YT YouTube2Text

YouTube2Text → Transcripts → Marketing Engineer: The $1M Job with AI Agents

Marketing Engineer: The $1M Job with AI Agents — Transcript

by Greg Isenberg · 5,845 words · 859 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00I think one of the most valuable people
0:02in tech over the next 18 to 24 months is
0:04going to be something called a marketing
0:07engineer. Now some people call it a
0:09forward deployed marketer and some other
0:11people are calling it an AI growth
0:13operator. I'm saying call it whatever
0:15you want. The name is probably going to
0:17change but the job won't. It's the
0:20person who can do a whole marketing team
0:22work with AI agents. And I think there's
0:25going to be a ton of money to be made in
0:27it. I actually think this becomes a
0:29250k, 500k, a milliondoll job because
0:33every company wants more leads. They
0:35want faster experiments. They want
0:37sharper positioning and they want to
0:39read on their customers and they want
0:41their just marketing to get smarter
0:43every week with a smaller team than a
0:46bigger team. Whoever can walk in and
0:49just build that with AI agents are going
0:51to get to name their price. So, if
0:54you're a marketer, this is how you
0:55become way more valuable. If you're a
0:58founder, you know this. You don't just
1:00want to vibe code something. You want
1:02people using your product. So, you're
1:04going to have a huge edge if you can use
1:06AI agents to do your marketing for you.
1:09By the end of this episode, you're going
1:11to know what a marketing engineer
1:13actually does. What do they build? What
1:16the tool stack looks like? how to use
1:19things like Grockbot and Claude and
1:21Codeex and Hermes and creative models,
1:23how they all play together within the
1:25context of a marketing engineer, and the
1:28exact 30-day plan I'd follow to learn
1:31from scratch if marketing engineering is
1:33interesting to you. Let's get into the
1:35episode. I can't wait to see what you
1:38[music] build.
1:46So something I can't stop thinking about
1:49is marketing keeps changing and having
1:53been a part of multiple cycles. I've
1:56started and sold three ventureback
1:57companies. You know, one was in the web
1:59era, one was in the social era, one was
2:01in the mobile era. Every time the
2:04technology changes, the most valuable
2:07kind of marketer changes with it. So,
2:10think about the traditional era of
2:12marketing. I actually think about it as
2:14like the Don Draper era where marketing
2:17was about making people care through the
2:19story, through the psychology, getting
2:22your product in front of people on
2:23whatever channels existed at the time.
2:26Things like traditional print media and
2:28radio. The best marketers at that time
2:31understood what people wanted, what they
2:33were insecure about, who they were
2:35trying to become and how to package a
2:37product so the market paid attention.
2:39Obviously, that skill matters a lot. But
2:42then the internet showed up and it
2:44created the digital marketer. So it
2:45evolved from traditional to digital.
2:47Suddenly you had websites, email, SEO,
2:50Google. In 2005, I think six, you had
2:53Facebook ads, landing pages, pixels. The
2:56best marketer became the person who
2:58could acquire customers through channels
3:02you could actually measure. And a lot of
3:04people didn't know these were new
3:05channels. So the the best marketers
3:08understood funnels targeting these new
3:10channels, analytics, things like Google
3:12Analytics and the very practical
3:14question about what happens after
3:16someone clicks.
3:18Then software created loops and then
3:22growth hacking became a thing around if
3:24I remember correctly 8 9 10 11 the best
3:28growth uh hacker people were all about
3:31activation referrals onboarding
3:33retention pricing there was a guy by the
3:35name of Dave Mccclure had this I think
3:37it was called the R framework activation
3:40and referral um that was the you know
3:44the marquee er uh the the the symbol of
3:47the time of the growth hacker era.
