Latest / AI DTC WTF / How to Hire Creative Strategists When AI Writes Every Sample | Jess Bachman, FireTeam

Transcript

How to Hire Creative Strategists When AI Writes Every Sample | Jess Bachman, FireTeam · AI DTC WTF

• 0:03
  If you work in DDC and you use AI and you're wondering
• 0:06
  what the F is going on every week, this is your podcast.
• 0:09
  The AI to see F podcast your home for tactical tips, strategies, and ideas
• 0:15
  that you can implement right now in your AI workflows to make your brand or agency
• 0:19
  more money.
• 0:20
  Okay, welcome to another episode of AI D to see what the.
• 0:25
  With Will and I and Jess Bachman from Fireteam Incredible episode already ahead.
• 0:31
  You could.
• 0:32
  Right now I can just tell you it's going to be good.
• 0:34
  So if you are thinking about
• 0:37
  leaving anywhere in this, it's on Poly Market.
• 0:40
  And if you're thinking about any leaving during this podcast at all,
• 0:43
  you're going to regret it.
• 0:44
  It's going to be the worst mistake of your life.
• 0:46
  Like the football coach in high school that I had,
• 0:48
  who told me that quitting football was the biggest mistake of my life?
• 0:51
  It's that it's that level and that still lives rent free in your head, right?
• 0:55
  You think about that?
• 0:56
  Yeah. I don't think about that. Ever.
• 0:59
  A natural athlete brought to my knees by the football coach.
• 1:03
  But we are here talking about AI with Jess.
• 1:06
  Jess has built a ton of stuff.
• 1:07
  Just as somebody who I love and respect and has an incredible agency.
• 1:11
  Awesome team. Jess, welcome to the podcast.
• 1:14
  Thank you.
• 1:14
  Love to be here with you guys.
• 1:16
  So let's talk about the thing that you most recently built,
• 1:18
  which is this AI hiring tool agent situation.
• 1:23
  What is this? Let's get into this.
• 1:26
  Yeah.
• 1:26
  So we are we are hiring a copy engineer.
• 1:30
  Or you could say that's a copywriter or something.
• 1:34
  And this is a process that we've done a lot.
• 1:36
  And also I'd lump in creative strategist hiring that.
• 1:40
  And so if you've ever hired a creative strategist
• 1:43
  I'm sure will you know, this is that like you asked
• 1:47
  you ask for work samples and you read their briefs and stuff
• 1:49
  and you cannot distinguish what is written by them or what it's written by AI.
• 1:55
  And even when like, it takes a lot of time to read samples and stuff.
• 1:59
  And so it's a it's a terrible heuristic to find people.
• 2:02
  And then also when we give people specific briefs to be like,
• 2:05
  all right, let's let's do a sample.
• 2:07
  Let's see how you work.
• 2:09
  Everything comes back as AI.
• 2:11
  And so it's I just did not want to run that
• 2:14
  playbook again hiring for this role.
• 2:17
  So what I did was I created a AI interviewer.
• 2:22
  And what it does is on the left hand side of your screen
• 2:26
  is a poorly written brief with all sorts of like errors and things inserted,
• 2:32
  and that it prompts you after it gets like your details of where you're
• 2:36
  where you live and whatever it prompts you to to find where the issues are.
• 2:42
  And so that like, I need to see your thinking
• 2:45
  about how you're looking at stuff,
• 2:47
  because that's the actual work that you need to do at
• 2:50
  our industry is find these particular errors and fix them.
• 2:54
  And so it will prompt you to find them.
• 2:57
  It'll prompt you to to fix them, but it'll do another thing
• 2:59
  where it prompts you to it'll like push on why you think that's an issue.
• 3:04
  Because I don't want to
• 3:06
  I don't want someone who's just coming in based off, you know, vibes or something.
• 3:09
  Like if you if you say like, you know, this is weak and it needs to be this way.
• 3:14
  Like what?
• 3:15
  What is the truth that you're bringing to that that makes you feel that way?
• 3:18
  Like, is it something you researched or previous winners or whatever?
• 3:22
  And so it'll do a whole bunch of turns of that probably takes 10 or 15 minutes.
• 3:27
  And, and then the last stage is to be like, all right,
• 3:31
  forget this brief, come up with a brand new idea
• 3:35
  for this client, and it gets a turn at doing that.
• 3:39
  So like editing and so like what I found
• 3:43
  in reviewing people's responses is like,
• 3:47
  there is so much signal in, in this stuff.
• 3:50
  I mean, it's kind of like a real interview,
• 3:52
  you know, but except I don't have to be there at all.
• 3:55
  And I can give it to like 50 people at once.
• 3:58
  And there's this, this is like they the people are.
• 4:02
  Let me interrupt like a second.
• 4:04
  They're they're getting the questions on there
• 4:07
  and they're speaking their responses or they're writing their responses.
• 4:10
  Yeah, they're typing it.
• 4:11
  It's just like typing the clod.
• 4:12
  It's the same like that sort of thing.
• 4:16
  One thing I had to put in there was paste detection and timing
• 4:21
  on their responses, because I was really worried that people would
• 4:26
  take the brief and then put that in the cloud,
• 4:28
  and then take the question and put that in the clod,
• 4:31
  and then take the answer and put that back in there.
• 4:33
  And so it detects if someone does a paste.
• 4:37
  And so I, I deploy this on Tuesday and I've already iterated a lot, a lot on it.
