwords.
• 13:19
  They will use LLMS to update the AdWords because every once in a
• 13:25
  while you get certain keywords that go stale or you have to,
• 13:31
  you have to continuously be checking your AdWords account
• 13:36
  and making tweaks and what not. And so they'll use LMS to help
• 13:41
  them optimize the spend there and just update these stale
• 13:49
  keywords. I've also seen it for keywords
• 13:54
  with SEO and use it a tool like a trefs or just a keyword
• 14:00
  basically. A semra?
• 14:01
  Yep. Exactly.
• 14:02
  SEMRA getting the CSV maybe for certain keywords that you're
• 14:09
  trying to rank for, and then the LM can tell you what other
• 14:14
  keywords you might want to be ranking for and which ones you
• 14:17
  should really try to optimize for because they seem like they
• 14:21
  have the best score. Yep.
• 14:23
  Yeah. And then there's also negative
• 14:24
  keywords, right. I mean, you know, I, I generally
• 14:28
  how I've seen this work in the anal, I think the analog to
• 14:31
  digital is a really good kind of construct.
• 14:34
  The way to think about this, like, you know, pre LLM, you
• 14:37
  know, post LLM, like pre LLM or analog version of this is
• 14:43
  downloading, you know, the CSV file of all of your keywords for
• 14:46
  the last 30 days or 90 days, whatever the window is going
• 14:50
  through. And basically, you know, marking
• 14:52
  the ones, the keywords that are, you know, that are leading to a
• 14:57
  conversion and the ones that are not leading that are not.
• 15:01
  And so those are what are called like negative keywords, right?
• 15:04
  And so you'd literally mean we go through and like check, make
• 15:08
  those checks and then re upload, you know, those keywords into
• 15:11
  your campaign. And that's kind of how you would
• 15:12
  do it. There's also other ways like you
• 15:14
  can let like Google decide for you, like just say actually I'm
• 15:18
  going to have Google optimize, you know, my my keyword list,
• 15:22
  but I don't know. But yeah, I think that's a
• 15:25
  that's a really cool use case. I expect to see that a lot more.
• 15:28
  I think what I'm hearing actually, even interestingly
• 15:32
  enough, like Google Adwords itself, I feel like is, you
• 15:36
  know, I've been, I don't know if you ever listen to that all in
• 15:38
  podcast, but there's a lot of conversation about what that,
• 15:41
  what search looks like in the age of, of LLMS and, and
• 15:45
  ChatGPT. And so now we can actually, I
• 15:47
  can see like the, you know, UTM parameters coming in for a lot
• 15:51
  of my clients because I'm like a, I'm, I'm definitely like a
• 15:54
  pretty firm believer that you have to UTM every link ever.
• 15:58
  So you know what's going on and so I can view a lot of the UT
• 16:03
  Ms. and I've been noticing more and more like the referral being
• 16:06
  open AI and like some of these. So I think the bigger
• 16:10
  conversation is how do you SEO for LLMS, like for open, you
• 16:15
  know, for ChatGPT or Gemini or what have you, What I'm hearing
• 16:19
  from, you know, people like that are SEO experts.
• 16:23
  I'm not an SEO expert, but people who are have told me that
• 16:28
  in general, it's following kind of the similar patterns like of
• 16:32
  getting indexed by search and that sort of thing.
• 16:35
  But there's still a bit, a little bit of an unknown about
• 16:39
  how these LLMS are indexing right, blogs and pages and that.
• 16:45
  But I definitely see like growth in terms of referral traffic
• 16:49
  coming from ChatGPT. It's kind of wild to watch that
• 16:52
  kind of like go up compared to like just Google search.
• 16:57
  Yeah, I've heard it called Answer Engine Optimization.
• 17:01
  So it's like AEO instead of FEO? Yeah, yeah, I'd be curious like,
• 17:07
  you know, maybe as a follow on, if you had someone, you know,
• 17:11
  that knows those things better, 'cause I know you have an
• 17:13
  extensive network of people who maybe can see behind the scenes
• 17:16
  with some of these things. I'd be very curious as a
• 17:18
  marketer to know to know that because definitely, I mean,
• 17:22
  sure, I go, I go to ChatGPT now more than for sure than Google
• 17:27
  search like to get answers for sure.
• 17:30
  So, yeah, I think that's a, that's a big kind of tectonic
• 17:34
  shift that's happening in marketing right now.
• 17:36
  Well, and the whole conversation is around, OK, our click through
• 17:43
  rate or basically our page views are lowered.
• 17:48
  We're getting less traffic sent to us from search, but we're
• 17:54
  still getting more relevant traffic.