3:49Basically, marketing moved closer to
3:52product because the product itself could
3:54become the growth engine. Now, we're
3:56walking into the marketing engineering
3:59era. And I feel like not a lot of people
4:01have spoken about this. That's why I
4:02want this to be the deacto episode about
4:05this whole era. The marketing engineer
4:09still needs all that old stuff. It still
4:11needs, you know, customer understanding,
4:13judgment, positioning, understanding
4:15distribution, uh, taste. Um, if
4:19anything, taste, you know, people talk
4:21about this all the time, but taste
4:23matters more now than ever because AI is
4:25about to make average marketing
4:27marketing just unbelievably cheap. The
4:30new part is that the marketing engineer
4:33also builds the system behind the
4:36marketing. So the marketing engineer is
4:38connecting uh customer data uh reading
4:41the results uh shipping little landing
4:43pages and uh you know calculators and
4:46then turning raw customer signal into
4:49content outbound positioning and product
4:52ideas. So the way I think about it is
4:55traditional marketing was about you know
4:58making people care. Digital marketing
5:00was acquiring customers through
5:02measurable new channels. Growth hacking
5:04was about using product and data to
5:06build these loops. And marketing
5:08engineering is about using AI, agents,
5:11data, code, and taste to build a
5:13marketing system that keeps learning.
5:16Um, and the last phrase is an important
5:18one because a marketing system that
5:20keeps learning is now actually possible
5:22in the agentic era. Now, most companies
5:25already have pieces of this lying around
5:28uh to their credit. So they've got, you
5:30know, tools and dashboards, calls,
5:32content, calendars, CRM, some SAS tools.
5:36Um, the problem is the learning is is
5:39pretty scattered. Um, you know, sales
5:42might hear one version of the market,
5:43support hears another. Uh, product sees
5:46the usage and marketing sees what got
5:48clicks and the founder remembers, you
5:51know, the one customer call that just
5:52hit him emotionally that week and just
5:54can't get that one customer call out of
5:57his or her head. I know that happens to
5:59me. Uh then everyone walks into the
6:02growth meeting with a slightly different
6:03version of reality. So the marketing
6:06engineer's whole job is actually to pull
6:08in these signals into one system and
6:12turn them into growth. So the way I
6:14define the role is this. uh a marketing
6:16engineer's you know is a marketing
6:19engineer is the person who turns market
6:23signal into pipeline using AI agents
6:26data code taste and that's really the
6:29job um and I'm going to get you know
6:31super tactical on how you can actually
6:32do this soon um if I were a founder
6:35right now uh the question I'd be asking
6:37myself is who be who on my team would be
6:41building the growth system for this
6:42company now I am a founder my uh myself
6:45so a lot lot of time I'm doing this
6:46myself. Um, and I just hope that you
6:49know if you're a founder listening here,
6:51uh, either you hire someone or you do it
6:53yourself. Um, and you know, because the
6:55companies that are going to win in this
6:56agentic era are going to learn the
6:59market faster than anyone else. So, it's
7:01kind of it's crucial to know. So if you
7:03see the customer pain earlier, you spot
7:05the winning language earlier, you're
7:07testing more angles using fed uh
7:09Facebook ads, shipping more surfaces,
7:12lead magnets, and understand what's
7:14working before the competitor even
7:15notices things, you have this unfair
7:18advantage. So the question I get asked a
7:21lot is, okay, but what is the first
7:22thing I would build? Okay, I want to
7:24become a marketing engineer. I want to I
7:26want to do more marketing engineering.
7:28What do I build first? And the first
7:30thing I would build is a growth repo.
7:33Yeah, I know it sounds a little bit
7:34nerdy. Um, but you know, even if you're
7:37non-technical, I believe you can do it.