• 4:43
  One, I found it there was a lot of people that were using paste.
• 4:47
  And what I found out is people that are using like dictation or
• 4:51
  like speech to texting will be detected as a paste.
• 4:55
  So I put in a question up front that's like,
• 4:57
  are you using a dictation tool?
• 4:59
  And if they say no and then they're still pasting stuff in there.
• 5:02
  Well, it's like I don't know what you're trying to pull here, but
• 5:07
  there was a lot of that going on to.
• 5:09
  So it's it's been a it's been a very interesting battle.
• 5:13
  But save me a ton of time and already found me some great candidate.
• 5:16
  Interesting.
• 5:17
  I sort of want to double click on that too.
• 5:19
  So like the problem that you're trying to solve here is like
• 5:22
  you were just getting candidates that were effectively writing
• 5:25
  or giving their answers to with Claude.
• 5:28
  And counterintuitively, what you're saying is, like,
• 5:31
  you use Claude to sort of solve for that.
• 5:33
  Right?
• 5:34
  And so I guess what I'm trying to understand is,
• 5:37
  like, do you think that having sort of like an
• 5:40
  AI screener would be like a deterrent for someone
• 5:44
  who thinks like, oh, you know, this person can't even put in the time to like,
• 5:47
  you know,
• 5:48
  just to play devil's
• 5:48
  advocate, put in the time to have my initial screening interview.
• 5:52
  Yeah.
• 5:52
  Well, funny enough, there was a video that went viral of
• 5:56
  like the next day of like this AI interviewer but it was like faking
• 6:00
  a real person and whatever
• 6:04
  this, this is, this is different, I think.
• 6:07
  I think just being in chat and not trying to do voice, where you're
• 6:10
  talking to some weird avatar or something is entirely different.
• 6:16
  But this is also more than an interview
• 6:19
  to get at, like your experience,
• 6:22
  like you're doing the work that you'd be doing here at fireteam.
• 6:25
  And so the the response that I've got from candidates was like,
• 6:29
  oh man, this was actually fun to do.
• 6:32
  So I mean, they could be lying
• 6:33
  to, to get the job, but that's what they're saying.
• 6:35
  Yeah.
• 6:36
  I mean, well,
• 6:36
  I mean, somebody that's going to be doing that
• 6:38
  eight hours a day or whatever, you know, your hours are it's like,
• 6:41
  oh no, you said
• 6:41
  you guys were 35 hours a week or whatever because you get half day off on Fridays
• 6:45
  or whatever. Yeah. That's right.
• 6:46
  Yeah.
• 6:47
  So, so whatever, 35 hours a week they're doing this.
• 6:51
  You need to make sure it's like they should find this fun
• 6:54
  because like ultimately what what like a value proposition.
• 6:57
  And, you know, in the marketplace
• 6:58
  that you guys have had for a long time, is that the work that you do is different.
• 7:02
  And it's fun.
• 7:03
  Like, you know, I mean, it's it's like what sets you apart from going to,
• 7:07
  you know, another place to, to do this kind of work
• 7:09
  and people who are good at their job, they like to be tested.
• 7:13
  They like to show off their skills if they're like nervous
• 7:17
  or worried about it.
• 7:18
  That's already like a red flag.
• 7:20
  That's probably not okay.
• 7:21
  So you also in this said that you have an AI
• 7:25
  brief and script writer that beats human copywriters.
• 7:30
  Can you tell us about this?
• 7:31
  What are your houses built?
• 7:34
  What's you know,
• 7:35
  how do you how do you kind of think of this idea and what sets it apart?
• 7:39
  Yeah.
• 7:40
  So I've been man ever since like chat you first came out, I was
• 7:44
  I was very much looking for a way to replace a lot of the work
• 7:48
  that I do, which is copywriting at the same time,
• 7:52
  like Fire Team is, is a quality over quantity shop.
• 7:56
  And so like I am looking to explore like what is the
• 8:00
  the quality levers that we can pull on AI because we just aren't a quantity shop
• 8:04
  and we're not going to be that way.
• 8:06
  So this is this is probably like the 50th iteration of where I'm at here.
• 8:10
  But all of our briefs are using this
• 8:14
  brief writing tool and I can go into how it works.
• 8:19
  But I'll tell you how I know it's effective is because we do.
• 8:24
  We do have a copywriter on staff who who's very good.
• 8:29
  She's a very good copywriter, but we treat her as a control
• 8:33
  and that she does not use this particular tool.
• 8:36
  And then everyone at fire Team has a, a number.
• 8:40
  It's called a Windex or a W index, which is measures their ability to
• 8:46
  affect
• 8:47
  winning ads or their ability to create winning ads.
• 8:50
  And it's usually a number.
• 8:51
  Hopefully it's over 100.
• 8:54
  And it's super complicated to get into that number in a sense.
• 8:57
  But the people that are using this tool, myself and a few others, have a much
• 9:02
  higher Windex number than this other brief writer who's very good.
• 9:06
  She's very good.
• 9:09
  And so I have determined
• 9:11
  that, like, like we've even brought in, like, copywriters that aren't that great.
• 9:15
  And still using this tool, they're able to put out
• 9:19
  good briefs and the good I'm trying to be as objective as possible.
• 9:22
  The good is not like I love that.
• 9:24
  It's like it wins in ad accounts, like that's the only thing I'm caring about.
