Latest / Sleeping Barber - A Marketing Podcast / SBP 235: The Sharp Cut - AI Won't Save You. Leadership Will.

Transcript

SBP 235: The Sharp Cut - AI Won't Save You. Leadership Will. · Sleeping Barber - A Marketing Podcast

• 0:00
  SPEAKER_00: Welcome to the Sleeping Baba podcast.
• 0:03
  SPEAKER_01: Welcome back to Sharp Cut, where we take marketing's biggest comfort blankets and cut them with research and real-world evidence to try and separate what many of us believe with uh the ev what the evidence actually says.
• 0:16
  SPEAKER_01: And so our topic today is really AI and leadership, which doesn't necessarily have a lot of evidence, which is why we're trying to get dig into this.
• 0:25
  SPEAKER_00: We're getting ahead of this.
• 0:26
  SPEAKER_00: Yeah.
• 0:27
  SPEAKER_00: But here the reality is like, here's what nobody wants to say out loud.
• 0:31
  SPEAKER_00: Most marketing teams haven't been transformed by AI, they've been paralyzed by it.
• 0:36
  SPEAKER_00: Ooh.
• 0:37
  SPEAKER_01: And the part that I think hurt hurts a lot of us is it's not because the technology is bad.
• 0:42
  SPEAKER_01: The models are pretty spectacular, actually.
• 0:45
  SPEAKER_01: The pilots are everywhere.
• 0:46
  SPEAKER_00: Yeah, MIT study this.
• 0:47
  SPEAKER_00: 300 deployments, 95% produced zero measurable return, unfortunately.
• 0:53
  SPEAKER_01: Yeah.
• 0:54
  SPEAKER_01: Which is not that checks with what the people I've been talking to have been saying as well.
• 0:59
  SPEAKER_01: It's kind of underwhelming.
• 1:01
  SPEAKER_01: And the 5% that worked, they weren't better at AI.
• 1:04
  SPEAKER_00: They were better at something else entirely.
• 1:06
  SPEAKER_00: So today on the sharp cut, why the tools alone won't cross that divide and what leadership really needs to do.
• 1:14
  SPEAKER_01: Yeah.
• 1:14
  SPEAKER_01: So I mean, I have a bit of a confession.
• 1:16
  SPEAKER_01: The story that most of us told ourselves about AI in 2024 and then 2025 was really a story about miracles.
• 1:24
  SPEAKER_00: Yeah, and the story we're living in 2026 is a lot smaller, stranger, more useful in specific ways, though.
• 1:31
  SPEAKER_00: Totally.
• 1:32
  SPEAKER_01: And the gap between those two stories is where most marketing leaders are stuck, I think.
• 1:36
  SPEAKER_01: Pretending that we're not stuck is the first thing that I think really has to go.
• 1:41
  SPEAKER_00: Yeah, you want to start with a confession here.
• 1:43
  SPEAKER_00: I'm sensing it already, Mark, don't you?
• 1:45
  SPEAKER_01: Yeah, I do.
• 1:46
  SPEAKER_01: I've been building and leading uh AI workshops for the past couple of years, and the honors report is that the transformation I expected didn't really happen.
• 1:56
  SPEAKER_01: But it's not zero.
• 1:57
  SPEAKER_01: No, it's not zero.
• 1:58
  SPEAKER_01: For specific tasks, the tools are honestly, it's remarkable.
• 2:02
  SPEAKER_01: Like I use it every day, and it's it's changing the way I work for sure.
• 2:06
  SPEAKER_01: Um, but for everything else, really, the average output is well, it's kind of average.
• 2:11
  SPEAKER_01: I invest a ton of time uh expecting a step change, and and I got a step, but not necessarily the change.
• 2:19
  SPEAKER_01: And what came out of the trough for you?
• 2:23
  SPEAKER_01: Well, there's really one line that the value of AI is less about the tool, and it's more about how it forces us to think.
• 2:30
  SPEAKER_00: That's an interesting reframe, which connects to something Roger Martin said when we interviewed him earlier in the year.
• 2:36
  SPEAKER_00: Yeah, why don't you walk us through it, B?