• 17:58
  And so even though it's less, we're getting relevant traffic
• 18:02
  because it is again, that answer that will link back to you from
• 18:10
  ChatGPT or Gemini output. And it's like you're 8 steps out
• 18:16
  of 10 on your journey and you don't need to sift through all
• 18:21
  this like super top of the funnel crap that wasn't getting
• 18:26
  you any actual paying customers. It was just getting your SEO
• 18:33
  agency a whole lot of coin. Yeah, you you've tapped into
• 18:37
  something that is it. That's just the that's just the
• 18:41
  so-called tip of the iceberg. I mean, there, there were so
• 18:44
  many, I would say circa like 2020-2021, like the last five
• 18:51
  years, I feel like that that whole game of top of funnel, you
• 18:56
  know, marketing has just gotten blown out, blown out completely.
• 19:00
  Like I would be very surprised if there's a single CEO or CMO
• 19:04
  that even like, you know, we used to call it vanity metrics
• 19:07
  and somehow it got like lost in the shuffle that it's still
• 19:10
  vanity metrics. And you know, and, and what
• 19:14
  would happen is, is not only would you get, I, I hate to say
• 19:18
  junk traffic, but you would, I mean, you even get bots and
• 19:21
  things right that would come in and people would pat themselves
• 19:23
  on the back, but even more so they then say, well, but at
• 19:27
  least they like signed up. And then you go and look at
• 19:29
  those sign ups and they're like all Gmail, yahooqq.com like, you
• 19:34
  know, and again, I hate to say garbage sign ups, but they were.
• 19:37
  And so one of the things that's actually been kind of a
• 19:39
  breakthrough with a lot of my clients is in fact using these
• 19:43
  AI agents to do research on those emails to actually figure
• 19:47
  out who these people are. And a lot of people don't
• 19:50
  realize like you can actually, I mean, you have to be really
• 19:53
  careful with PII and all of that, but you can figure out who
• 19:56
  they are now. And you can say, oh, this is
• 19:58
  actually a competitor that's in there trying to like get
• 20:01
  information. And so we, we didn't know that
• 20:03
  like with a few of our clients and you can kind of put them on
• 20:07
  a different path, so to speak. So.
• 20:10
  How are you doing that? Just by creating workflows that
• 20:14
  are researching the Gmail addresses and.
• 20:19
  Yeah, yeah. So you can basically use their,
• 20:24
  I mean, they give you a couple pieces of information.
• 20:26
  It's, I think there's a guy who does a lot of the training for
• 20:29
  Clay's name is Yaz. And he says like one of the back
• 20:34
  in my college days, I used to work for UPS to pay my way
• 20:37
  through school. And when you're stacking, I used
• 20:39
  to load these containers like full of boxes.
• 20:41
  And they used to say there's like a cornerstone box.
• 20:44
  And like the program was actually called cornerstone.
• 20:46
  And so he has to take a box and you start in one corner of the,
• 20:50
  of the trailer, right? And then you build your wall
• 20:52
  from there. So the boxes don't fall on you
• 20:54
  when you're, when you're loading.
• 20:56
  And that's kind of what this is like.
• 20:57
  You have a cornerstone piece, which is basically, you know,
• 21:00
  for now anyways, as your LinkedIn profile.
• 21:04
  And so basically you can, you can try to search using these AI
• 21:10
  agents, their Gmail. And then you can see also based
• 21:16
  on their, you know, first and last name, if they give that to
• 21:18
  you. And there's ways you know, to,
• 21:21
  to figure out who they could be. And it's not 100%.
• 21:24
  I think we get about a, I'd say the industry average right now
• 21:27
  is about 4045% hit rate. So you still don't, there's
• 21:30
  still a bunch that you don't know.
• 21:31
  But the bottom line is like we're, we're implementing this
• 21:34
  idea like, I don't know, you know, most people have like, you
• 21:38
  know, MQLSSQLS, they have lead scoring.
• 21:42
  Like that whole like domain is completely getting upended by
• 21:48
  this AI kind of evolution, right?
• 21:52
  And I call it evolution, not a revolution, because it is.
• 21:54
  It's evolving. Why?
• 21:57
  Why is it no more NQLS? Well, because it's a, it's such
• 22:02
  a road, you know, analog model, right?
• 22:05
  It's like, well, they, they clicked on 4 links or they
• 22:09
  visited your website and they also attended A webinar and they
• 22:12
  downloaded a white paper. But you know, that's not really
• 22:15
  like they're ready to purchase. They could just have a lot of
• 22:17
  free time on their hands. I don't know, right.
• 22:20
  Like, you know what I mean? Like, and that's probably true.