• 9:28
  And that's the only thing I'm really measuring here.
• 9:31
  So that's how I know it's working.
• 9:34
  Or better than copywriters manually.
• 9:37
  Makes sense 100%.
• 9:40
  Let's let's dissect.
• 9:42
  So I tend to agree,
• 9:46
  I found that copywriting has been the hardest unlock
• 9:51
  by far in terms of creative strategy.
• 9:55
  Like I was very good at like pointing you in the direction of like iterate on this,
• 10:00
  you know, build on this.
• 10:02
  Here's a net new persona to test.
• 10:03
  But like the copywriting,
• 10:05
  you know,
• 10:05
  I think I'm 90% of the way there, but you seem to have sort of cracked the code.
• 10:08
  So like, let's talk, you know, what model are you using?
• 10:12
  What is what is like
• 10:13
  sort of like the step by step process that the model is sort of going through.
• 10:17
  And then it sounds like you're almost like training junior strategists
• 10:22
  with this tool to write, to make them write more effective copy.
• 10:27
  And is that making them better copywriters,
• 10:29
  or is that just having someone just, like, better the machine?
• 10:34
  Yeah, I don't know if it's making them better copywriters,
• 10:36
  but it's certainly making them better at producing winning ads.
• 10:40
  I'm not
• 10:41
  I don't know if those things are linked, but I'm certainly going for
• 10:44
  the latter here. So I'll tell you how it works.
• 10:46
  So like all all of our all of our ads, they or their ideas,
• 10:51
  they start with basically a gist,
• 10:53
  which is just a small encapsulation of what the idea is.
• 10:58
  This gist is actually pulled out of three hours
• 11:02
  of brainstorming that we have a week where
• 11:05
  we'll just we'll just say into the ether
• 11:08
  like project name, rejuvenate, you know, bedwetting ad
• 11:12
  and then it'll it'll make that into our project
• 11:16
  management system sort of automatically from the transcript.
• 11:19
  It will add the gist about what we talked about, whatever.
• 11:22
  That just will then
• 11:23
  get served to the copywriter in the tool when they're ready to write.
• 11:27
  So we got that for their.
• 11:30
  Yeah.
• 11:30
  I just also want to like pause and say like a three hour brainstorming session.
• 11:33
  Like,
• 11:34
  I know this isn't about AI,
• 11:35
  but are you and your team just like hopping on a call
• 11:36
  for three hours and just like, mind melding?
• 11:38
  Yeah, it's it's one two hour call and then one one hour call. Wow.
• 11:44
  I mean, that's that's the that's the differentiator, right?
• 11:47
  Like immediately right is part of it.
• 11:49
  It's the impulse. Yeah. This is this is an aside.
• 11:52
  But the one thing that I found that, like AI cannot do well is come up with ideas.
• 11:57
  It's just like terrible at it.
• 12:00
  So it is very much a mind meld with like 13 people all yelling on zoom or whatever.
• 12:06
  So the the copywriter will get this gist in the system,
• 12:10
  and then they will go back and forth with buddy,
• 12:14
  which is the name we give to our AI copywriter about this particular idea.
• 12:20
  And buddy will reference previously winning briefs for this
• 12:23
  client or whatever, and it will probe the person be like,
• 12:28
  what?
• 12:29
  Like what about this is important?
• 12:31
  And then they'll like speak into it
• 12:32
  or type into it until it gets to a point where buddy feels like
• 12:37
  the copywriter has amplified the idea in the right ways to do the right thing
• 12:43
  that it needs for the generation, which is the next part.
• 12:47
  So it'll be like a sort of a simple idea, but what the copywriter needs to be like,
• 12:51
  what, like what's really going to get someone to tension or
• 12:55
  what's the tension here and put that in there.
• 12:58
  So then it will go into the generation and I'd probably say
• 13:03
  like 75% of what makes it effectively is that from the generation,
• 13:09
  it's pulling all of the learnings
• 13:13
  that we have that are both per client and across clients.
• 13:18
  So we have like for each client, there'll be like 10 to 40 learnings
• 13:23
  about what actually has been effective in previous ads
• 13:27
  and things that we've tested.
• 13:28
  And it'll pull that into the particular brief
• 13:31
  and write the brief and make sure a lot of that is in there.
• 13:35
  So to generate a brief based off this, this huge system prompt that we have,
• 13:40
  then we'll do a quality control check on itself based off things
• 13:44
  that I have found are generally defective when AI is trying to do a first pass.
• 13:50
  So I'll do its own second pass before the the human gets a chance to edit it, then
• 13:56
  it'll finally give like the full version in the in the format that we like.
• 14:01
  One thing that I found is kind of crazy is that you have to tell the AI
• 14:05
  to show the work on the learnings that you put in there, and why,
• 14:11
  because if you don't, it's going to get lazy
• 14:14
  and it's going to like, you know, fabricate things like, do you
• 14:18
  you really have to work with it to get it to do exactly what it wants and,
• 14:21
  and stress it out, to get it to do the thing that you want.
• 14:25
  And so after it's generated, there's a whole interface where the copywriter
• 14:29
  can go in and like, change a bunch of lines or do heavy revisions or,
• 14:33
  you know, talk back to buddy to get it to fix everything.
• 14:37
  And then at the final stage, they press the button
• 14:39
  and it just goes right into our project management system.
• 14:42
  Should take about a half hour per brief if you're doing it right.