7:40So, you're going to want to go and
7:41create a GitHub repo. Um, or honestly
7:44just a structured folder. Uh, you can
7:46call it something like growth OS. And it
7:49becomes a place where the company's
7:51marketing memory is going to live. The
7:54problem it's going to be solving is that
7:56most people use AI in these random
7:58chats. So they'll open up a chat GBT or
8:00Claude Gemini. They'll ask for 10 posts
8:04and maybe they'll copy and you know copy
8:06one into a doc that they like and then
8:08the work just disappears. Next week the
8:12AI is starting from scratch again. Uh
8:14when what it really needed was the
8:16performance data and the founders voice
8:18and the objection from the sales calls
8:20and the language that actually created
8:23replies. So the growth repo is going to
8:26fix that. Um, and inside it, what we're
8:28going to have is a customer truth
8:30folder, and that's going to have our
8:33sales calls notes or support tickets,
8:35maybe some churn notes, interviews, um,
8:38even uh, live product feedback can go in
8:41there. So, you've got uh, a content
8:43engine folder with the founder voice
8:46guide with the winning hooks and the
8:49scripts and notes on what performed
8:50before. You've got an outbound engine
8:53folder with the ICP, your ideal customer
8:56profile. Uh the account research, the
8:59trigger events, maybe some approved
9:01angles could be good to have there. Even
9:03actually ban uh band language is good to
9:06have as well. Um because you know AI
9:10outbound gets weird fast. If you let it
9:12talk like an overexited SDR who just
9:15discovered personalization, you know,
9:17sometimes bad things could happen. So,
9:19you've got a creative testing uh folder
9:22for ad angles um and things like landing
9:26page tests and hooks and offers and
9:29results. And you've got an agents folder
9:31where you define the jobs your AI
9:33workers do. And that repo is the
9:37difference between hey AI helped me make
9:40a thing and AI is helping the whole
9:43company get smarter. That's how a growth
9:46or a marketing engineer uh you know is
9:49thinking about it. And then the prompt
9:52gets way better. So instead of hey you
9:55know write me 10 LinkedIn posts, you say
9:58you know read me read the customer truth
10:00file, read the founder voice file, read
10:02the last five uh posts that drove
10:04qualified replies and draft five new
10:07posts around Payne's buyers that were
10:10actually mentioned this week. So, it's a
10:12totally different level of output. Um,
10:15because the agent is now having real
10:17context. What tools do I need if I want
10:20to become a marketing engineer? Well,
10:24I'll tell you some of the most important
10:25ones and how to think about, you know,
10:27where all the tools fit uh and your tool
10:30stack. So, you know, Grockbot is new,
10:32but it's just an incredible uh product.
10:35So I think of Grockbot as the growth
10:39operating system that lives close to the
10:41internet. So marketing is a living
10:43system. The marketing is moving.
10:45Competitors are moving. Culture is
10:47changing. Customers are changing their
10:49language. Uh creators are picking up new
10:51formats. Um you know Grockbot is
10:54especially useful in that world because
10:56it is connected to the X ecosystem. If I
11:00were setting this up as a founder, I'd
11:02give it a few clear lanes. So, I'd say
11:04one bot watches competitors and tells me
11:06what changed. One is going to watch
11:08customer language across X and Reddit.
11:10One watches the creators in the niche
11:12and finds formats worth testing. And one
11:15watches ads and landing pages. Um, you
11:18know, [clears throat] basically wherever
11:19there's a connection to the internet,
11:22you know, Grockbot is going to be extra
11:24good there. That doesn't mean you can't
11:26use Grockbot to do everything. You
11:28totally can. Um, and I think, uh, you
11:32know, I'm one of those people that, you
11:34know, say like, you know, basically, you
11:36know, pick an ecosystem that you like,
11:38that you feel comfortable with. If
11:39Grockbot feels good for you, you know,
11:41just do everything in there as well. The
11:45way I think about it, this is just the
11:46way I'm thinking about it. So, uh, hope
11:49it gets the creative juices flowing. You
11:51know, for me, I use Claude and Codeex
11:54and products like that in in a different
11:56part of the system. So, they're going to
11:58help me build the repo and generate the
12:00landing pages, writing scripts, building
12:02the little internal tools that I was
12:05talking about. Um, and then, you know,
12:07turn that repeatable work into something
12:09durable. Um, you know, I've talked on
12:12this channel about Hermes before. Hermes
12:14style workflows are still extremely
12:16valuable, uh, especially when you want
12:19scheduled operations with memory and
12:21approval. So something like every Monday
12:24morning build me a market brief or every
12:26Friday afternoon review the experiments.
12:29Um every time a fresh batch of sales
12:32calls land, you know, maybe put it in a
12:34folder and then pull the objections and
12:36update the positioning file. Then you
12:39have creative models. Then they're going
12:41to help you move faster on ads,
12:43thumbnails, mockups, and video concepts.
12:46Um there's a bunch of those that exist.