• 2:38
  SPEAKER_00: Yeah, Roger called AI a mode-seeking entity, not the mean, not the medium, the mode, the most frequent answer, essentially.
• 2:46
  SPEAKER_01: Uh in his words, I'm paraphrasing a little bit, but it can do anything, AI, that is, without breaking a sweat.
• 2:52
  SPEAKER_01: But it begs the question: is that what you want?
• 2:54
  SPEAKER_01: Do you want the modal approach?
• 2:56
  SPEAKER_01: Wouldn't you like to have the best approach?
• 2:58
  SPEAKER_01: It's not set up to give you that, though, and that's the challenge.
• 3:01
  SPEAKER_00: And the reality is marketing's not even a mode game.
• 3:04
  SPEAKER_00: Brands don't grow by being modal.
• 3:06
  SPEAKER_00: Distinctiveness is the asset, mental availability is the asset, fame is the asset.
• 3:12
  SPEAKER_00: None of those are produced by averaging, though.
• 3:14
  SPEAKER_01: Yeah, which is where a lot of the people get stuck.
• 3:16
  SPEAKER_01: If the modal gives you the mode, in other words, the middle, the most popular, um, and you're not paying attention, your work regresses to the mean.
• 3:24
  SPEAKER_01: Well, it looks like you're improving, which is a hard one to differentiate there.
• 3:29
  SPEAKER_00: Yeah, David Lansfield wrote a piece for HBR this August that puts on scene on the same argument.
• 3:36
  SPEAKER_00: A CEO he calls Susan was preparing for a town hall speech after a very difficult quarter.
• 3:41
  SPEAKER_01: Yeah, and her chief of staff had AI draft it, which is normal.
• 3:44
  SPEAKER_01: A lot of us are starting to do that too.
• 3:46
  SPEAKER_00: Yeah, so pulling from her past speeches, internal memos, the draft was coherent, measured, and Susan read it and felt something was off.
• 3:53
  SPEAKER_01: So what did Lansfield say the problem was?
• 3:55
  SPEAKER_00: The draft was too safe.
• 3:57
  SPEAKER_00: Rather than responding to the realities of the moment, it stitched together reassurances she offered in the past.
• 4:03
  SPEAKER_00: That's the mode-seeking machine doing exactly what it is what it is designed to do.
• 4:08
  SPEAKER_01: Yeah, producing the modal answer, the most offensible, most familiar version of what Susan had already said.
• 4:14
  SPEAKER_01: In a difficult quarter, the modal answer isn't always the right answer.
• 4:18
  SPEAKER_01: In fact, it could be totally the wrong answer.
• 4:20
  SPEAKER_00: Yeah, you did something during your research.
• 4:22
  SPEAKER_00: I want you to walk us through that.
• 4:24
  SPEAKER_00: You asked the modals themselves.
• 4:26
  SPEAKER_00: Oh, yeah, this is funny.
• 4:27
  SPEAKER_01: So I got tired of all the stupid things.
• 4:30
  SPEAKER_01: I'm not saying stupid things.
• 4:32
  SPEAKER_01: Saying things like, good catch.
• 4:34
  SPEAKER_01: You know, I should have thought of that.
• 4:35
  SPEAKER_01: You know, every time I questioned something, so I asked why.
• 4:38
  SPEAKER_01: And then I did, you know, why, why, why.
• 4:40
  SPEAKER_01: I kept asking why it's five whys.
• 4:41
  SPEAKER_00: I gotta know, I gotta know, like, what did it actually say though, after?
• 4:45
  SPEAKER_01: So each one of them, yeah, each one of them came up with something slightly different.
• 4:48
  SPEAKER_01: But the one that Claude gave me was really interesting.
• 4:52
  SPEAKER_01: Um, so well, it said I'm designed to be accommodating and non-confrontational, prioritizing, prioritizing a cooperative tone over critical independent debate.
• 5:02
  SPEAKER_01: And Chat GPT had a really great one too.
• 5:05
  SPEAKER_01: And so it had said, if I were a two-person research team, I'd be the one who always says, good point, instead of let's rerun the data.
• 5:13
  SPEAKER_00: The tool is telling on itself, essentially, confessing that the mode-seeking behavior in its own words.