• 14:46
  I mean, incredible process.
• 14:47
  One theme that I'm that I hear you talking about, as far as you know,
• 14:51
  you're very clear within anything that you're using AI for
• 14:55
  to make sure that you're putting guardrails on and definitions around
• 14:59
  where you believe it's going to screw up or like where it could lead people astray.
• 15:03
  And, you know, so can you tell me about the common ones
• 15:06
  that you run through that you're like, okay, you know, I mean, I mean, you know,
• 15:10
  well, there's a lot of things you don't allow of directions,
• 15:12
  but what are the common ones that you utilize?
• 15:19
  This episode of
• 15:20
  Ada is brought to you by runneth from motion.
• 15:24
  And, you know, there's so many interesting things about runneth
• 15:27
  that I absolutely love.
• 15:28
  The best marketers I know right now are not sitting there going,
• 15:31
  how do I get ad answer my questions?
• 15:33
  They're thinking in systems.
• 15:35
  It's I have taste, I have judgment, I know my customer better than anyone,
• 15:39
  and they're putting that into runoff and having it work on it
• 15:43
  in a really sophisticated way.
• 15:44
  It's putting it in there
• 15:45
  and runneth is amplifying what they can do with that particular data.
• 15:49
  It's not a chatbot, right?
• 15:51
  It's not just Claude, it's not just ChatGPT.
• 15:53
  That's the thing I love about runneth
• 15:55
  is that it pushes it further, and it has your ad data,
• 15:58
  your competitor data,
• 16:00
  your creative strategy, it has your product data, and it helps
• 16:03
  you understand all of these things together
• 16:05
  so that you can genuinely train it on how you think.
• 16:08
  And it becomes useful from an operational level,
• 16:11
  saving you time, giving you context, and giving you important information
• 16:15
  so that you can make the most absolute best decisions
• 16:20
  as a marketer today, to help make your business run more efficiently.
• 16:24
  Motion can do so much run if can do so much on top of it.
• 16:28
  So I encourage you to give it a shot.
• 16:29
  If you haven't run a calm and go and see what you can build.
• 16:33
  Yeah.
• 16:33
  So the latest it's very interesting because the latest iteration is not just
• 16:38
  guardrails on the AI but guardrails on the human, because previously,
• 16:43
  like what I did not want it to be was a skill in how you prompt AI.
• 16:48
  I did not want that because different people
• 16:50
  have different interpretations and how they talk to it.
• 16:53
  So the way it works now is like you have a conversation
• 16:56
  with the AI, and from that conversation
• 17:00
  it creates the input, which is basically, you know, half of the prompt
• 17:04
  that then goes into the generation.
• 17:06
  So it's not like the the writer is, you know, taking their own spin
• 17:11
  on what the input is or becomes like, they don't know how to.
• 17:15
  It's just like anything when you do when you talk with AI like
• 17:18
  or the image generation, like, I'm not going to write in an image prompt,
• 17:22
  I'm going to tell
• 17:22
  Claude to write me an image prompt because it knows how AI talks and thinks
• 17:25
  and whatever. So that's like the way to do it.
• 17:27
  So reverse prompting.
• 17:29
  So that is sort of built in there.
• 17:31
  And that has sort of negated the skill issue, where
• 17:34
  even our account managers who are not writers
• 17:37
  can use this tool very effectively, but a good writer can certainly plus it.
• 17:41
  Okay.
• 17:42
  I mean, all of the organization and the logic makes sense, but like,
• 17:46
  what's the source?
• 17:47
  Like, how are you getting this to write better copy.
• 17:49
  Like what is the framework?
• 17:51
  I guess I'd push back on what is better copy because this is not like
• 17:55
  in the prompt.
• 17:55
  It says you're the ghost of David Ogilvy or whatever.
• 17:58
  Like in the prompt it is, it's pulling
• 18:02
  from the highest value signal that we could possibly get,
• 18:06
  which is previous like learnings that have previously won on this account.
• 18:11
  And it's like folding those in there.
• 18:14
  Like, like we have some clients that have like 50 learnings,
• 18:17
  even like a great creative strategist isn't going to have
• 18:20
  all of that in their head and know when to deploy all of those.
• 18:23
  Whereas this can absolutely like jam
• 18:26
  packet in all the different layers of what it needs to do.
• 18:29
  And so
• 18:32
  that's really the source.
• 18:33
  It's not like I can't give you
• 18:34
  I could, I could like totally open source the prompt and give it to you.
• 18:37
  And it's not going to be the same thing because you don't have the same data
• 18:41
  that we are feeding into the into the machine.
• 18:44
  Got it. Yeah.
• 18:44
  I mean did is obviously the wedge in today's day and age.
• 18:48
  So it sounds like you sort of have like two data sets that you're feeding this
• 18:52
  like again, like a time series analysis of like all of the metadata.
• 18:56
  Like I'm sure you like pulling transcripts from winners and like correlating
• 19:00
  that to like sort of CPA row as thumbs up, Ctrl whatever.
• 19:03
  And then on the other piece of the data set is like the brainstorm.
• 19:07
  And so you're sort of combining those two elements.
• 19:09
  And based on that, you know, the, the tool is seeing like this is, you know,
• 19:14
  this is the top performing copy in this,
• 19:17
  in this ad account,
• 19:18
  like this is the top performing hook because I put the transcript with Gemini
• 19:21
  or, you know, you know, sonnet or whatever.
