Latest / AI DTC WTF / From Persona Research to AI Media Buyer: The Full-Stack Agency Automation Playbook

Transcript

From Persona Research to AI Media Buyer: The Full-Stack Agency Automation Playbook · AI DTC WTF

• 0:03
  If you work in D2C and you use AI and
• 0:05
  you're wondering what the
• 0:07
  F is going on every week,
• 0:08
  this is your podcast, the AI D2C WTF
• 0:11
  podcast, your home for
• 0:13
  tactical tips, strategies, and
• 0:14
  ideas that you can implement right now in
• 0:17
  your AI workflows to make
• 0:18
  your brand or agency more money.
• 0:22
  Hey, welcome to another episode of AI D2C
• 0:25
  WTF. So glad to have you
• 0:27
  here. This podcast is all
• 0:29
  about ensuring you have tactical advice
• 0:32
  to take away and utilize in your
• 0:35
  e-commerce agency or brand
• 0:37
  right away. That's really what we're
• 0:39
  doing here at the AI D2C podcast. So I'm
• 0:42
  excited to have you,
• 0:43
  Freddie. For those of you that don't know
• 0:45
  Freddie, he is a German badass on AI,
• 0:48
  somebody that I respect
• 0:49
  and who is absolutely always building
• 0:51
  incredible things. Freddie, thank you and
• 0:53
  welcome to the show.
• 0:53
  Thank you so much for having me, Andrew.
• 0:55
  I'm glad to be here.
• 0:57
  So what have you been up to lately? I
• 1:00
  mean, I know that you're
• 1:01
  doing a lot of creative strategy.
• 1:02
  So have you been building tools on
• 1:03
  creative strategy or what
• 1:05
  kind of stuff have you been
• 1:06
  building that's with AI that's been
• 1:08
  making your life a lot easier?
• 1:09
  Yes, doing lots of creative strategy,
• 1:10
  kind of transitioning away
• 1:12
  from that and just building
• 1:13
  more with AI, helping other agencies
• 1:15
  implement AI. And, you know,
• 1:17
  there are so many, you know,
• 1:19
  important use cases or useful use cases,
• 1:22
  especially nowadays with
• 1:23
  the models getting better,
• 1:24
  Claude being there. And the first thing
• 1:26
  that comes to my mind is just like,
• 1:29
  probably research agent, you know,
• 1:31
  research skill that sits in
• 1:32
  Claude. And it just, you know,
• 1:33
  does persona research. It looks up words,
• 1:36
  it looks up reviews, and it
• 1:37
  basically builds this whole,
• 1:39
  you know, persona sheet that, you know,
• 1:42
  other skills that I'm using,
• 1:43
  such as like a script writer
• 1:44
  or whatever can kind of access. And then
• 1:46
  the outputs of the stuff that I'm getting
• 1:48
  from AI are way more,
• 1:50
  you know, tailored to the
• 1:51
  brands that I'm working with.
• 1:53
  So this is, so you have, so talk us
• 1:55
  through how you're doing this on the
• 1:56
  research agent. So this
• 1:57
  thing, like, give me the tactical
• 1:59
  rundown. So this thing, you've built it
• 2:01
  how, it sits in Claude.
• 2:02
  And what is, how did you, like, how are
• 2:04
  you telling it to go find things? What
• 2:05
  kind of stuff is it,
• 2:06
  you know, doing tactically?
• 2:08
  Yes. So it's a, it's basically a Claude
• 2:10
  skill. So, you know, it's super easy to
• 2:12
  create. You don't need
• 2:13
  any like coding experience or whatsoever,
• 2:15
  really. And it's connected to foreplay.
• 2:19
  Foreplay has basically an MCP server,
• 2:20
  which you can connect
• 2:21
  essentially in one click to Claude.
• 2:23
  Then it can use, you know, foreplay for
• 2:25
  you, basically. It's
• 2:26
  also connected to another
• 2:28
  kind of program that is called Firecrawl.
• 2:31
  And Firecrawl also has an MCP server,
• 2:34
  but you can also connect in one click. So
• 2:35
  everything is pretty
• 2:36
  straightforward. And,
• 2:37
  you know, firecrawl can essentially
• 2:39
  scrape websites for you. So this, you
• 2:42
  know, research agent
• 2:43
  is, you know, connected to Firecrawl,
• 2:47
  connected to foreplay, and then has
• 2:49
  instructions for me,
• 2:50
  you know, and stuff that I want. So for
• 2:51
  example, you know, words
• 2:53
  that the brand uses, you know,
• 2:54
  and it extracts reviews from, for
• 2:57
  example, Trustpilot or the brand's
• 2:58
  website. And, you know,
• 3:00
  it scans competitors, you know, looks at
• 3:02
  competitors ads using foreplay. And then
• 3:05
  at the end comes up with
• 3:06
  different personas. And I basically give
• 3:08
  a thumbs up or thumbs down, you know,
• 3:10
  whether I like the persona
• 3:11
  or not. And then it gets added to kind of
• 3:13
  my database of personas
• 3:15
  for this brand. And then
• 3:16
  from there on, I can reference it in
• 3:18
  other skills that I'm
• 3:19
  running, essentially.
• 3:20
  Cool. Just jumping in here, Andrew, I
• 3:22
  think this is super
• 3:23
  interesting and super cool. So
• 3:25
  I'm not so familiar with the foreplay. I
• 3:27
  played around in Firecrawl,
• 3:28
  which is just amazing to just
• 3:30
  get a lot of data out pretty quickly. But
• 3:32
  can you explain a bit
• 3:34
  more what foreplay is?