12:49There's foul AI, there's Higsfield,
12:52there's a bunch of them. And local AI
12:55matters when the data is, you know,
12:57particularly sensitive or there's p
12:59private customer transcripts or
13:01regulated notes, pricing plans,
13:03basically anything a company would feel
13:06weird sending into a cloud tool. Um,
13:10also things that are expense, too
13:11expensive to do into a cloud tool. I'm
13:13going to do a whole separate episode on
13:15local AI. So, so stay tuned for that
13:17over the next one or two weeks, you
13:20know, and subscribe. Uh, so that comes
13:22into your feed. The tools are going to
13:24keep changing, but the workflow is the
13:27thing to actually learn. So, where it
13:30gets really interesting is when the
13:31agent connects to live business data and
13:35the tools obviously to actually do the
13:37work. So, take SEO content. The beginner
13:42version is asking an AI to write a blog
13:45post about a keyword. So, a marketing
13:47engineer isn't going to do that. A
13:48marketing engineer is going to check
13:50Google Search Console, pulling, you
13:53know, keyword data from Hrefs or or SEM
13:56Rush, look inside this CMS to see, you
14:00know, if it already exists. It's going
14:02to rank opportunities by volume and by
14:05buyer intent and it's going to research
14:08what's already ranking uh in you know
14:11hopefully adding the founders's point of
14:13view and draft the post write the
14:16metatitle suggest internal links and
14:19just send the whole thing for approval.
14:21Um that's a pretty big jump but you know
14:23the agent has a job the job has inputs
14:26and the inputs come from the business
14:29and the output goes somewhere useful.
14:32So, every agent is going to need a real
14:35job spec. And I'd write it out almost
14:38like I was hiring a person. Here's the
14:40data sp here's the data source. Here's
14:43when you run it. Uh here's what you
14:45filter out. Here's the output I expect.
14:47Like here's what good looks like. Here's
14:49what's going to need human approval.
14:51Here's the metric that matters. And
14:53here's what you write the result so the
14:56system gets smarter next time. So for
15:00you know maybe a competitor engager
15:02agent that might be every weekday
15:05morning check these 20 LinkedIn accounts
15:08and pull the people who commented on new
15:11post enrich them drop the drop the bad
15:15fit leads and draft 10 messages tied to
15:18a specific post they engaged with. Oh
15:21and then obviously write that write the
15:23results to a file for approval. The
15:26metric is going to be positive replies
15:29from qualified accounts because a
15:31marketing engineer cares about business
15:33results, right? Not activity counts. Uh
15:36messages sent is activity. Quality
15:39qualified replies is going to be your
15:41signal. And the whole point the
15:43marketing engineer is trying to do is to
15:45generate pipeline demand. And you train
15:47these agents the same way you train a
15:50new hire. You start with small tasks.
15:52You watch it work. You correct the
15:54mistakes. You add the correction to
15:56memory because now we have memory and
15:58then you expand the scope as you
16:00increase your comfort level. If the
16:02outbound agent writes a first line that
16:05sounds like fake, for example, you got
16:08to add the rule to the repo. And if the
16:10content agent keeps writing these
16:12generic intros that sound like generic
16:14AI, you give it three good examples and
16:17three bad ones. If the customer truth
16:20agent makes a claim with no evidence,
16:22you know, we got a problem here. You got
16:24to add the rule to that. Every insight
16:28needs a quote or a link or a source.
16:31Every correction becomes part of this
16:33operating system, this growth, you know,
16:36marketing engineering uh uh system. And
16:40that's how this whole thing compounds.
16:42And going back to like how does a, you
16:45know, marketing engineer make a million
16:47a year or $500,000 a year or $1.5
16:49million a year for their own startup.
16:51It's because they're building this and
16:53it's so darn valuable. But let's let's
16:56actually get into a concrete example so
17:00that just this gets solidified into your
17:03head. So imagine a vertical SAS startup
17:07selling software to commercial HVAC
17:10contractors. These are companies
17:12managing technicians and service calls
17:15and maintenance contracts
17:18uh and and dispatch. So, it's a real,
17:21you know, B2B market. Um, the buyer has
17:24a lot of money. The workflows are messy
17:26and the language is specific, which is
17:28why I wanted to use this uh example. The
17:31marketing problem for that company is
17:34usually a little bit more sharper than,
17:36hey, we need some more content. Um, the
17:39real problem that they're facing is
17:41usually something like which pain gets
17:43the owner to take a a demo. Uh, maybe
17:47it's dispatch chaos, or maybe it's late
17:49invoices. Maybe it's that the owner has
17:52no idea which jobs were profitable until
17:55the month is over. Or maybe it's
17:58actually that the technician finishes a
18:00service call, spots a replacement
18:03opportunity, and the follow-up quote
18:05just never gets sent. Um, that's
18:07interesting just because it's really
18:09specific. So, when you have something
18:11specific, you know, it's it's just
18:14interesting.