• 22:23
  And then the alternative is then you get like a SQL, which is a
• 22:26
  sales qualified lead, right? And they're going to say, well,
• 22:29
  this is a really good lead because I know a guy who knows a
• 22:32
  guy and, and, and we think that they're really, you know, a
• 22:37
  perfect target for our solution and and so on and so forth, you
• 22:42
  know, but really what you want is an AIQL, which kind of
• 22:47
  removes a lot of that subjectivity.
• 22:48
  I think. I mean, there's a new type of
• 22:50
  subjectivity coming from AI, but you know it, what it can do is
• 22:54
  like I said, you have to feed it a bunch of signal, right?
• 22:58
  Like I said, so job postings, public filings, you know, org,
• 23:03
  like there's the org chart, right?
• 23:04
  So there's a there's a website calledtheorgyouknow.com.
• 23:08
  So there's ways to figure out like who's who in the zoo.
• 23:10
  Like that's a a buddy of mine used to say that at IBM all the
• 23:13
  time, which kind of got annoying after a while, but anyways, but
• 23:16
  it's true. It was like, who's who in the
• 23:17
  zoo and who has the budget and what are they, what's the, what
• 23:21
  are they willing to pay? And do they have a competitor
• 23:23
  already? And what's the relationships
• 23:25
  like? Because yeah, like I'm, I'm
• 23:27
  sure, you know, like it's, it's a very complex thing to sell,
• 23:32
  you know, sell enterprise software into a large
• 23:34
  organization. Well, one thing that I am trying
• 23:39
  to wrap my head around is you get all these signals, you get
• 23:43
  all this data and some of it may be incomplete.
• 23:50
  And then where are you throwing all of this?
• 23:53
  Just into a context window and then asking an LLM to say are
• 23:58
  they in the buying cycle or not? And then they get a score that
• 24:02
  gets fed to the sales team. Or how does that process look
• 24:07
  of, hey, we got all these signals, this person looks like
• 24:10
  a green light, we should reach out.
• 24:12
  Yeah. We automatically reach out with
• 24:15
  AI and we get something back and it only gets surfaced when they
• 24:18
  respond because we've done all these cool memes and this audio
• 24:23
  generation from the founder like paint.
• 24:26
  Full picture in the end for me. Well, we have.
• 24:30
  So one of the other things that we developed is a unified kind
• 24:34
  of inbox, right? And so, so we feed all of the
• 24:38
  kind of the outreach, if you will, into kind of a single
• 24:41
  inbox and anyone can view in the in the company.
• 24:44
  And so, you know, when someone replies back, like to one of
• 24:48
  these, you know, tactics, we can immediately have someone jump in
• 24:53
  and say, hey, yeah, I saw, you know, that you are interested
• 24:56
  in, in such and such or so and so, and we can immediately start
• 24:59
  that conversation. But yeah, they hit a certain
• 25:02
  threshold where, you know, they, they put out a job posting that
• 25:06
  mentions, you know, for example, like for one of our clients
• 25:11
  they're targeting like my sequel shops because my sequel is known
• 25:16
  to basically, you know, basically tap out after a
• 25:21
  certain level of, of concurrent users, right, for an
• 25:23
  application. So we're always looking to say,
• 25:25
  OK, they're trying to do some implementation of my sequel for
• 25:28
  some sort of SAS application. So we see that kind of job
• 25:32
  posting and they kind of mentioned some of these things.
• 25:35
  Then I, I actually take, like you said, it's within that,
• 25:38
  within that context window. And then we actually integrate
• 25:40
  into Slack. And so we have our own little
• 25:42
  bot, if you will, within art, because that's where people live
• 25:45
  nowadays And Slack, as I'm sure you know.
• 25:48
  And so we have a Slack that actually pings the rap and says,
• 25:50
  hey, like as a person would, as an SDR would, it says, Hey, I
• 25:54
  noticed, you know your cover, you know, your, your coverage,
• 25:58
  this area or you're the AE, the account executive for this
• 26:01
  account, right? You should know that they just
• 26:04
  put out this job posting and here's the hiring manager.
• 26:07
  You should here's their LinkedIn contact.
• 26:09
  And by the way, here's some interesting facts you should
• 26:10
  know about them. They happen to be a marathoner
• 26:12
  or they like to paint in their free time or they have kids or
• 26:16
  what have you. And then you can ask your they
• 26:18
  they themselves can also write a really nice note which increases
• 26:21
  their odds of of acceptance. And so you're getting that ping.
• 26:26
  Do you have the capability? Because I can see this being
• 26:29
  like the marketer's cursor in a way where you.
• 26:33
  Can. It's a really good analogy.