• 19:25
  And then based on the brainstorming session we just had about this client,
• 19:27
  like, let's translate everything we just said and reformat it into
• 19:33
  that exact or not reformat it.
• 19:36
  But just like re contextualize it
• 19:40
  based on the winning copy that we've already written effectively.
• 19:44
  Yeah.
• 19:45
  Except the the coming up with the winning, like what are the learnings
• 19:49
  is very much a manual process done entirely by Rachel, our media director.
• 19:54
  Like I said, I also do not trust
• 19:57
  AI with analysis.
• 20:00
  And like I get enough cloud slop from clients to begin with that.
• 20:04
  And I've tried to like, do a lot of my own, my own analysis.
• 20:07
  And it's very, very hard to to get at a right place
• 20:12
  that isn't like you just find one loophole and then it's like, oh, you're right.
• 20:17
  It's totally the reverse of what you just said or whatever.
• 20:19
  So all of our learnings are are manually sourced by a human.
• 20:25
  And I think and like when you, when you when she puts it
• 20:29
  in, she can put whether it's like, you know, a strong suspicion or
• 20:34
  you know the different there's like levels of learnings.
• 20:36
  And then the LM will use that information as well.
• 20:40
  And like lean into the long stuff, the,
• 20:42
  you know, the strong stuff and stuff like that.
• 20:44
  So it's a combination of very human and
• 20:49
  AI parts.
• 20:50
  Funny.
• 20:51
  That's that's so interesting because. Yeah. Wow.
• 20:54
  That's very, very cool.
• 20:55
  Because I feel like, you know, we've sort of come down on the other side
• 20:59
  where I found the AI is very good at analysis but sucks a copywriting.
• 21:02
  So like I have it sort of do the first 10% and say, like, here
• 21:05
  are all the new personas to test and like, you know, how are we doing that?
• 21:08
  It's just like time series, analysis app and Reddit cross-referencing the two.
• 21:13
  What personas haven't we tested that people in subreddits are talking about?
• 21:17
  And so then it's like
• 21:18
  strategist would have that context to go write the copy, but then it's like,
• 21:22
  I don't necessarily feel comfortable just yet,
• 21:26
  you know, giving them like a tool to write the copy.
• 21:29
  So I guess, like when you sort of think about, you know, fire team
• 21:33
  in five years, like,
• 21:34
  you know, let's just assume that you want to keep doing this thing
• 21:37
  and definitely like, how do you sort of think about, like,
• 21:39
  training talent and like building people
• 21:41
  if they're just like sort of relying on this to generate better winning ads?
• 21:45
  It's it is interesting because I think we're at a crossroads where
• 21:51
  the people that I've who are able to use
• 21:54
  AI best are the people that had those skills beforehand. Yes.
• 21:58
  And they know how to edit or basically not trust the outputs.
• 22:03
  You know,
• 22:03
  I'm not sure what's going to happen when we get to a stage
• 22:05
  where there is no beforehand and those people have frigging retired or moved on.
• 22:10
  You know, I do worry about that, but maybe we just keep hiring
• 22:15
  older and older copywriters, I guess.
• 22:19
  I'm not sure.
• 22:20
  Yeah.
• 22:20
  Age bias starts going to other direction, right?
• 22:24
  So I mean, like, what do you so what do you see this
• 22:29
  all of these things with the AI that you're doing changing
• 22:33
  in the offerings that your agency has as time
• 22:38
  goes on to because, you know, you guys do obviously,
• 22:42
  this very manual process of research and iterative ideas and, and, you know,
• 22:47
  and then you go in, you know, you're shooting stuff, you're making great ads.
• 22:52
  You know,
• 22:52
  as far as I know, you're doing very little in reference to like creating AI
• 22:56
  is actually creating the statics or anything.
• 22:58
  I know there's a little bit of that,
• 22:59
  but that's very small percentage of what you do.
• 23:03
  And how do you see this evolving in in your agency as time goes on?
• 23:08
  I think the North Star is that
• 23:12
  we're a performance based agency.
• 23:15
  We get paid when things work and when they don't, we don't get paid more.
• 23:19
  If we deliver more, like we get paid more if things work better.
• 23:24
  And so with that framing, we're constantly looking for ways to work better.
• 23:29
  And like we like we have two designers on staff
• 23:32
  and we have two video editors, and they both use AI all the time,
• 23:36
  but not for the purpose of like pumping out, you know, 800 statics.
• 23:41
  That's that's not something that we've seen be effective.
• 23:45
  And so I think we're going to constantly, you know, like we track our win rate
• 23:49
  across clients and our hit rate and like our people on
• 23:53
  staff are looking to improve their own Windex scores and stuff like that.
• 23:56
  And so I think if you build the incentives internally to do this
• 24:01
  and it ladders up to the incentives of the agency, which is to,
• 24:05
  you know, win more, you don't have to, you know,
• 24:08
  there's a simplicity and a complexity that's like wrapped up into that.
• 24:11
  And I don't have to, you know, people will find their stride
• 24:15
  at their agency, at this agency and do what they need to do.
• 24:18
  I don't know, I'm kind of just rambling,
• 24:20
  but let me know if this is making any kind of sense whatsoever.
• 24:23
  I mean, it's yeah, I mean, I think it's your
• 24:26
  what I hear you saying is you're providing the framework of this is what we do.