18:16you know, my bunny ears go up. Stop
18:18losing replacement revenue after every
18:21service call is obviously a much sharper
18:23angle than run your HVAC business
18:26better. So, this is where your marketing
18:28engineer is going to earn their keep,
18:30right? It's going to start with the
18:31customer truth system. That first system
18:34is going to be the customer truth
18:35system. And every startup says they
18:37understand the customer and you talk to
18:39five people and they get five different
18:41results. We talked about that. But the
18:43marketing engineer is going to pull
18:44those signals into one place. The output
18:48is a file called what the market is
18:51telling us.md. I tweeted about this
18:53idea. It went viral. I'm glad people
18:55liked it. It's basically this a markdown
18:58file which updates every morning or
19:00every week depending on how much signal
19:02the c uh the company is going to have.
19:04And it's reading the sales calls and
19:06support tickets and churn notes. Uh even
19:09stripe movement. Um, oh, CRM notes is a
19:13good one. Uh, and also social data,
19:16especially if it's more consumerry, and
19:18its whole job is to show what's changed.
19:21So, maybe the buyers are using a
19:23different phrase than they were you
19:25using a month ago, or maybe the trial
19:27users keep getting stuck before they
19:29invite a teammate. You're just going to
19:30get some insight
19:33and you're going to ask the agent to
19:36show quote snippets, uh, ticket links,
19:39event counts. Um, what you don't want is
19:43obviously a vague summary, which a I've
19:45seen a lot of people do this. They just
19:46get these summaries and it's pretty
19:48vague. Like in this case, you'd get
19:50something like u customers want better
19:53collaboration.
19:54You want something way more sharp than
19:56that. I want the thing that's going to
19:58make the business, you know, harder to
20:01lie to. So, for the HVAC company, you
20:03know, good memo might be something like
20:05five sales calls this week mentioned
20:08emergency dispatch, but the calls that
20:10actually converted all talked about
20:13missed follow-up quotes after the tech
20:16left. Um, just a lot sharper.
20:21The second system is the founder content
20:24engine. So, a lot of companies have uh
20:27raw material, great raw material, like
20:29the founder has opinions. Um, and you
20:32got customer stories. Um, but you know,
20:36you're not really capturing all the
20:37stuff. So, the marketing engineer could
20:39build the loop. So, you can record
20:42founder founder talking to customers.
20:45You can pull from podcasts, extract the
20:47strongest ideas, and then have the
20:50system watch what performs. you know
20:52which hooks you know people keep
20:54watching because you have this data
20:56right and then you create uh content out
20:58of that for the HVAC company um imagine
21:03something like the loss replacement
21:05revenue insight becoming you know five
21:09things a founder post about the hidden
21:11revenue leak in service businesses a
21:13short video on why contractors lose
21:15money after the first visit a landing
21:17page line that says every completed job
21:20uh should create the next quote a cold
21:22email angle could be good and a simple
21:25calculator that estimates the loss
21:26revenue. Um, so the system here is
21:30learning the the third system is the
21:32outbound signal engine. So bad outbound
21:36uh usually starts with a spreadsheet a
21:39spreadsheet full of names. But good
21:41outbound starts with timing. So, who
21:43just raised money? Who's hiring for the
21:45exact problem you solve? Who posted
21:48publicly about, you know, a pain point?
21:52Um, and then who fits your ICP and has a
21:55real reason to care this week. You know,
21:57these people are, you know, they they've
22:00got the pain. You're selling
22:01painkillers, not vitamins with with when
22:04timing uh hurts.