• 26:36
  And you can go back to the bot and say draft me up a a LinkedIn
• 26:43
  message that is relevant like you know what we're doing and
• 26:50
  the job post has X amount of crossover.
• 26:54
  Yeah, I, I think that's a great analogy.
• 26:57
  I don't think, I think Clay's probably the closest thing I
• 26:59
  see, but it's still pretty like Workflowy, you know what I mean?
• 27:04
  Like there's still a lot of, I don't think we've really, now if
• 27:07
  someone has that Y Combinator or somewhere, somewhere else wants
• 27:10
  to talk to me and we want to start it, start something
• 27:12
  because I don't think there is a cursor for marketing yet.
• 27:15
  I do. I do think there should be, I do
• 27:16
  think there should be a, you know, a way to like, you know,
• 27:20
  like you can with replit or, or cursor, right?
• 27:23
  Just go in there and say, I want to launch a campaign to, you
• 27:26
  know, cybersecurity companies and it'll automatically create,
• 27:30
  you know, the entire, you know, campaign.
• 27:32
  That includes, you know, ad copy and creative, the messaging, you
• 27:36
  know, target audience, the landing pages, the white papers,
• 27:40
  the content based on all that. Like I'm not aware of anything
• 27:43
  that exists today. I have to base, I mean, that's
• 27:44
  why I have an agency, right, is because it doesn't exist and I
• 27:48
  have to do that essentially manually, but stitch all these
• 27:52
  things together. So there's a lot of bespoke
• 27:54
  solutions that that touch like what we've all what we've been
• 27:57
  talking about. But I feel like I'm more of like
• 28:00
  a system integrator for marketing today, right?
• 28:04
  Is this really what the job is today as a marketer?
• 28:07
  It's you're, you're a system integrator period.
• 28:09
  That's how I see it. Yeah, it's funny that you
• 28:12
  mentioned that because I was just talking to a friend, Sean,
• 28:15
  last week on how if you look at any job, what people are doing
• 28:21
  in their job is just getting, they want the end result.
• 28:27
  You want the job to be done type of thinking where you don't buy
• 28:31
  a drill to drill things. You buy a drill because you want
• 28:35
  to put a picture up on the wall and so you need to drill a hole
• 28:38
  in the wall and then then you can hang that photo.
• 28:43
  So you want an end result, but there's all these steps in
• 28:46
  between. There's these tasks in between.
• 28:48
  And maybe some of it is moving data from over here and putting
• 28:52
  it over here or it is creating something over here.
• 28:57
  And so all these tasks that you can break out in that whole end
• 29:01
  to end flow is where you are starting to see now AI play a
• 29:10
  part in some of these tasks. Whether or not it is helping
• 29:15
  with the productivity of you're able to spin something up faster
• 29:18
  or it's just it can go ahead and do it for you.
• 29:23
  And then you can give it the sign off.
• 29:25
  Or sometimes people don't even give it sign off, they just
• 29:28
  throw out shit. Yeah.
• 29:29
  I think that's a really, really important point.
• 29:32
  I think there's two points in there that you made that are
• 29:34
  really important. One is you absolutely need to
• 29:37
  still sign this stuff off because it's still not 100%.
• 29:40
  I think I had AI had a salesperson share with me the
• 29:44
  other day that we totally got the name wrong of, of this
• 29:47
  person for whatever reason. And I was just like, you guys,
• 29:50
  that's not my name and that's not what I do crap.
• 29:54
  And it happens, right? I mean, there's still some
• 29:55
  margin of error with these things And, and so it happens.
• 29:58
  But you know, you, you basically own it and you say you're right,
• 30:03
  you know, this was we we got too ahead of ourselves.
• 30:07
  And you know, but I think even humans make mistake.
• 30:09
  I always think about like the argument against like self
• 30:12
  driving cars. It's like, oh, there, you know,
• 30:14
  they, they, there's one accident this year.
• 30:16
  And so we should just cancel the whole program.
• 30:18
  Like wait a second, like how do how are humans doing out there
• 30:22
  with with the car driving and the it's in the marketing
• 30:26
  situation? Like, I don't think it's 100%
• 30:28
  either, last I checked right. So we have this weird, like I
• 30:32
  thinking that, you know, we should hold AI to this higher
• 30:35
  standard that we do humans for some reason.
• 30:38
  And I think that's a little bit misplaced.
• 30:39
  I'm like, I get it, like we should do better.