• 5:19
  SPEAKER_01: Yeah, and we handed it to every marketing team on earth at the same time, well, really every team on earth at the same time.
• 5:24
  SPEAKER_00: Yeah, no kidding.
• 5:25
  SPEAKER_00: Everyone's coding the 95% number as proof that AI is overhyped.
• 5:30
  SPEAKER_00: One quick clarification though, before we go further, when people in marketing hear five and ninety-five, most of them think of John Dawes' 595 rule, that in any given category, only about 5% of buyers are in market to buy right now, and the 95% aren't.
• 5:46
  SPEAKER_00: That's a completely different finding.
• 5:48
  SPEAKER_00: This 595 is the MIT NADA Gen AI divide study, which says 95% of AI pilots produced no PL impact.
• 5:58
  SPEAKER_00: 5% did.
• 6:00
  SPEAKER_00: Same numbers, different research.
• 6:02
  SPEAKER_01: Yeah, it's a good call.
• 6:02
  SPEAKER_01: Thanks for just clarifying that.
• 6:04
  SPEAKER_01: So let's walk through the MIT piece then.
• 6:06
  SPEAKER_00: Yeah, so MIT's uh NADA's Gen AI, it's a mouthful divide study from 2025, 300 enterprise AI diplomats, 52 case studies, 153 leader surveys.
• 6:19
  SPEAKER_00: The headline: 95% of enterprise AI pilots delivered no measurable PL impact.
• 6:26
  SPEAKER_00: The bottleneck they identified wasn't infrastructure, though, wasn't regulation, and it wasn't talent.
• 6:32
  SPEAKER_00: It was learning.
• 6:33
  SPEAKER_00: Organizations were buying tools, not building operating models.
• 6:36
  SPEAKER_00: So the successful 5% shared three traits tightly scoped initiatives, domain-specific focus, and smart external partnerships.
• 6:44
  SPEAKER_00: MIT frames it as an approach difference, not a technology difference.
• 6:49
  SPEAKER_01: Yeah, it just a slight anecdote on my end, too.
• 6:51
  SPEAKER_01: Like I think that was one of the biggest problems I had with AI.
• 6:54
  SPEAKER_01: I was trying to do too many things.
• 6:56
  SPEAKER_01: The narrower I've been focusing it, the better it seems to be, which is kind of interesting.
• 7:02
  SPEAKER_01: So once you unpack that approach, though, you find leadership underneath it.
• 7:06
  SPEAKER_01: So somebody had to decide what to tight scope tightly, and somebody had to define what a good partnership looks like.
• 7:13
  SPEAKER_01: And so those are really calls that a leader makes, not necessarily a procurement process.
• 7:18
  SPEAKER_01: And then work published something in May that put the same finding out there, but in a slightly different language.
• 7:24
  SPEAKER_01: So in the multiplier playbook, uh, they worked with analytic partners, Barra AI, Profit, and System One.
• 7:30
  SPEAKER_01: They called the phenomenon, the say do gap.
• 7:32
  SPEAKER_01: So in marketing theory, a lot of CMOs know the theory, but they have a hard time applying it for a whole bunch of reasons.
• 7:39
  SPEAKER_01: And there's cultural, procedural, and structural barriers that are really preventing them from getting that implemented.
• 7:46
  SPEAKER_00: Two independent research bodies looking at different layers of the same problem, arriving at the same conclusion.
• 7:52
  SPEAKER_00: MIT names it structurally, work names it culturally.
• 7:56
  SPEAKER_00: Both say the technology isn't the constraint, the organization is.
• 8:00
  SPEAKER_01: Yeah, and the stakes are worse than most people realize.
• 8:02
  SPEAKER_01: So Gartner's 2025 CMO spend survey has marketing budgets flatlined at seven, oh seven and a half percent, let's call it, of company revenue.
• 8:11
  SPEAKER_01: And inside of that flat budget, organizations are now starting to reallocate to AI tooling and away from media.
• 8:18
  SPEAKER_00: Yeah, so combined the IPA evidence, which says cutting brand investment and downturns produces lasting damage, with the Gen AI divide, which says 95% of AI investment produces nothing, man, you got a nothing covered.
• 8:40
  SPEAKER_00: Yeah.