• 3:36
  Yes, foreplay is basically just a tool
• 3:38
  that allows, you know, you
• 3:39
  to kind of save Facebook ads
• 3:40
  from the ads library, right? But the
• 3:42
  great thing about it is
• 3:43
  that they have an API or,
• 3:44
  you know, an MCP server that you can
• 3:45
  connect to. So you can automatically
• 3:48
  crawl your competitors,
• 3:49
  you know, ads library, for example, you
• 3:51
  get the transcripts of the videos easily,
• 3:53
  you can download the videos. And then,
• 3:55
  for example, you know,
• 3:56
  downloading the videos,
• 3:56
  you can throw these videos into Gemini,
• 3:59
  which, you know, you can
• 4:00
  also connect with an MCP.
• 4:01
  And then you can basically visually
• 4:03
  analyze the videos, for example, what's
• 4:04
  going on in the hooks,
• 4:05
  you know, what's going on in the first
• 4:06
  three seconds, what kind
• 4:07
  of angles are they using?
• 4:08
  What kind of visuals maybe are they
• 4:09
  using? And what kind of
• 4:11
  text overlay? So, you know,
• 4:12
  this gets super, you know, super helpful.
• 4:15
  And for you kind of
• 4:15
  analyzing competitors.
• 4:17
  Okay, wow. So basically, where Firecrawl
• 4:19
  goes and looks at websites
• 4:21
  and gives you the information,
• 4:22
  foreplay is a way where you can basically
• 4:24
  scrape ads libraries and give you that
• 4:28
  data in a structured
• 4:29
  way so you can work from there.
• 4:31
  Yes, exactly. You can see how long ads
• 4:33
  have been running, etc., you
• 4:34
  know, which is not always a
• 4:35
  telltale sign for, you know, whether
• 4:36
  something is a winner. But especially,
• 4:39
  you know, if it's a video,
• 4:39
  it's been running for quite some time.
• 4:41
  And maybe the ad is running
• 4:43
  like the same kind of concept
• 4:45
  over and over again, you know, I'm pretty
• 4:46
  sure that this is a winning
• 4:47
  concept. And now I can have
• 4:49
  it analyzed using the API inside of my
• 4:52
  cloud, and then kind of
• 4:53
  can, you know, identify winning
• 4:54
  patterns this way.
• 4:56
  So there, so you're going through running
• 4:58
  this research agent, it's
• 4:59
  doing tons of stuff for you.
• 5:02
  It's looking at these tools, what other?
• 5:05
  I mean, one, one thing
• 5:06
  I've always wondered about the
• 5:07
  research tool, looking at Reddit is that
• 5:09
  can it get into Reddit
• 5:10
  threads, like logging in and that
• 5:12
  kind of thing? Because sometimes the
• 5:13
  stuff that's publicly available is not as
• 5:15
  good. And does it actually
• 5:17
  like, what's the output of it? Does it
• 5:19
  come out with different
• 5:20
  angles you can pursue? Like,
• 5:21
  is it a written document? Like how,
• 5:23
  what's the output you have it
• 5:24
  going into? Or is it actually
• 5:25
  just going into creating the ads?
• 5:28
  Yeah. So there's also basically for
• 5:30
  Reddit, there's like a separate,
• 5:31
  basically MCCP server
• 5:32
  that you would connect to. And in terms
• 5:35
  of output that I'm getting,
• 5:36
  and maybe I can like link this
• 5:37
  in the show notes or whatever, but it's
• 5:38
  basically like a 10 page
• 5:39
  document, which may sound overkill
• 5:42
  at first, but especially with context
• 5:44
  windows being so large, nowadays for
• 5:46
  these larger models,
• 5:47
  especially, you know, 1 million context
• 5:48
  windows for Opus, it can,
• 5:50
  you know, get a lot of context
• 5:51
  and it can kind of remember all of the
• 5:52
  concept. So basically the
• 5:54
  outcome is like a brand overview,
• 5:56
  like, you know, the product mechanics,
• 5:57
  the any like studies associated with it,
• 5:59
  what kind of is the overarching brand
• 6:01
  voice? And then like a
• 6:02
  market analysis, you know,
• 6:04
  how is the market sophistication level?
• 6:05
  How are competitors and
• 6:08
  stuff like this? Then again,
• 6:09
  it looks at reviews, it extracts those,
• 6:11
  then like it basically
• 6:12
  saves full-end reviews in like
• 6:13
  another document. Then based from those
• 6:15
  review, it like kind of like
• 6:16
  creates a like language bank,
• 6:18
  which is essentially just like words that
• 6:20
  the brand is or that, you
• 6:21
  know, at least reviews are
• 6:22
  mentioning over and over again. So you
• 6:24
  can kind of try to identify
• 6:25
  patterns from that. Then it looks
• 6:27
  again, it looks at competitors.
• 6:28
  Automatically it finds those and kind of
• 6:30
  sees what kind of ads,
• 6:31
  you know, they have running using
• 6:32
  foreplay again. And then
• 6:34
  based on all of that information,
• 6:35
  it creates, you know, persona documents,
• 6:37
  which feature like the angles, the, you
• 6:42
  know, benefits for that
• 6:43
  specific angle that you can use and stuff
• 6:45
  like this essentially. And,
• 6:47
  and are these then stored as
• 6:49
  just MD files locally? Or do you have
• 6:52
  like your own database of things where
• 6:53
  you're organizing? Or
• 6:55
  how are you kind of organizing the data?