• 30:41
  But on the other hand, we should really be with the second
• 30:44
  comment that you made about buying the drill, again, I think
• 30:47
  it's if you're absolutely right. And, and you know, using that
• 30:50
  kind of analogy, what's the biggest thing, at least for me,
• 30:54
  when I'm hating hanging a painting, like missing the stud,
• 30:57
  right? So again, you have to find a
• 30:59
  signal because I can, I can build amazing campaigns, but if
• 31:03
  I, if I miss that stud or if I miss the target, all of that's
• 31:08
  for not right. And so that's why these AI
• 31:10
  agents are so critical, right, for marketers is because you can
• 31:15
  spend, I've seen literally like people spend millions and 10s of
• 31:19
  millions of dollars on campaigns because they had the wrong
• 31:23
  messaging or they had the wrong audience or they had the wrong
• 31:26
  offering. And you know, and all of that
• 31:29
  could have been avoided had there been a bit more research
• 31:31
  done at the start to really deeply understand the pain of
• 31:35
  what that prospect is going through.
• 31:39
  And that's what I think, you know, a lot of marketers just
• 31:41
  don't have the time to do because they're so focused on
• 31:43
  like doing the tactical stuff, right.
• 31:47
  So I'm actually like think there's like a renaissance
• 31:49
  coming of marketing because we're not going to be, you know,
• 31:53
  charged with having to spend 90% of our time just like finagling
• 31:57
  things and and building. We can actually spend more of
• 32:01
  our time like really understanding that customer,
• 32:03
  which is I think what marketing is supposed to be doing at the
• 32:06
  outset. But anyway.
• 32:08
  It it's just funny that you mention, hey, I'm a systems
• 32:12
  integrator because it goes back to what my friend Sean was
• 32:17
  saying is that we're all going to become architects.
• 32:21
  Yeah. And we are going to be as good
• 32:25
  as we are able to architect our system.
• 32:27
  And so if we can put all these disparate pieces of AI
• 32:35
  capabilities together and make a cohesive system out of it, then
• 32:40
  we're going to get so much more lift from that.
• 32:44
  Yeah, yeah, 100%. And I, I think what, like I
• 32:46
  said, I think we're really at the early, at the early days of
• 32:49
  this, but who am I to say? Like I just, I just have my
• 32:51
  little purview. I'm sure there are people that
• 32:53
  are doing amazing things that they don't talk about, you know,
• 32:56
  because it's probably their IP. But yeah, I, I would say that
• 33:02
  it's, it's really quite fun. I mean, that's the other part
• 33:05
  about this. Like I remember, like marketing
• 33:07
  used to be really fun, like in the 2000s, I guess, like in the
• 33:12
  age of the Internet and all of that.
• 33:14
  And then like we ran like some amazing campaigns and people had
• 33:17
  really fun with it. And then like it got it, it
• 33:20
  stopped being fun. Like I think developers like
• 33:22
  lash back against marketers. Like it was like, you know, with
• 33:26
  pitchforks, like, you know, and, and going after that.
• 33:28
  I was in AI was at a conference not too long ago and I was
• 33:32
  sitting in a room. Now I'm trying to remember which
• 33:35
  one it was, but it was a while back.
• 33:37
  And oh, I, I think it was a, it was a Kubernetes conference in
• 33:41
  Chicago. That's where it was.
• 33:43
  And so these are pretty hardcore developers.
• 33:45
  And so I was sitting in this like investor room.
• 33:48
  And I mean, if he's just like, I didn't, I almost like wanted to
• 33:52
  like sneak out of there 'cause I'm like, they're just like just
• 33:55
  bashing marketing. Like, oh, like marketing always
• 33:58
  gets developers wrong and don't hire them and they're terrible.
• 34:00
  Like, Oh, no, got to get out of here.
• 34:03
  But I, I still think that's, there's a little bit of that and
• 34:06
  I, I get it. Like I think there's been, you
• 34:08
  know, times. Look, I lived through the IBM
• 34:10
  Watson era, quite honestly, right?
• 34:12
  So I was, I can tell you like there's, there was a a bit of
• 34:16
  hand waving at that time, quite honestly, right, Which has been
• 34:21
  I think is fairly well understood at this point.
• 34:24
  And, and so, yeah, I get it. Like, you know, marketers are
• 34:27
  trying to push the envelope and get people excited about a new
• 34:30
  technology or a new service or new product and they they
• 34:34
  sometimes go too far. And that's something you have to
• 34:36
  be really, really cautious of. But going back to this point, I
• 34:40
  just wanted to close the loop on the cursor for marketers or the
• 34:44
  Devon or whatever, insert your favorite code agent for
• 34:50
  marketers. I think that is something that
• 34:57
  would be incredible because if you can prompt the whole
• 35:02
  campaign. Yeah, I mean there's there
• 35:04
  already is ways to do that. Like for example there is an
• 35:08
  integration right now between clay and web flow.