• 24:30
  These are the outputs we need, and this is how we're tracking it.
• 24:34
  And the way that you get there
• 24:38
  is, you know, sort of up to you to a degree.
• 24:43
  And if you're using AI, there needs to be certain guardrails on it.
• 24:46
  And here's a way that we think about that.
• 24:47
  But you're giving them the freedom to be creative and be productive to
• 24:53
  to try to, you know, increase their own Windex or whatever else.
• 24:57
  And I think if our offer was a set
• 25:01
  volume of creative, we'd be a very different type of agency and,
• 25:07
  you know, different that looks different on a margins perspective.
• 25:10
  And we just operate entirely different
• 25:12
  in a way that I would literally probably hate my job trying to do.
• 25:16
  So I very much like what we're where
• 25:19
  we're at and feel very confident in our offer, I guess.
• 25:24
  Yeah, absolutely.
• 25:25
  I mean, it's always been as the volume game is always going to be there, right?
• 25:30
  And it's like as a hedge, you guys are like, look, this is a manual process.
• 25:33
  It's different.
• 25:34
  Is it is it, is it always going to be there
• 25:37
  I think well I think I think if you look in any marketplace like there's always
• 25:40
  there's always going to be room for people that want the volume style
• 25:46
  and want like, yes, let's make a bunch of this or,
• 25:49
  you know, let's give this and maybe, maybe they're not going to pay
• 25:52
  as much for it and it'll continue to decrease in value.
• 25:55
  But I think there's always this idea that more is better from a lot of folks.
• 25:59
  Right.
• 25:59
  Like, you know, like I think over time it maybe isn't going to perform,
• 26:04
  you know, as well.
• 26:04
  And people are like it turns into a negative thing, I guess.
• 26:07
  I think I think the pendulum will swing back if it hasn't already.
• 26:12
  The volume game is not coming from I'm telling you,
• 26:16
  it's not coming from Andromeda or a need for
• 26:22
  you know what meta wants.
• 26:23
  It's coming from the ability to create this stuff at scale,
• 26:27
  which agencies absolutely freaking love because it's so cheap,
• 26:31
  and then that gets wrapped and then that gets twisted into what like rep say.
• 26:36
  And then all of a sudden it's like attached to Andromeda,
• 26:40
  but like you go back historically, it has always been a quality game.
• 26:45
  Like it is never been a volume game.
• 26:48
  That's true.
• 26:49
  And so I think you're right.
• 26:50
  I guess what I'm saying is, I think there's always
• 26:52
  this idea that people are like, yeah, I really like that.
• 26:54
  It can do 50, I can get 500 of these, you know what I mean?
• 26:58
  Like, I mean, it's like it's a there's a novel concept to it.
• 27:00
  But in terms of actual quality, I think you're right.
• 27:03
  And and I think you raise a good point about what
• 27:06
  how it gets twisted in terms of how it gets talked about.
• 27:09
  Yes. You need to have more creative and, you know, whatever.
• 27:11
  And I think a lot of people are listening to this.
• 27:13
  Would totally agree about the creative and quality piece.
• 27:15
  I guess to me the question is is always, you know, how do we turn it into,
• 27:21
  you know, how do we research brands better and make sure that we understand
• 27:26
  angles that haven't been explored and make quality ads
• 27:29
  to speak to those people that have not been addressed previously?
• 27:32
  That's like the challenge of any creative creative shop creative strategist today.
• 27:37
  And how do you present them in a way that's going to make that company
• 27:40
  more money, and you don't really know in terms unless you start sort of testing
• 27:45
  and digging into that, what do you think creative strategy will look like?
• 27:50
  I mean, the role of a creative strategist
• 27:51
  has already changed in the last 365 days, dramatically.
• 27:55
  So like, what's the what's this?
• 27:57
  It's like a Rorschach inkblot, whatever that term means. Yes.
• 28:01
  Yeah. Right. Yeah.
• 28:04
  Yeah. It's interesting.
• 28:05
  Like, I just wonder what it's I just wonder how it's
• 28:07
  how the how it's going to evolve creative strategy overall.
• 28:09
  I mean, I think it's obviously AI enabled, but I,
• 28:13
  I wonder sometimes there's certainly a lot of talk about
• 28:16
  100 X employee right now to like the person can do a lot more.
• 28:21
  Do you feel that the volume
• 28:23
  of tasks that a person can accomplish
• 28:28
  equate to better quality for the end user, for the client volume?
• 28:33
  No, I my general lens of of AI
• 28:37
  is is like more and better.
• 28:40
  Like you can certainly do more, but you also have to do better as well.
• 28:46
  And and whenever we talk to employees, it's always like
• 28:51
  where where in your process can we insert AI to
• 28:55
  allow you to do more, but also, you know, better at the same time
• 28:59
  in terms of like the 100 X employees,
• 29:01
  this is something that I've really struggled with
• 29:03
  because I love building AI and I want my employees to do it.
• 29:08
  But I've really found that, like you, you cannot
• 29:12
  turn someone into a builder.
• 29:15
  Like you can identify the drive to like,
• 29:19
  improve things, but like, you can't take like a frigging,
• 29:23
  you know, even like a B+ employee and throw them in a freaking cloud code
• 29:27
  harness and and be like, look at all the things you can do.
• 29:30
  And you can't spark that desire
• 29:33
  to improve something, you know, on the one sense,
• 29:37
  because they're an employee and you know you're the owner, that's one thing.