• 6:57
  Because one thing is to have these
• 6:59
  amazing skills do all the
• 7:00
  crawling, but kind of how do you make
• 7:02
  sure you don't lose any data
• 7:04
  on the way? And then can you
• 7:06
  kind of take a look at the big picture?
• 7:08
  Yep. So I think like the first step, you
• 7:10
  know, when you start
• 7:11
  using this, probably just save them
• 7:12
  locally, right? And then whenever you
• 7:14
  basically open up a new chat,
• 7:16
  you know, what I would do is like create
• 7:17
  a project and cloud for your
• 7:18
  brand. And then whenever you
• 7:20
  start in your chat, just insert them
• 7:21
  manually. And over time, that obviously
• 7:22
  gets tedious. So I think
• 7:24
  the next step from there is basically
• 7:25
  taking it to Superbase, which is kind of
• 7:27
  like a, like database,
• 7:29
  it also has an integration, an official
• 7:31
  integration, even with Cloud.
• 7:32
  So you can just connect it to
• 7:33
  Cloud, tell it like, hey, I want to
• 7:35
  create a new database based
• 7:36
  on this, or like you can even
• 7:37
  connect the skill to Superbase. So the
• 7:40
  way that I've set it up is
• 7:41
  like any kind of research doesn't
• 7:42
  even get saved for me locally, it
• 7:44
  automatically gets uploaded to Superbase.
• 7:46
  And because the skill
• 7:47
  has those instructions, like, hey, here's
• 7:49
  the Superbase server I
• 7:49
  want you to use, upload all of
• 7:51
  that information, etc. Okay, so you have
• 7:53
  that. So you're talking about doing all
• 7:55
  this other consulting
• 7:55
  with agencies. And obviously, Persona,
• 7:58
  unlocking Persona is a massive
• 7:59
  thing that people are getting
• 8:00
  into. What are other things that you've
• 8:02
  been building or that you're helping
• 8:04
  agencies instill that are
• 8:05
  like blowing their minds? Because this is
• 8:08
  a big one. If you can
• 8:09
  go to the client and say,
• 8:09
  hey, look, I've got a whole bunch of new
• 8:10
  Persona research, that's
• 8:11
  awesome. But what else are you
• 8:13
  building or messing around with that
• 8:15
  people have really been loving?
• 8:16
  Yeah, a lot of stuff. I think that first
• 8:19
  thing is probably like
• 8:20
  automatic select reports,
• 8:22
  right? So I'm building this for one
• 8:24
  agency who, you know, they want their
• 8:27
  strategist to basically
• 8:28
  kick off, you know, the day with, you
• 8:30
  know, some automatic
• 8:31
  reporting of like, hey, this is how,
• 8:32
  you know, the ad account is going. And
• 8:34
  these type of scripts are
• 8:35
  working well, you know, here,
• 8:36
  maybe some ideas for videos that you can
• 8:38
  do some iteration ideas. And
• 8:40
  so this one is then connected
• 8:42
  to the official like meta API. So it can
• 8:44
  basically, you know, download or it gets
• 8:46
  all of the information
• 8:47
  from the ad account, such as like amount
• 8:48
  spent, hook rate, etc.
• 8:50
  Right. And then saves all of this
• 8:51
  information in Superbase as well. So this
• 8:54
  way basically build like a whole like
• 8:56
  interconnected system,
• 8:57
  right? You save your like Persona's in
• 8:58
  there, you save, you know, your well
• 8:59
  performing ads in there,
• 9:01
  you save the ad copies, the headlines and
• 9:03
  all of that data that
• 9:03
  is associated with this
• 9:05
  ad account in there. And then each, you
• 9:07
  know, morning they wake
• 9:08
  up to like a new report of
• 9:09
  like, hey, this ad is trending, this ad
• 9:11
  was maybe a recent launch, it's getting a
• 9:12
  bunch of spend pretty
• 9:13
  fast. Or, you know, this ad is, you know,
• 9:16
  30% below our CPA threshold, it's
• 9:18
  performing super well.
• 9:19
  Here's some some patterns that we've
• 9:20
  identified, let's try angle X, Y, Z or
• 9:23
  forever, let's do more ads
• 9:24
  for, you know, with the split screen
• 9:26
  hook, that seems to be performing well,
• 9:28
  like in that direction.
• 9:29
  And so this is based on the on what Meta
• 9:33
  is saying, though,
• 9:34
  right? That's the downside of
• 9:35
  it is it's based on their reporting. So
• 9:38
  it's like, it's not a third party or
• 9:39
  anything. But could you
• 9:40
  like cross reference it against, you
• 9:42
  know, North Beam or something like that?
• 9:44
  If the client wanted to?
• 9:46
  I think there is, I haven't played around
• 9:47
  with North Beam, actually,
• 9:48
  but I think there's a North
• 9:49
  Beam API. So you can definitely like play
• 9:51
  around with it. I know also
• 9:52
  that, and I've tested this,
• 9:54
  that Motion is coming out with like an
• 9:56
  API and an MCP server. And so you can
• 10:00
  connect with that as
• 10:00
  well. And Motion is obviously also
• 10:04
  connected with North Beam and North Beam.
• 10:06
  So yeah, I think that's
• 10:07
  basically not going to help in the long
• 10:09
  run. For now, it's, it's
• 10:10
  basically just, you know,
• 10:11
  quote unquote Meta. And I saw, you know,
• 10:13
  it's definitely still more
• 10:14
  actionable than just looking
• 10:15
  at foreplay and like maybe even scraping
• 10:16
  your own ad library.
• 10:17
  Okay. And sorry, one more question. How
• 10:19
  do you like tactically do
• 10:20
  this? Like, what are you using?