• 35:11
  I don't think it's a great integration yet as far as I
• 35:13
  know. We personally like at single
• 35:16
  store we use Mutiny. So we were actually building
• 35:20
  these pages on Mutiny that could basically customize yeah, like
• 35:27
  based on on what we knew about about the account, right.
• 35:31
  And so we can speak to speak to their pain points and use like
• 35:36
  we literally update the like the customer case studies, right,
• 35:42
  based on the customer that came in right.
• 35:44
  You'd want to find a look alike right to the customer that
• 35:48
  you're targeting saying, you know, did you know your
• 35:51
  competitor is using us to do this right?
• 35:53
  And they go, oh, OK, Like there's always like a FOMO
• 35:56
  there. But yeah, I, I saw some more
• 35:59
  recently someone said that vibe marketing is a thing already.
• 36:02
  I mean, I, I feel like it's happening.
• 36:04
  I'm just not like I've been like, so heads down that
• 36:07
  sometimes you forget to kind of look up and look around like,
• 36:09
  OK, what what other people doing?
• 36:11
  Because I'm like, so in the weeds of building these things,
• 36:15
  like, you know, going to builder mode and then you, you pause and
• 36:18
  Oh, no, someone just like fix this over there.
• 36:22
  So like, I got to go like grab jump on that train now.
• 36:24
  So yeah, it's, it's happening, man.
• 36:26
  I think this, it's good that we're having this conversation
• 36:29
  because I, it's, it's often interesting to me how I do meet
• 36:34
  founders, how I do meet very savvy CE OS at, you know, growth
• 36:40
  companies as well, who aren't really recognizing the, the
• 36:46
  opportunity that exists. Like right now, like the second,
• 36:49
  right? They're still investing like in
• 36:51
  a lot of these more traditional like marketing tactics, like
• 36:55
  quite honestly, and no offense to any friends of mine out
• 36:57
  there, but like, like going to these giant trade shows, quite
• 37:01
  honestly, it's like, I get it. Like there's a, there's a level
• 37:05
  of, you know, it's, it's cool that you can like connect with
• 37:08
  people in real life. I think that's fine.
• 37:10
  But boy, it's, it's, it's, it's very resource intensive versus
• 37:15
  like coming in with like more of like a, a scalpel and like kind
• 37:18
  of figuring out how to find, you know, how to, how to generate
• 37:22
  that conversation differently. So yeah, I think, I think
• 37:25
  there's definitely going to be, you know, there, there's
• 37:29
  definitely going to be a group of people who don't transition
• 37:33
  fast enough to, to AI agent marketing, quite honestly, or
• 37:36
  vibe marketing, if you want to call it that, versus those who
• 37:40
  kind of stick with the, with the traditional, you know, approach.
• 37:44
  And I can already see it, I can see in a couple of places.
• 37:48
  But I'm convinced that those who jump on the the AI stuff sooner
• 37:52
  than later will likely likely jump out ahead within the next,
• 37:55
  you know, few months. I think it's a real thing.
• 38:01
  One thing that's fascinating and I'm trying to grapple with in my
• 38:07
  mind is how we have this whole systems and very much like
• 38:13
  marketing OPS type thing where you are tactical and you're
• 38:22
  creating very specific and like you said, almost like the sniper
• 38:30
  type of bullets that are going out.
• 38:33
  But then you also have the thought leadership in the brand
• 38:35
  building and the stuff that I think goes a really long way,
• 38:43
  especially because, as you said, 90% of folks that are finding
• 38:50
  out about you or or trying, you probably aren't in the buying
• 38:54
  cycle. Yeah.
• 38:55
  So you want to figure out how you can keep that conversation
• 39:00
  going with them until they are in the buying cycle and you can
• 39:03
  become that trusted advisor. And that's where I feel like you
• 39:09
  cannot use AI in that type of thought leadership per SE, or at
• 39:18
  least maybe in the way that I'm thinking about it, which would
• 39:21
  be a podcast, for example, or going to maybe it's giving talks
• 39:28
  on stages is also a little bit harder to use AI in that regard.
• 39:36
  AI can help you write the talk. AI can help you distribute the
• 39:40
  podcast or create clips from the podcast, but it's not
• 39:43
  necessarily like you can use notebook LM and now you have
• 39:49
  thought leadership because of your notebook LM podcast.
• 39:54
  Right, right. But this is what I would say to
• 39:55
  that. So 100% I would say 100% that
• 39:59
  the key to using. You know, AI agents in your, in
• 40:03
  your workflow and in marketing and frankly in business is, is,
• 40:07
  and This is why you know, it, it's, it's, it's assistive
• 40:11
  versus replacement, right? And, and it's because you still
• 40:17
  need that authenticity, right? There's no replacing it.