22:07So having an agent watching those
22:09signals, researching accounts, uh
22:13drafting specific angles and sending to
22:15human for approvals, that's the type of
22:17thing that for the HVAC company would be
22:19awesome. So like watching for
22:21contractors, hiring dispatchers or
22:23opening new locations, getting bad
22:25reviews, and then having the agent
22:26actually go and reach out outbound uh is
22:30going to be huge. The fourth is the
22:32creative testing engine.
22:35So, uh, you know, taking one offer and
22:38spinning up 20 hooks, 10 ad angles,
22:41recording the results, and testing them.
22:43So, a lot of people say like, "Facebook
22:45ads don't work for me." Uh,
22:49yeah, maybe. Or maybe the creative,
22:52you're just not testing enough creative
22:53with the right angle. So, a good uh
22:56marketing engineer, you know, could
22:58create thousands of pieces of creative
23:01uh based on, you know, your positioning.
23:04Um, it's basically like having
23:08uh creative becoming this like learning
23:10system, not really a treadmill that you
23:12actually have to do, you know, have to
23:13do. You're going to have it on repeat,
23:16having these agents go and create
23:17creative based on uh just how the world
23:20is changing and how that data is is
23:23changing too. And again like another
23:26like huge insight around like wow this
23:28is like a new way of doing marketing
23:30marketing engineer. The fifth system is
23:33AI search visibilities. So, you know,
23:35you now have billion I mean there's a
23:38billion plus people using chat GPT
23:40asking ju just chat GPT. I'm not talking
23:43about Google AI uh AI overviews or or
23:46Gemini or Perplexity or Claude. Uh I
23:50just saw Sam Alman said they have a
23:51billion users. It's insane. So you got
23:54to think about whether your company is
23:56even understandable to those systems and
23:59then having agents actually go pull in
24:03that data and actually create content
24:06and optimize your website such that
24:09you're you know ranking high there.
24:11Getting cited by AI is like such a huge
24:15opportunity and something that a
24:17marketing engineer is thinking about. Of
24:19course the sixth system is the growth
24:22cockpit. So you know this is like a
24:24weekly view that tells the team what has
24:27changed and what to do about it. What's
24:29con what content has worked? What
24:31campaign created real conversations?
24:33Which objection came up again? What test
24:35won? How many tests won? What percentage
24:37of tests won? What competitors moved? Uh
24:40what customer pain is getting louder and
24:42what to test next. You know, for the
24:45HVAC company, the cockpit might say
24:47something like, "Hey, you know, this
24:49week the lost uh replacement revenue
24:51angle drove fewer clicks than the
24:53dispatch angle, but twice as many demo
24:56requests from owners with more than 20
24:58tech. So, that's the kind of memo that
25:01if you're an executive, you want to wake
25:03up to that. Um, and that's super super
25:05valuable. So, if you've gotten this far,
25:08what are some ideas on how you've
25:09actually could get can make money with
25:13marketing engineering? And you know, I
25:15think there's a few ways that you can do
25:16it. The first is becoming the person
25:19inside the company. So, if you're
25:21already a marketer or a RevOps person, a
25:24growth person, a creator, um or just
25:27honestly like a curious
25:30um marketing-minded person, this is one
25:33of the clearest ways to become way more
25:36valuable uh because this work sits
25:39directly next to uh revenue. All the
25:43ideas that we talked about, all the
25:44systems that we talked about is things
25:46around creating pipeline, lifting
25:48conversion, uh cutting wasted spend is
25:51huge with things like uh these marketing
25:54agents. And you've got this direct line
25:56to business value. And that's how
25:58someone becomes a $500,000 hire because
26:01they look at it and they're like, "Well,
26:03if I'm going to save $2 million and I'm
26:05going to increase revenue uh this this
26:07amount this much and I'm going to double
26:09the conversion rate, uh that's a huge
26:12huge like it's it's a win-win
26:14situation." Um so you know why in the
26:18original I said I think that there's
26:20going to be people make a million
26:22dollars doing this and I actually think
26:23that's conservative. I think there will
26:25be versions of that of the best
26:27marketing engineers
26:29uh making millions of dollars a year is
26:31because
26:33uh they're going to be just driving
26:35insane amounts of value uh in the same
26:39way that forward deploy engineers are
26:41driving insane amounts of value for
26:43companies right now. The second way is
26:45just do consulting. So you embed, you
26:48know, you create an offer, you embed
26:50with a founder company, maybe it's 30,
26:5260, 90 days. Uh you build one growth
26:55system and then you sell the outcome. So
26:58hey, we'll build your customer true
27:00system and turn it into weekly campaigns
27:03or we'll build your founder content
27:05engine or we'll build your outbound
27:07signal engine. some of these ideas that
27:08we talked about, you just embed
27:10yourself, you build it, uh, and you
27:13charge, you know, 5, 10, $30,000
27:16a month depending on what you're
27:18actually building. The third, uh,
27:21somewhat less talked about is
27:22productized services. So, you can pick
27:24one, uh, one wedge and then repeat it.