• 10:23
  I mean, I have, I know I have a general
• 10:25
  idea, but like, what are
• 10:25
  people wanting to set this up?
• 10:27
  How do, how are they,
• 10:28
  you know, setting it up?
• 10:29
  Ooh, setting this up. So for this, for I
• 10:34
  think an easy version,
• 10:35
  a super easy version is
• 10:39
  getting two minute reports, which is just
• 10:41
  kind of like Supermetrics
• 10:42
  was back in the day. It's just
• 10:47
  you create a report to get the data in
• 10:49
  Google Sheets. And you
• 10:50
  don't even need, you know, to
• 10:51
  connect to any Meta API or whatever. It's
• 10:53
  super straightforward. And
• 10:55
  then you use APFi, which is
• 10:58
  kind of like a scraping website. And they
• 11:00
  haven't, they have a kind of
• 11:01
  official Meta ad scraper as
• 11:03
  well. And then you get the post links
• 11:06
  from two minute reports,
• 11:08
  just give them to APFi. It's
• 11:11
  basically scrapes the, it scrapes the
• 11:14
  like URL of the asset. So you
• 11:16
  can like download the video,
• 11:17
  you can download the image, and then just
• 11:19
  give that to Gemini. And
• 11:22
  then you can basically add
• 11:23
  that data. Like you get the, all of the
• 11:25
  information that you
• 11:26
  need, like it's plus the like
• 11:28
  downloadable files. And you can just add
• 11:30
  that to Superbase. So you
• 11:31
  don't need any like Meta API.
• 11:33
  You don't need to custom code anything
• 11:34
  really. Just pre-stress
• 11:35
  for about two minute reports,
• 11:36
  you know, set up the reports that you're
• 11:38
  looking at anyway, like maybe
• 11:41
  high level account overview.
• 11:42
  And then also you look at like daily,
• 11:43
  daily ad spend or
• 11:44
  whatever. And then you can just add
• 11:46
  that to Superbase. But do you then run
• 11:48
  like a scheduled roaming
• 11:49
  cloud co-work or how do you
• 11:52
  actually trigger these reports? Yeah. So
• 11:54
  for these reports, so two
• 11:55
  minute reports, you can schedule.
• 11:56
  So that runs automatically, you know,
• 11:58
  any, you know, every day, every week,
• 11:59
  however often you want
• 12:01
  to run. And then for getting the data
• 12:05
  into Superbase, there's
• 12:06
  another service that I'm using.
• 12:10
  It's basically called, it's called
• 12:11
  Railway. So what you can do
• 12:13
  on there, you can connect like
• 12:16
  your GitHub and then also basically set
• 12:19
  up a schedule. So it
• 12:20
  also runs every single week.
• 12:21
  And then it, that's for like the Slack
• 12:24
  automation to get the report into Slack.
• 12:27
  Costs like $5 a month. So it's basically
• 12:28
  free. And then it runs
• 12:30
  every week and basically creates
• 12:31
  that Slack report essentially. Okay,
• 12:33
  cool. What I think is
• 12:36
  really interesting when I talk to
• 12:38
  different people like you, Freddie, is
• 12:41
  that what we're learning now
• 12:43
  is kind of the processes and
• 12:45
  what's useful and what's not useful. And
• 12:49
  I feel like the tools are
• 12:52
  constantly changing, but we're
• 12:53
  realizing kind of what is, what do we
• 12:56
  need help with and what we need to get
• 12:58
  done. And of course,
• 13:00
  the models will be better, the MCPs will
• 13:02
  get better. But what I
• 13:03
  think is really cool is that
• 13:05
  it's getting like week by week, it's
• 13:07
  getting more and more useful and we kind
• 13:09
  of nail down what we
• 13:10
  want. And we can just like switch out
• 13:12
  small little tools along the
• 13:13
  way to make sure that just the
• 13:16
  output and the results are just getting
• 13:18
  better and better as
• 13:19
  the technology evolves.
• 13:21
  Yes, 100%. I think at the beginning or
• 13:23
  maybe last year, like lots of
• 13:24
  stuff was maybe a little bit
• 13:25
  unnecessary. I think now it's actually
• 13:26
  turning into like, you know, which
• 13:28
  workflows are maybe like,
• 13:30
  you know, taking up too much time or
• 13:31
  which can be automated and
• 13:32
  then using AR to automate,
• 13:34
  you know, those workflows. And one
• 13:37
  example that I have is also like as a
• 13:39
  creative strategist,
• 13:40
  you know, reviewing ads, there was a big,
• 13:42
  big, you know, chunk of
• 13:43
  the job, I guess, you know,
• 13:45
  and one part of that is just like
• 13:46
  checking for grammar issues on the ads,
• 13:49
  you know, maybe some
• 13:49
  of the transcripts, they have like
• 13:50
  grammar issues or maybe on the ads, maybe
• 13:52
  like a word is spelled
• 13:53
  in the wrong way or whatever. And if
• 13:57
  you're using, you know,
• 13:58
  certain, maybe Google Drive,
• 14:00
  for example, you know, you know, as long
• 14:01
  as they have an API,
• 14:03
  I'm using frame for this,
• 14:05
  is that you can basically automate that
• 14:08
  kind of review part, right? So I'm
• 14:11
  setting up an automation
• 14:12
  right now that is, you know, the brand is
• 14:14
  using ClickUp as soon as it
• 14:15
  gets turned into, you know,
• 14:16
  ready for review or whatever. And we have
• 14:20
  the frame link linked in
• 14:21
  the ClickUp task. And as soon
• 14:23
  as something turns to ready for review,
• 14:25
  it goes into frame,
• 14:26
  basically using the official API,
• 14:28
  looks at the image and or video and kind
• 14:31
  of checks is there if
• 14:32
  they're like any grammar issues.