• 8:41
  SPEAKER_00: How do you have that conversation though with a CFO?
• 8:45
  unknown: I don't know.
• 8:47
  SPEAKER_00: That's a great question.
• 8:48
  SPEAKER_01: But we do have some answers actually coming up in a future podcast.
• 8:52
  SPEAKER_01: Uh you start by admitting, I think, that you can't defend most of the AI spend.
• 8:58
  SPEAKER_01: And then you start to ask what would need to be true for the 5% pattern to show up in your organization.
• 9:03
  SPEAKER_01: And that's a really important leadership conversation.
• 9:06
  SPEAKER_01: Again, it's not a procurement conversation.
• 9:08
  SPEAKER_00: Roger Martin wrote a piece for HBR, I think it was back in 2013, that we both keep coming back to from time and time again.
• 9:16
  SPEAKER_00: He called the modern white collar organization a decision factory.
• 9:20
  SPEAKER_01: Yeah, I I love this article.
• 9:21
  SPEAKER_01: It's so good.
• 9:22
  SPEAKER_01: Um his argument really is um, and and not just for marketers, but a lot of us in in knowledge work type of positions don't make things anymore.
• 9:32
  SPEAKER_01: In the case of marketers, we don't make ads anymore.
• 9:35
  SPEAKER_01: We make decisions about the ads.
• 9:38
  SPEAKER_01: And they sit in meetings and producing memos and slides that become inputs to other meetings, producing other memos.
• 9:45
  SPEAKER_01: And we've all seen that show before.
• 9:47
  SPEAKER_01: Shuffle some paper around, move it, and move it, move it, and see what happens at the end of the chain.
• 9:51
  SPEAKER_00: And when we interviewed him earlier this year, he expanded on that.
• 9:55
  SPEAKER_00: Knowledge workers were 15% of the U.S.
• 9:57
  SPEAKER_00: economy in 1959 when Peter Drucker coined the term.
• 10:01
  SPEAKER_01: Yeah, and it's shocking when you start looking at it.
• 10:03
  SPEAKER_01: Um, by 2012, knowledge workers made up of about 40% of the workforce.
• 10:09
  SPEAKER_01: I think it was by headcount, um, and maybe closer to 45%, but 70% of the wage bill.
• 10:17
  SPEAKER_01: Yeah, the decision factory is the economy.
• 10:20
  SPEAKER_01: Yeah.
• 10:20
  SPEAKER_01: And his framing on why it's inefficient was the line that really stuck with me.
• 10:24
  SPEAKER_01: He said that decision factories are are really inefficient.
• 10:27
  SPEAKER_01: Factory factories, I keep getting more efficient, uh, which makes sense.
• 10:31
  SPEAKER_01: But the productivity growth in factories is going up ever, except for in decision factories, they are because they're not treated with the same productivity goals and metrics and discipline that regular factory factories are.
• 10:42
  SPEAKER_01: And then the AI beat, which was harder to hear.
• 10:46
  SPEAKER_01: Yeah.
• 10:46
  SPEAKER_01: He made a really strong argument that managers had spent too much time on algorithmic tasks.
• 10:53
  SPEAKER_01: So AI arrived, uh, did those tasks facts faster, more accurately, and then at a fraction of the cost.
• 11:00
  SPEAKER_01: And the work that would have made them irreplaceable, the mystery made the managers irreplaceable, uh the mystery work, the heuristics, the like figuring out how are we gonna how are we gonna do this kind of stuff, which is based on his knowledge funnel, that was the work that they'd never been taught or rewarded to do.
• 11:19
  SPEAKER_00: So they get exposed by the tool that could have augmented them if they'd be doing the harder work all along.
• 11:26
  SPEAKER_01: Yeah, and so Roger's point wasn't celebratory.
• 11:29
  SPEAKER_01: It was clear that this isn't the promising path.
• 11:32
  SPEAKER_01: The promising path is augmentation in his mind.
• 11:36
  SPEAKER_01: AI combined with people and how they do things, not replacement.
• 11:40
  SPEAKER_00: Yeah, I want to push on that framing though, just a little bit because decision factory is a big claim.
• 11:47
  SPEAKER_00: It works if you're inside a large enterprise with meetings that reproduce meetings.