• 29:41
  But also in general, like you kind of need the hunger in the fire to like push
• 29:46
  this stuff, which is very opaque and hard to work
• 29:49
  with and confusing at times to get what you want.
• 29:52
  And you might you might need to like, hire those people or look
• 29:57
  for those those people in the signal
• 29:59
  when you're hiring, because it is very difficult,
• 30:01
  if not impossible to motivate those people, just giving them a tool
• 30:06
  that can now ten x, you know what they can do.
• 30:09
  Yeah,
• 30:11
  they're the way that they think.
• 30:12
  And the way a person thinks is still a massive competitive advantage.
• 30:15
  Like the way that they're going to reason through something
• 30:19
  and sort of look at the why of it, etc..
• 30:22
  You know, I agree, and I've seen that definitely happening of
• 30:25
  like I've given them all the access,
• 30:26
  I've given them all the tools and they're still not doing what I need.
• 30:29
  It's like I also to me, that's also
• 30:32
  like a lack of strategy, that there isn't the will.
• 30:35
  And I've talked about this in the podcast
• 30:37
  like there's there's this there's no strategy.
• 30:39
  It's just like come up with ideas like do more and you're like, well,
• 30:43
  they don't see it the way that you see it.
• 30:44
  Like it's not, you know, you have to sort of
• 30:46
  lay out like, this is why this is important.
• 30:48
  Here's the idea.
• 30:49
  Come up with iterative ideas on that and like that or, you know, whatever.
• 30:53
  Like making it more narrow so the person can sort of
• 30:55
  start to understand, like what they're trying to accomplish.
• 30:57
  But yeah, it's interesting.
• 30:59
  Yeah. You can also
• 31:03
  instead of just ten axing, you can almost like negative
• 31:05
  ten x some employees because it allows them to create
• 31:09
  so much like downstream work or you know, there's error.
• 31:14
  Like they just allows them to create so much more stuff that if they were a
• 31:19
  C plus to begin with.
• 31:20
  Now there's like they're just generating all types of like work.
• 31:24
  And it can really it can really like gum up your team.
• 31:28
  Yeah. I mean I completely agree with that.
• 31:31
  It's like definitely can create a race to the bottom very quickly.
• 31:35
  And that's like one,
• 31:36
  you know, downside of saying like, oh, we're AI enabled or like we use AI.
• 31:40
  You get these people
• 31:41
  like, oh, I can just like fucking phone it in and just like, you know,
• 31:45
  not have to think so.
• 31:48
  I mean, yeah.
• 31:49
  Anyway, it is it is sort of crazy.
• 31:51
  Yeah.
• 31:52
  I think, I think if, if I'm a listener and I'm maybe like looking for a job
• 31:58
  right now, you know, at a brand or an agency, you know, something you said is
• 32:01
  like you can't force people to sort of like ten, like who has this like sort of.
• 32:05
  And you briefly sort of touched on this,
• 32:07
  but I really want to sort of drill down on it like,
• 32:09
  like what sort of person
• 32:10
  has this really worked for, and who's a person that this hasn't worked for?
• 32:14
  And like, what is a gap and the difference between the two of them?
• 32:17
  Is it like their copywriting skills, or is it something like more inane
• 32:21
  that they sort of innately have? What do you mean work for?
• 32:24
  Like what type of person?
• 32:25
  Yeah.
• 32:25
  So is it like and I don't want to just like,
• 32:27
  you know, lead the horse water here, but like, is it someone who is like
• 32:30
  intellectually curious about, like, you know, getting better or like,
• 32:35
  yeah, sort of that idea like, is this going to take like
• 32:39
  D plus person to an A
• 32:41
  or you know, or, or like a plus person to a B or is it,
• 32:45
  you know, only going to really work if you're a B, you know, to an A and like
• 32:49
  if you are like a D plus now like,
• 32:51
  how can you get yourself up to be a B like using AI?
• 32:54
  I'm not sure. I'm not sure you can.
• 32:56
  I think, you know, the the floor is probably
• 32:58
  AI is just raising the floor to a C,
• 33:01
  and if you're a deep person, maybe you can get up there.
• 33:03
  But it's hard to get higher.
• 33:05
  And like if I was if I was evaluating a creative strategist,
• 33:10
  like if I really want to pay a very good creative
• 33:14
  strategist to change my business,
• 33:17
  unfortunately, I would be looking at their GitHub commits
• 33:21
  and seeing how much, how much stuff they are like pushing,
• 33:24
  because that is the signal that that they're looking
• 33:28
  to like improve and change and actually be on the forefront of using this AI,
• 33:33
  like AI is like anything, it's it's
• 33:37
  the people who are well, there's like
• 33:40
  there's there's two levels to that to a, there's like being so bought in
• 33:45
  that you kind of like don't know what you're doing
• 33:48
  and you can't evaluate it and you're just like,
• 33:51
  you just like so and you know, we have a lot of clients that are even,
• 33:54
  you know, like this as well.
• 33:55
  And then there's like the really effective layer is one layer below that where
• 34:00
  you have a very healthy sense of skepticism,
• 34:03
  but you're still building and you're still testing.
• 34:05
  And that is like the true, productive, productive layer.
• 34:09
  And so I've seen people like,
• 34:10
  you know, I built an AI and it's like it's based on consciousness and love.