• 14:33
  Now, it sounds like such a small like
• 14:36
  issue. But in the end, that
• 14:37
  is like, you know, saving me at
• 14:39
  least speaking from my own perspective,
• 14:40
  like a whole lot of time, if I don't,
• 14:42
  manually have to like check the
• 14:43
  transcript or, you know, look at a whole
• 14:46
  two minute long video
• 14:46
  to see if there's like any grammar issues
• 14:48
  going on in that video.
• 14:49
  That's great. I mean, so one thing I'll
• 14:52
  just jump in on a question,
• 14:53
  Freddie. So I think a lot of
• 14:55
  people are thinking about sort of an AI
• 14:56
  media buyer. And I'm curious if you think
• 14:59
  this is close. I mean,
• 14:59
  I know Thomas has done this and Thomas
• 15:02
  can tell you briefly about
• 15:03
  his process. But, you know,
• 15:06
  so the AI media buyer would go ahead and
• 15:09
  look at optimization
• 15:10
  ideas and correlation between
• 15:12
  like in platform, MTA, MMM, and then it
• 15:17
  would be like approved by a
• 15:18
  person. So if you have you set
• 15:20
  this up for people? And if so, like how
• 15:23
  outside of what you just
• 15:24
  talked about in terms of like
• 15:25
  automated reporting, what other things
• 15:27
  would you set up to make sure? And how
• 15:31
  what tools would you use
• 15:31
  to because I think a lot of people are
• 15:33
  thinking about this. It's
• 15:34
  like, bring me the ideas and then
• 15:35
  let me say yes or no.
• 15:41
  AI D2C WTF is brought to you by Motion.
• 15:46
  And this episode is brought to you by
• 15:47
  Runneth from Motion.
• 15:49
  And if you run paid social, I mean,
• 15:50
  obviously, you know, Motion, it's a
• 15:51
  creative analytics platform
• 15:53
  that everybody uses. But the thing I've
• 15:55
  been obsessed with lately
• 15:56
  is Runneth. And it's their
• 15:57
  AI agent that lives right inside of your
• 15:59
  ad data. The simplest way to
• 16:01
  explain it, most people use AI
• 16:03
  by asking it one thing, like, hey, what
• 16:04
  did we do yesterday? How'd
• 16:05
  it go? Runneth is the upgrade
• 16:07
  of that. You just tag it and ask it to do
• 16:09
  almost anything. It goes into your
• 16:10
  account, does it like
• 16:12
  build me this week's creative recap, pull
• 16:14
  the data and write a
• 16:15
  brief on what to iterate next.
• 16:16
  Tell me which ads to make next. And it
• 16:19
  goes on and does this in
• 16:20
  plain English. So two things that
• 16:22
  actually make it really special, right?
• 16:23
  So under the hood, it's Claude, right?
• 16:25
  And anything Claude can
• 16:27
  do, Runneth can do basically even better
• 16:29
  as long as you get access.
• 16:30
  And second, it's wired from
• 16:31
  everything Motion has. So the AI tags,
• 16:34
  the creative, the
• 16:35
  frame by frame breakdowns.
• 16:36
  And then it's trained by $14 billion in
• 16:38
  ad spend on Motion's
• 16:39
  customers, right? So it's pretty insane.
• 16:42
  So the way that, you know, we're using
• 16:44
  it, and I know some other
• 16:46
  colleagues are using it,
• 16:47
  like I looked at what they're doing, and
• 16:48
  we talked through it. People are, you
• 16:51
  know, having it drop a
• 16:52
  morning brief in Slack on how the
• 16:54
  account's performing. It
• 16:55
  gets smarter every time it runs.
• 16:56
  It's less like a tool and more like a
• 16:58
  creative strategist
• 16:59
  working while you sleep.
• 17:00
  So go see it at motionapp.com. That's
• 17:02
  Runneth from Motion. And make
• 17:04
  ads that win without getting
• 17:05
  lucky.
• 17:11
  Yeah, I don't actually have set it up
• 17:12
  yet. But I also saw Zach's
• 17:14
  tweet, I think, about him,
• 17:15
  like automatic media buying as well. I
• 17:17
  don't know if you saw that, Andrew.
• 17:19
  Zach Stacks. But I think we are like
• 17:23
  basically almost there. I
• 17:24
  think a lot of the stuff that
• 17:25
  helps with this is basically giving the
• 17:28
  LLM or the AI context,
• 17:30
  like giving it your own SOPs,
• 17:31
  like basically, like what you would give
• 17:32
  your own media buyer, you
• 17:33
  need to give it to the AI,
• 17:35
  right? And then maybe not the, maybe the
• 17:36
  first run isn't the
• 17:37
  greatest, or maybe not the second,
• 17:39
  but you give the AI feedback. And then,
• 17:41
  you know, it gets the memory, it becomes
• 17:43
  basically a self-learning
• 17:45
  kind of, you know, agent, essentially.
• 17:47
  And then maybe in a month or so, you can
• 17:50
  automate a lot of the
• 17:50
  media buying. And just because every
• 17:54
  brand is different, every
• 17:54
  brand obviously has different
• 17:56
  like media buying principles and what
• 17:57
  they act up on, you know,
• 17:58
  maybe it's CBO, maybe ABO.