• 40:19
  Like people respond to other people, like people, you know,
• 40:23
  our human beings are not human doings, right?
• 40:25
  So they're looking for connection.
• 40:27
  They're looking for, you know, building relationships.
• 40:30
  They want to build trust. And So what I see is that, you
• 40:35
  know, AI and, and these agents are, are just really helping you
• 40:39
  to distill information. So for example, I'm working with
• 40:41
  a startup here at Oregon State University, where we're filming
• 40:44
  from today. And they have a really cool
• 40:48
  computer vision technology that allows you to, you know,
• 40:54
  measure, monitor and and manage aquaculture.
• 40:59
  So if you're, if you're unaware, fish farming or seafood farming
• 41:04
  is more than half of all of the world's food food supply, which
• 41:08
  is absurd. Like it's crazy to think about,
• 41:09
  right? And already 85% of the wild food
• 41:15
  seafood out there is already basically is, is, is no longer
• 41:19
  available, right? So only like 15% of the world's,
• 41:21
  you know, can go with wild fish. But they were, you know, we were
• 41:25
  trying to figure out, they were trying to figure out like doing
• 41:27
  all these like customer interviews, like what to build
• 41:30
  and how, who, how to focus their go to market.
• 41:32
  So on the research side again, and so we conducted, and this is
• 41:36
  where, you know, the AI part came in a, we were able to use,
• 41:41
  you know, AIA just to go out and find these types of operations,
• 41:45
  right? Where they have large enough
• 41:46
  operations that they could use technology to count fish or
• 41:49
  count sea urchins or count, you know, whatever.
• 41:52
  And so we were able to get in, you know, get conversations with
• 41:55
  these CEOs or these, you know, these kind of seafood farmers
• 42:01
  all over the world. Like that's amazing, right?
• 42:03
  And, and so we interviewed them using Microsoft Teams.
• 42:08
  And so we got the transcripts, right?
• 42:10
  And we fed those into our into our AI system and they brought
• 42:16
  back like, here are the concepts that you should focus on.
• 42:19
  And so to your point, like, I don't know, you know, Demetrios,
• 42:21
  like when you come up with, you know, who your guests are going
• 42:24
  to be or other topics you should be talking about, You know, I
• 42:28
  don't know how you, you know, decide that maybe you're just,
• 42:30
  you're probably a genius is probably a big part of it.
• 42:32
  But also it's like, but how do you like take all those
• 42:35
  conversations you have and synthesize them?
• 42:37
  Like now we have gong and now we have all these other things that
• 42:41
  can like do that for you, but you still need to marry that
• 42:44
  with other, you know, information, right about, you
• 42:47
  know, these macro moves and these other things that are
• 42:50
  going on, right. And I think that's the part
• 42:53
  where, you know, these AI agents can help you craft a better, you
• 42:57
  know, message and and think about what you actually should
• 43:00
  be talking about as a thought leader to your to your point,
• 43:05
  right. So you're always kind of staying
• 43:06
  on the cutting edge. That's.
• 43:10
  Cool, but again, it doesn't abstract away the actual time in
• 43:19
  conversations with folks. It is more the originality or
• 43:24
  the ability to come up with new ideas.
• 43:27
  You have a sparring partner now that you can call up or spar
• 43:33
  with at any moment of the day. Yeah, yeah.
• 43:36
  I mean, it's a, it's a great, you know, what's the famous
• 43:40
  line? Like Steve Jobs says that, you
• 43:42
  know, computers are basically a bicycle for your mind, right?
• 43:46
  I mean, this is really, this gets to that, right?
• 43:48
  I think more so than anything like it's, you know, working
• 43:51
  with these AI agents and and obviously ChatGPT being, you
• 43:55
  know, another example that it's, it really is like it's a it's an
• 44:00
  exercise for your mind. Like you get much deeper
• 44:03
  insights. I hate to use that word, but you
• 44:05
  do you get much deeper insights from working with this stuff
• 44:09
  than you would, than you would normally.
• 44:11
  So yeah, I, I think that's you're, you're totally right.
• 44:13
  I don't think it's a replacement though.
• 44:15
  I've, I've come to realize that like I get pretty excited with
• 44:19
  tech like, Oh yeah, we can just like ramp this up and we can do
• 44:22
  all AISDRS and never have to, you know, pay a Commission
• 44:25
  again. We don't need sales people.
• 44:26
  No, that's not that's not true at all.
• 44:28
  Like there's a big like product LED growth.