27:28So, for example, outbound signal engines
27:31for vertical SAS, it's what you focus
27:33on, or founder content engine for B2B
27:36CEOs.
27:37uh customer truth repos for seedstage
27:40startups before they hire a a full
27:43marketing team. So the tighter the
27:45wedge, the easier it is to sell, deliver
27:48and repeat. And that's like the only
27:50thing that you focus on. That's why it's
27:52called productized service services
27:54because it's not like you're doing
27:55services custom things for everyone.
27:57There's this one thing you do for this
27:59one niche and you charge x amount of
28:01dollars for it. The fourth is software.
28:05Um, I think the biggest outcomes are
28:07going to come from this, but I do think
28:10that I would start with services first.
28:11So, you do the work by hand, you build
28:14the same system for five companies, 10
28:16companies, and you notice the pain that
28:19repeats, and then that's when you turn
28:21it into software. Um, that's and that's
28:24also how you avoid building something
28:26that nobody wants. Uh, the fun part is
28:29all these ideas actually stack together.
28:32You can start by consulting to learn
28:34what actually works. You can notice the
28:36same system every client needs. You
28:38productize it. You eventually turn it
28:39into software like you know set of
28:42agents. Um if I were doing this uh
28:46tomorrow morning, I would keep the first
28:48version almost painfully simple. You
28:51know, I would, you know, build that
28:54growth OS folder. I'd have five of those
28:56files. customer truth, founder voice,
29:00uh, experiments, agent jobs, and then I
29:03would paste 20 real customer notes, um,
29:08or or call call summaries, and then I
29:10would ask the agent to do one job. I'd
29:12say, tell me what's changed, show me the
29:14receipts, suggest one marketing test
29:17that could create pipeline this week,
29:19not next week, not a month from now. And
29:21then build one thing from that output.
29:23you know, for that HVAC company I was
29:25talking about. Maybe it's the loss
29:27replacement revenue calculator
29:30um or something like that. The first
29:32goal is just to prove the system can
29:35turn this messy market data into one
29:39useful uh action. So, if you listen to
29:42this and you're like, "Wow, being a
29:43marketing engineer sounds really cool. I
29:46want to go hone my skills in the next 30
29:48days to become a marketing engineer. Be
29:50it as an employee, as a founder,
29:53whatever it is. Here's the plan that I
29:55would run. Week one, I would do an
29:58audit. So, I'd pick one real company. It
30:01could be yours, a friends, uh, whatever
30:03you can get access to. I would study the
30:06website, the offer, the ICP, uh, the
30:09founders content if there is any. Um,
30:12oh, sales calls and support tickets if
30:14you can get them obviously. Um, and then
30:17you output, you know, a market map.
30:20Who's the customer? What pain do they
30:22describe? What words do they use? And
30:25what would you test first? Uh, what are
30:27they buying instead of your product or
30:30this product? Where's the funnel leak?
30:33And what would you test first? So, what
30:35week one is just studying all that
30:37stuff. Um, week two is the growth repo.