• 17:59
  Just make sure that you give good context
• 18:01
  to the AI of like how
• 18:02
  you do stuff. And then,
• 18:06
  yeah, I'm pretty sure you can automate
• 18:08
  media, you know, automate
• 18:09
  media buying to a good chunk.
• 18:11
  For sure. A hundred percent. And I think,
• 18:15
  I think this is definitely
• 18:16
  kind of the next phase of what
• 18:17
  agencies are going to do. So one thing
• 18:19
  you're talking about, we've
• 18:20
  been talking about previously,
• 18:21
  like doing research or finding typing
• 18:23
  mistakes and, and, and all
• 18:25
  that stuff, right? And actually
• 18:27
  pushing the buttons is, is possible today
• 18:30
  through most of these
• 18:32
  different APIs, right? And I think
• 18:35
  we're going to see big, big, big moves in
• 18:39
  that space just within a
• 18:40
  few months. I've been playing
• 18:42
  around myself with, I have six open
• 18:44
  machine, Mac meanies on my,
• 18:45
  in my office around me at this
• 18:48
  moment. And they're doing media buying
• 18:50
  now for, for 10 of my, my
• 18:53
  clients from my agency. And,
• 18:54
  and it's still a little bit scary, but it
• 18:57
  actually is working. And
• 18:58
  we have, of course, our team
• 19:00
  monitoring and checking in everything. So
• 19:02
  I think that if me with a
• 19:04
  little bit of duct tape are able
• 19:06
  to set something like this up, I think
• 19:08
  there's going to be a lot
• 19:08
  of interesting services going
• 19:12
  around the, around this for sure. Because
• 19:15
  you're talking to so
• 19:15
  many different agencies,
• 19:18
  Freddie, what I would like to get your
• 19:20
  kind of gut feeling on is how
• 19:23
  in front of this are agency,
• 19:26
  DTC agencies today? Are they, are they
• 19:29
  afraid of this? Are they
• 19:32
  like surfing the wave? Or are
• 19:34
  they just waiting kind of what's the
• 19:35
  general vibe you get out there?
• 19:37
  Hmm. Yeah, I think they're excited for
• 19:39
  it. Like some of the
• 19:40
  stuff that I've seen,
• 19:40
  like agency owners build is, is super
• 19:43
  crazy to me. Like they've built some
• 19:44
  really advanced stuff.
• 19:46
  And I think everybody's like excited that
• 19:47
  they can maybe also have
• 19:49
  their employees do more important
• 19:50
  work, you know, whereas before maybe they
• 19:52
  had to, you know, name ads or
• 19:53
  like even launch ads or like,
• 19:55
  you know, do media buying, maybe you can
• 19:56
  do more analysis, or maybe
• 19:59
  you can do more like high
• 20:00
  level stuff that's actually moving the
• 20:01
  needle more than like naming an ad.
• 20:03
  Um, or, um, stuff like this. I think the
• 20:07
  agency process is like, is
• 20:09
  coming down to getting rid of
• 20:11
  sort of the, the, the noise. And I think
• 20:13
  a lot, a lot of where, you
• 20:15
  know, I've seen agencies step
• 20:16
  in and say like, okay, I've got to figure
• 20:17
  this out. I got to get
• 20:18
  better at this is a lot of people
• 20:20
  spend time, like you said, getting
• 20:21
  together reports and
• 20:22
  reporting. And I think it, you know,
• 20:25
  sort of wrapping all those into a place
• 20:27
  where a client can take
• 20:28
  action, I think is the next step.
• 20:30
  A lot of it is like the analysis is
• 20:32
  happening. You're giving people these
• 20:33
  reports on personas.
• 20:34
  Um, you're putting it, you're putting it
• 20:36
  in front of them on a
• 20:37
  platter and the client's like,
• 20:38
  oh, cool. And like, really the next thing
• 20:41
  of this is being able
• 20:42
  to say like, yes, like,
• 20:45
  you know what I'm saying? I don't know
• 20:46
  what you think in terms of
• 20:47
  talking to other agencies.
• 20:48
  Um, you know, obviously like, I assume
• 20:52
  these are German
• 20:53
  agencies or maybe they're not,
• 20:55
  but they're probably, if they are,
• 20:56
  they're quite orderly in
• 20:57
  terms of like getting people
• 20:59
  the information that they need. How do
• 21:01
  you see this evolving? Like
• 21:03
  in terms of other tools people
• 21:05
  are using and things that they can, um,
• 21:07
  you know, get on and get
• 21:09
  learning to make sure that they're
• 21:10
  ready for, for this.
• 21:12
  Yeah. Um, so I think like one of the
• 21:15
  agencies that I'm working with, they
• 21:16
  basically have their own
• 21:18
  like custom software, I guess already,
• 21:21
  right. They have
• 21:21
  basically kind of like a place for
• 21:22
  their strategists and where basically,
• 21:25
  you know, those reports
• 21:26
  essentially generate ad copies,
• 21:28
  they generate scripts, um, and stuff like
• 21:30
  this. And then essentially
• 21:31
  they approve of the, you know,
• 21:33
  of those concepts. And if they approve,
• 21:34
  it goes basically to the
• 21:35
  client and, um, you know,
• 21:38
  if the client likes it, they basically
• 21:39
  produce that type of content. So I think
• 21:40
  kind of that's like,
• 21:41
  you know, the evolvement, you obviously,
• 21:42
  you know, you have the
• 21:43
  reports and I think that's
• 21:45
  already pretty helpful, but obviously,
• 21:46
  you know, next step is
• 21:47
  like basically generating that
• 21:49
  meaningful work, which I think for, if
• 21:51
  we're talking about
• 21:52
  creative is that, for example,
• 21:53
  script writing or like finding patterns,
• 21:55
  for example, and then,
• 21:56
  you know, they may have like
• 21:58
  ads manager, like creative strategists,
• 22:00
  agents, they find those
• 22:01
  patterns, you know, he has like
• 22:02
  winning ads and then it goes straight to
• 22:04
  the, um, creative
• 22:06
  strategist and they approve of those
• 22:07
  patterns. They approve of the new hooks
• 22:08
  and then go straight to the client. Um,
• 22:11
  so basically kind of, kind of like a mini
• 22:13
  says, I guess, for
• 22:14
  those clients of the agency.