• 44:30
  Everyone wants to be like data dog.
• 44:32
  Everyone wants to be like sneak. Everyone wants to be like Docker
• 44:36
  and like, Oh yeah, we don't need sales people.
• 44:37
  Like we're just going to put our product out there.
• 44:40
  It's going to be great. It's going to sell itself.
• 44:42
  You know, we're just going to put all of our money and
• 44:44
  resources in a product development and we'll never have
• 44:46
  to deal with sales and marketing people ever again.
• 44:48
  Thank goodness. No, it's just not.
• 44:50
  It's just not real. People buy from people.
• 44:52
  I really believe that, you know, and whether they're buying, you
• 44:56
  know, whether you talk to them or not, they're still, you know,
• 44:59
  looking up the company's brand and they're still thinking
• 45:02
  about, you know, who is behind the product, right.
• 45:05
  And so again, like, like you said, like if you're not out
• 45:07
  there, you know how, you know, basically nurturing or, or
• 45:11
  creating your own personal brand as a founder, right, then yeah,
• 45:16
  you're, I think you're likeliness of seeing that type
• 45:18
  of growth is going to be stimmied.
• 45:20
  I don't think it's going to be the same.
• 45:23
  One thing that I wanted to bring up was that when you think about
• 45:29
  and you hear this trope, it's almost cliche these days on
• 45:32
  like, oh, now we're not going to need humans to do that.
• 45:35
  And it's very clear to all of us that you still need humans to do
• 45:42
  stuff. But what is happening that I
• 45:45
  have been seeing in a few friends reference to me was that
• 45:50
  you're now looking at when maybe there was traditionally a Python
• 46:00
  developer that would create some kind of a mock up in Streamlit.
• 46:06
  All of a sudden because of cursor, they can have three
• 46:11
  different types of MVPS that are built with React.
• 46:18
  You know, it's like you're not now magically freed up with your
• 46:23
  time. I think things just accelerate,
• 46:25
  man. I think they just accelerate.
• 46:27
  So for example, with this, with this startup, they're called C
• 46:30
  Stereo. But with this startup, I was
• 46:34
  like, they, they have a team of, of developers and engineers and
• 46:38
  you know, they people spend, they're spending, you know, a
• 46:42
  good portion of their time on like, I would argue like mundane
• 46:46
  stuff, you know. And so I think like what you're
• 46:47
  describing to me is a lot of the remedial work.
• 46:51
  And I, you know, I don't mean to be sound like derogatory or
• 46:53
  anything when I say that. I just mean a lot of the grunt
• 46:55
  work of like standing up a website and creating like a UIUX
• 46:59
  for like a dashboard, like those sorts of things.
• 47:02
  Yeah. Like I don't think a lot of
• 47:04
  people necessarily want to work on those things.
• 47:06
  And so, for example, I was we drove out to one of these, you
• 47:13
  know, kind of design partners, basically like a, an early
• 47:16
  customer of, of our company. I just went and repple it and
• 47:19
  I'm like, all right, describe to me like what you'd want to see
• 47:22
  as your dashboard. And I, I kid you not, like I
• 47:24
  I'll send you the link. Like it basically created an
• 47:28
  unbelief and I said I wanted to look like that.
• 47:30
  I don't know if you've ever heard of that company called
• 47:32
  front Point, but it's basically like a do it yourself security
• 47:35
  software company where you go around you like put in all the
• 47:38
  different sensors and anyways, it has cameras and things
• 47:41
  whatever. And so like I said, yeah, look
• 47:43
  at this UI and make a version of of this for like aquacultures.
• 47:48
  Dude, it spun it up in like less than an hour.
• 47:50
  It was unbelievable. And the guy was like, Oh yeah, I
• 47:51
  like this, I don't like this. So we just tweak it.
• 47:54
  And then that gives, you know, the the engineers actually
• 47:57
  something to build too, right? So, yeah, I think, I think with
• 48:01
  marketing like it's, it is, it's like vibe coding and vibe
• 48:03
  marketing. It's is lowering the threshold
• 48:06
  for experimentation, which is something that I've always
• 48:09
  loved. I love experimenting.
• 48:11
  And I think that's if we can experiment more, we tend to be
• 48:15
  able to find ground truth that much sooner, you know, and
• 48:20
  that's what it's all about. So I think a lot of these, you
• 48:22
  know, breakthroughs are really allowing us, like I said, to
• 48:27
  spend more time thinking much more deeply than we ever have
• 48:32
  about these user experiences or these interactions or our
• 48:36
  messages or even the products that we're building, right?
• 48:40
  And that's, I think what that's what excites me probably the
• 48:42
  most, you know?