30:41So create it, add the folders and build
30:44your first what is the market telling us
30:46uh markdown file. Uh you can use
30:48whatever tools you like. Could be claw
30:50chat chat gbt grockbot uh gemini local
30:53models whatever it is. Uh the tools
30:56actually matter less than the workflow
30:58here. The goal is to basically just to
30:59turn that scattered signal into the meta
31:02with real receipts and actually just
31:04start uh you know feeling like a re a
31:08true marketing engineer. Week three is
31:11your first machine. So you can pick one
31:13system and actually build it. You know,
31:16it could be the content engine, the
31:17outbound signal engine, a landing page
31:19tester. Obviously, this is going to vary
31:22depending on what the company needs and
31:23wants, but just pick one because you
31:27know you're going to get better outcome
31:29with one uh and one working system is
31:32going to beat five like half-built ones.
31:34And then week four, week four is just
31:36all about results. Like what what
31:38changed? Okay, you did this thing. Did
31:39replies improve? Did meetings get
31:41booked? Did any conversion lift? Uh, did
31:44the founder sound sharper? You know, did
31:46the founder like the post? Um, at the
31:49end of your month, you should have a
31:51case study that sounds something like,
31:53you know, I audited this audited this
31:55company's growth. We built this customer
31:57truth repo I found was like one high
32:00intent pain that they didn't know about
32:02and I turned it into an outbound signal
32:04engine which shipped you know 75
32:08targeted messages got nine warm replies
32:11booked three calls and I documented
32:13everything what I learned um and then
32:16you're showing like a real business
32:18result tangible value um and that's how
32:21you get hired that's how you get clients
32:22and that's how you become credible I
32:25think the Best marketing engineers are
32:27going to feel like part marketer, part
32:30product person, part revops, part data
32:33analyst, part creator, and part
32:35engineer. So they can talk to a
32:38customer, they can build the workflow
32:40that uses that insight. They can write
32:42the positioning. They can wire the
32:44automation. And they can make the
32:46landing page. And they can read the
32:47conversion. And they can set the
32:49outbound agent. And they know when
32:51personalization sounds fake. And they
32:53can use AI to make more. And they've got
32:56the judgment and taste to know what
32:58should exist in the first place. The
33:01agents are going to be a commodity at
33:04some point. Your judgment about what to
33:06point them to is the moat. And that's
33:07the job of the marketing engineer
33:09really. And I think it's going to be one
33:12of the most valuable jobs out there. Um,
33:15if this is if you're a marketer, uh,
33:18this is how you become the person your
33:20company literally cannot run without.
33:22And if you're a founder, this is how you
33:24get agents running your marketing for
33:27you. Um, I think there's a real edge
33:30that uh you can have when you're
33:32actually using marketing agents to
33:35actually grow your startup ideas because
33:38people are still stuck in the old growth
33:41hacker or even even worse digital
33:44marketing era of marketing. I think this
33:46window is open right now. I think a lot
33:49of people haven't built the machine and
33:52I wanted to give you the sauce so that
33:54you can uh internalize it so you can
33:56process it so you get your hands dirty
33:58around building some of these agents,
34:00some of these marketing agents because
34:02it's all about increasing your
34:04probability of success when it comes to
34:06building your own startup. And I thought
34:10that that, you know, hey, if you can get
34:13a promotion, if you can, if you can, you
34:16know, have more fun being an employee
34:18working within uh an organization, why
34:21not? Why not do this? Um, so hope this
34:25has been helpful. Obviously, I could
34:27have gone deeper in so many parts of
34:30this episode. There just wasn't enough
34:32time. Um, but do let me know what you
34:36want me to go deeper in. Is it the
34:37Grockbot point, you know, part? Is it,
34:40you know, different uh, you know, the
34:42markdown files, skills? Uh, you let me
34:45know. I live to serve. I'm here to just
34:48give that information to you. Hopefully,
34:52uh, you enjoy it. Hopefully,
34:55um, it gets your creative juices
34:57flowing. And if you haven't liked,
34:58comment, and subscribed
35:01at this point, I don't know what you're
35:02doing. H, hook it up. You're hooking
35:05yourself up. you're getting more quality
35:07content in your feed, less slop. So, uh,
35:12thank you for giving me your time. Hope
35:15it's been helpful, and I'll see you next
35:17time.

About this transcript

This page contains the full transcript of Marketing Engineer: The $1M Job with AI Agents by Greg Isenberg, generated from the public captions YouTube serves with the video. The transcript has 5,845 words across 859 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.