• 22:16
  It's so true. I think it's, it's, it's
• 22:18
  interesting to me because I think the,
• 22:21
  each piece of this that
• 22:22
  is, is getting more automated. So in
• 22:25
  terms of creation of like,
• 22:26
  okay, we're building this,
• 22:27
  you know, the persona builder. And then
• 22:29
  it's like the next phase
• 22:30
  of it is now it's like,
• 22:31
  it's building the persona builder. You
• 22:32
  have the brand guidelines,
• 22:33
  it's got it all in there.
• 22:35
  It's stored in super base. And then now
• 22:37
  they're able, it's able to
• 22:38
  like build and potentially
• 22:39
  publish and start to optimize ads. Um,
• 22:42
  you know, as, as you train
• 22:43
  it over time, I think that
• 22:44
  that's a really valuable one to think
• 22:47
  about as you kind of go
• 22:48
  through and learn this stuff,
• 22:50
  Freddie, like obviously you're deep in
• 22:51
  it. How are you
• 22:53
  continuing to upskill yourself?
• 22:55
  Like, how are you making sure that you're
• 22:57
  staying on top of it? Um,
• 22:59
  as much as you possibly can.
• 23:01
  Yeah. Or is it just messing around?
• 23:04
  I mean, I get FOMO on the daily about
• 23:06
  like, there's so much new
• 23:07
  stuff always coming out.
• 23:09
  I think, um, basically kind of committing
• 23:11
  to like one kind of like platform, like,
• 23:13
  you know, Claude and then also like
• 23:15
  Codex. Um, for example, I
• 23:16
  haven't played around that much
• 23:17
  with OpenClaw yet. Um, and then kind of
• 23:19
  staying on top and then
• 23:21
  also just being a lot of,
• 23:21
  on Twitter. Like there's so much stuff
• 23:24
  being built on Twitter as well
• 23:25
  that other people are sharing.
• 23:26
  Like every day I probably save like 10
• 23:27
  bookmarks. And then also
• 23:29
  just a lot of like, like, Hey,
• 23:30
  I have this idea, for example, like the
• 23:32
  creative automation would like
• 23:33
  automatically reviewing
• 23:34
  the ads. Like I think that popped in my
• 23:36
  head yesterday. I'm
• 23:36
  like, I don't want to review
• 23:38
  grammar anymore. Just type it into Claude
• 23:40
  Code. Like here's
• 23:40
  what I want. Here's what,
• 23:42
  which software I'm using. Um, can you
• 23:44
  help me build this? And
• 23:46
  then 95% of the time it's just,
• 23:48
  yes, it can help me build this. Um, and
• 23:51
  then I think the next
• 23:52
  question is just like,
• 23:52
  how can I integrate that into my, into
• 23:54
  my, my workflows? Like
• 23:56
  how can I make use of this?
• 23:57
  Not as a standalone workflow, but
• 23:59
  actually kind of part of my daily
• 24:02
  routine, I guess. Right.
• 24:03
  So it's not like one-time build and it's
• 24:05
  helpful for like 5% of the cases, but
• 24:07
  like actually helpful
• 24:08
  on the daily.
• 24:08
  And the creative stuff that you're built.
• 24:10
  So you see in terms of
• 24:12
  Facebook advertising today,
• 24:14
  creative being obviously the main lever,
• 24:16
  which I think a lot of us
• 24:17
  are going on. Do you think
• 24:18
  that there's going to be a period of time
• 24:20
  in which like the
• 24:20
  optimization or the, um, automatic stuff
• 24:25
  that we're doing with AI is going to be
• 24:27
  able to, is going to be adjusting bids
• 24:29
  and things like that?
• 24:30
  Or do you think that's mostly going to
• 24:31
  live within, like, it might
• 24:32
  be a leading edge to do it now,
• 24:33
  but like mostly meta is going to be doing
• 24:35
  that itself and you're not going to have
• 24:36
  to mess with that as
• 24:37
  much. Like, is it mostly going to be
• 24:38
  around continued optimize
• 24:40
  creative? And also on top of
• 24:41
  that, like, have you done anything
• 24:42
  messing around with like the
• 24:44
  other parts of the funnel, like
• 24:45
  websites or whatever, cause you can do
• 24:47
  all you can on the meta side, but you
• 24:49
  know, like obviously
• 24:50
  the landing pages are massive as well in
• 24:52
  terms of getting that flow
• 24:53
  and the match between those
• 24:54
  personas. Yep. Um, so regarding meta, I
• 24:57
  mean, I think, you know, you can
• 25:00
  definitely set up the,
• 25:01
  I don't know if the API has bit caps
• 25:02
  actually, if you can change those on the
• 25:04
  API, but like, I think
• 25:05
  like, like, like it can, the stuff can be
• 25:07
  automated for sure. You know, whether
• 25:08
  it's meta or whether
• 25:09
  it's like, you know, through an API or
• 25:11
  through some sort of skill,
• 25:12
  um, you know, I don't, you know,
• 25:14
  that's yet, you know, yet to come
• 25:15
  basically. But I think in terms of
• 25:17
  landing pages, like definitely,
• 25:19
  you know, I've played around with this.