• 34:14
  And it's like, that's not the
• 34:16
  that's like the wrong level, but like, show me what you've built.
• 34:19
  Whether you are an AI or a strategist or a video editor or whatever.
• 34:24
  And if it's like a real tool or something
• 34:27
  like that, that's probably the highest signal for employees.
• 34:30
  I think that's I mean, I think that's always been the case of,
• 34:33
  you know, going back thinking about when we were in the early days of meta
• 34:37
  ads and Facebook ads, like I was always and still to this day, of the most curious
• 34:42
  with about people that are doing crazy shit like that
• 34:47
  are trying really strange stuff that are doing, you know, and it's like,
• 34:49
  and I think if you know, the technical, if it's a person
• 34:52
  that's competent on writing and can think through things
• 34:55
  and then has this capability of almost technical AI curiosity,
• 35:00
  then it's like a winning combo because it's
• 35:02
  the tools are obviously going to continue to change, you know, and it's like
• 35:06
  they're going to continue to to evolve and you have to to go for that.
• 35:09
  So I think looking at GitHub commits is like super interesting idea
• 35:12
  and one that I was,
• 35:14
  you know, voiced to another person the other day about, you know,
• 35:18
  trying to get their employees to be more bought into this.
• 35:20
  Like, hey, can we look at that as a metric?
• 35:22
  Because I think that's and for the creative strategist,
• 35:25
  like for copywriting, it might be one thing, but for a creative
• 35:29
  strategist, which is a role that spans multiple departments and disciplines,
• 35:34
  like that is absolutely a place where AI enablement can can like force
• 35:39
  multiply all sorts of outcomes versus purely like copywriting or I know we're
• 35:45
  sort of nearing time here, but sorry, I do want to sort of drill down into this
• 35:48
  because I think it's fascinating sort of the GitHub
• 35:53
  commit sort of hypothesis.
• 35:55
  And I again, just playing the devil's advocate, I don't agree or disagree, but
• 35:59
  like, you know, obviously tech companies for a while were gauging their employees
• 36:03
  success by the amount of tokens they were using internally.
• 36:06
  And obviously like that caused the employee that costs
• 36:10
  the company's a fortune.
• 36:11
  So they stopped doing that in this case.
• 36:13
  Like, how do you know that the GitHub
• 36:16
  repo that they're creating are actually accretive?
• 36:19
  And they're not just like in this AI psychosis state where they just like
• 36:22
  are building because it's like fun or just like for the sake of building.
• 36:26
  Yeah.
• 36:27
  That's a great question.
• 36:29
  I think it's it's probably a lot easier to like, run up a huge token bill
• 36:33
  than, you know, do commits for days and days and months and months.
• 36:37
  But I think you I think probably, probably what I would do,
• 36:42
  I would create an AI interviewer that interviewed them
• 36:46
  specifically on their their thoughts around
• 36:49
  AI and probably put some like, like, here's a problem.
• 36:53
  How would you solve that with AI?
• 36:55
  And if they're like, oh,
• 36:56
  you need Jack into the matrix, then they're like the wrong person.
• 36:59
  But that's probably what I would do.
• 37:02
  Yeah, so much of it.
• 37:03
  And hiring is always thinking like figuring out how somebody thinks anyway,
• 37:06
  you know, so just doing that a little bit different speed.
• 37:09
  So just one more question on the GitHub thing.
• 37:12
  What is the Willy Wonka golden ticket?
• 37:15
  If you opened a GitHub and you saw X, you know, would you say
• 37:19
  like this person has got the jobs to be a great credit strategist?
• 37:23
  I would say, I would say because a lot of this
• 37:29
  vibe coding is still pretty new.
• 37:32
  I just want to see activity to start.
• 37:36
  I would, I might say that like too much activity or if you're like,
• 37:42
  you know, contributing to open source or something,
• 37:45
  it's like you might be too far in the dev side of things, but like,
• 37:48
  I really just want to see the fire and the determination,
• 37:52
  even the determination to fricking for a quote unquote creative strategist
• 37:56
  or someone who feels like their creator, or if they're a UGC creator
• 37:59
  and they want to become a creative strategist,
• 38:01
  even to freaking get set up on GitHub and start like pushing
• 38:04
  stuff requires you to push through some uncomfortableness
• 38:08
  about what your new identity is, or what I'm doing, or what this future is,
• 38:12
  and like seeing their ability to do that on a on a consistent basis,
• 38:18
  regardless of what the, you know, the work product is, is, is a very good start.
• 38:22
  Well, I'd like to introduce you to interviewer by Fire team.
• 38:27
  The world of the the world of interviewing is never going to be the same.
• 38:31
  Or I'm trying to think about all the exact the scripts from the exact same video.
• 38:35
  So we've lost.
• 38:36
  Well, I'm more than happy to share the like I posted in the channel, like the the
• 38:40
  the prompt of how to sort of
• 38:42
  put this together in a huge EMD file on what I've learned in doing it.
• 38:44
  And you can just throw it in your car and say, build it.
• 38:47
  So I'm happy to share that too. Nice. Well thank you.
• 38:49
  We'll put that in the in the notes.
• 38:51
  Jess, appreciate your time and appreciate you being here.
• 38:54
  If anybody has questions or whatever they can reach you on X,
• 38:57
  we'll put that in the show notes as well.
• 38:59
  And as always, thank you to our listeners.
• 39:02
  Always my pleasure guys.