• 25:20
  Um, it's pretty fascinating
• 25:22
  what you can build. Like I
• 25:23
  spend like, you know, one afternoon
• 25:25
  basically built like kind of like a, an,
• 25:27
  a CRO landing page agent,
• 25:29
  which basically, you know, generates
• 25:31
  landing pages, um, for a client. So
• 25:33
  basically based on the
• 25:35
  research from the research agent, um,
• 25:37
  that kind of landing page
• 25:38
  agent uses that research
• 25:40
  and the words, the reviews, et cetera,
• 25:42
  then you could give it like
• 25:43
  one other landing page that
• 25:44
  you are having, you know, whether it's a
• 25:46
  PDP or like an actual
• 25:47
  landing page. And basically it goes
• 25:49
  then in there, analyzes that landing page
• 25:51
  and like iteratively comes up with
• 25:53
  improvements. Um, you
• 25:54
  know, you can task it to make like small
• 25:56
  improvements, kind of like a CRO agency
• 25:58
  would do it basically.
• 25:59
  Or you can also kind of completely, you
• 26:01
  know, restart new landing
• 26:02
  pages. Um, and those landing
• 26:04
  pages are so on, like, I was surprised
• 26:07
  myself, like how on brand
• 26:08
  they were. Like it was using
• 26:10
  the exact, like, like even the copy, like
• 26:12
  it's basically split. So
• 26:13
  it has like a copy agent,
• 26:14
  which is basically only responsible for
• 26:16
  the copy that it kind of, kind of has
• 26:18
  like a layout agent,
• 26:19
  which is only responsible for like for
• 26:20
  the sections for the layouts. Like, you
• 26:22
  know, how was the hero
• 26:23
  structure? Um, like how was the kind of
• 26:25
  buying section structure, et
• 26:26
  cetera? And it was pretty good.
• 26:27
  Yeah. You saw the tweet. Yeah. Yeah.
• 26:29
  Well, so what do you, what tools did you
• 26:31
  use to build that one?
• 26:32
  What tools? I just like clock code,
• 26:34
  basically like typed it in like, Hey,
• 26:35
  like, uh, I want to build
• 26:37
  like a landing page agent, um, for this
• 26:40
  brand. Like here's the
• 26:41
  research that I've done on that brand.
• 26:43
  Please make sure you follow those, those,
• 26:44
  uh, instructions. And
• 26:46
  then thought of like how,
• 26:47
  because I think like in terms of like
• 26:48
  building this, you need to have
• 26:49
  specialized agents, which are
• 26:51
  responsible for one specific task. I
• 26:52
  think, you know, one landing page agent,
• 26:54
  which does everything
• 26:55
  probably won't be as good as having like
• 26:57
  three different agents,
• 26:58
  but in the same landing page,
• 26:59
  you know, like a copy agent section, like
• 27:01
  section building
• 27:02
  agent, and like a CRO agent,
• 27:03
  um, which checks for like CRO improvement
• 27:05
  stuff. So I think like giving
• 27:07
  one agent a very specific task
• 27:09
  is like super helpful with this type of
• 27:11
  stuff, right? Like you have,
• 27:13
  you don't have one creative
• 27:14
  strategist agent. You have like one,
• 27:16
  which is responsible maybe
• 27:17
  for analyzing the ads. And then
• 27:18
  another one, which is responsible for
• 27:19
  like writing the
• 27:20
  scripts and stuff like this.
• 27:21
  Just, uh, it's so exciting to just see
• 27:24
  what's possible and the
• 27:25
  quality that's just becoming so
• 27:27
  much better every day. So it's just like,
• 27:30
  uh, yeah, I'm, I'm, um,
• 27:33
  I'm both kind of scared and
• 27:36
  really, really excited at the same time
• 27:38
  because it's just changing so fast and
• 27:41
  it's, it's just, um, um,
• 27:44
  hard to keep yourself updated. So I think
• 27:46
  that the FOMO you're, uh,
• 27:48
  you're referring to, Fred,
• 27:49
  is, is real with everyone. And I think,
• 27:52
  uh, the more everyone of
• 27:53
  us can kind of share best
• 27:55
  practices and also kind of all our fuck
• 27:57
  ups and that, you know,
• 27:58
  that, that's what we want is just
• 28:00
  kind of help each other navigate and
• 28:02
  understand how to get
• 28:03
  maximize this to, to help as much as
• 28:05
  possible with our
• 28:06
  clients and our businesses.
• 28:08
  Yeah. Well, uh, Freddie, thank you for
• 28:10
  coming on. Freddie, Freddie
• 28:12
  will, uh, we'll do a quick
• 28:14
  loom overview of some of the things he
• 28:15
  talked about on this that
• 28:16
  we'll put in the show notes
• 28:17
  and, uh, we'll share that along with any
• 28:19
  links to the other
• 28:20
  things that he talked about.
• 28:22
  Um, and if you want to contact Freddie, I
• 28:23
  can put you in touch with
• 28:24
  him. You can always email me
• 28:25
  andrew at foxwelldigital.com and I'm
• 28:27
  happy to connect.
• 28:28
  Freddie, you're a rockstar dude.
• 28:30
  Super appreciate having you on and, uh,
• 28:32
  we'll talk to you soon.
• 28:33
  Thank you so much for having me.
• 28:34
  Thank you, Freddie.