• 11:52
  SPEAKER_00: But a lot of marketing teams aren't that.
• 11:55
  SPEAKER_00: Six people, one PM tool, one shared doc.
• 11:58
  SPEAKER_00: Is Roger's frame still the right one for them, though?
• 12:01
  SPEAKER_01: I I that's fair.
• 12:02
  SPEAKER_01: I think the honest answer is probably that the size changes.
• 12:06
  SPEAKER_01: Um, but the pattern still is true.
• 12:08
  SPEAKER_01: Like, and really, knowledge workers are knowledge workers in a lot of cases.
• 12:13
  SPEAKER_01: Even a six-person team has to uh has to make decisions and they have to hand off decisions from one person to the next.
• 12:20
  SPEAKER_01: The brief goes to creative, the creative goes to media, media goes to measurement, measurement goes back to the brief.
• 12:24
  SPEAKER_01: And even if you're an individual, you're still doing all those things, but just passing along to yourself.
• 12:29
  SPEAKER_01: And every one of those handoffs is a mini decision factory, and AI attacks the material at each handoff, it doesn't touch the handoffs.
• 12:36
  SPEAKER_00: So the bottleneck is between people, not inside them.
• 12:39
  SPEAKER_01: Yeah, and this is where leaders are getting it right or gaining advantage.
• 12:43
  SPEAKER_01: I think they're not treating AI as a productivity layer on top of existing ones.
• 12:47
  SPEAKER_01: They're actually redesigning the role and the workflow and the decision rights together.
• 12:52
  SPEAKER_01: And so McKinsey's 2025 State of AI report keeps surfacing this.
• 12:57
  SPEAKER_01: Um, the value comes from the workflow redesign, not necessarily the tool adoption.
• 13:03
  SPEAKER_00: Which is the hardest thing an organization can really do.
• 13:06
  SPEAKER_01: Yeah, I think so.
• 13:07
  SPEAKER_01: And and it's I think why only 5% of the people are pulling ahead or organizations are pulling ahead.
• 13:13
  SPEAKER_01: The tool is actually the easiest part of all this thing.
• 13:16
  SPEAKER_01: Like getting a subscription is so simple.
• 13:18
  SPEAKER_01: Yeah.
• 13:19
  SPEAKER_01: Um, but redesigning what the tool sits inside of is the actual job.
• 13:24
  SPEAKER_01: There's an onboarding component, like there is with regular staff.
• 13:28
  SPEAKER_00: Absolutely.
• 13:28
  SPEAKER_00: Most of the AI capability conversation marketing is built on one comfortable assumption, though.
• 13:34
  SPEAKER_01: Yeah.
• 13:35
  SPEAKER_01: That AI is really good at the boring stuff, the volume stuff, the first draft stuff.
• 13:39
  SPEAKER_00: And human marketers stay valuable by handling the strategic, the creative, the senior level thinking, which is kind of odd.
• 13:47
  SPEAKER_00: Yeah.
• 13:47
  SPEAKER_01: Well, the data doesn't support that assumption anymore, though.
• 13:50
  SPEAKER_00: You're right.
• 13:51
  SPEAKER_00: In September 2025, OpenAI released a benchmark called GDP Val.
• 13:56
  SPEAKER_00: The top model matched or beat human experts 47.6% of the time.
• 14:02
  SPEAKER_01: Yeah, and this is shocking.
• 14:03
  SPEAKER_01: Like when I started writing this Pi-shaped marketer article, V, I was trying to like figure out what this is.
• 14:08
  SPEAKER_01: It's it will go into this more detail, but like literally in six months, a the number now is 66 or 67% of that.
• 14:17
  SPEAKER_01: So two-thirds of the time the models beat the humans.
• 14:21
  SPEAKER_01: So it's it's crazy.
• 14:22
  SPEAKER_01: It's how amazing how fast these things are rolling out.
• 14:25
  SPEAKER_00: Yeah, that's 20 percentage points in a year from a benchmark that was already testing frontier models.
• 14:31
  SPEAKER_01: Yeah, and that's the number that should keep leaders up at night, not the 67.7, it's the rate of change.
• 14:38
  SPEAKER_01: I think you said this a few times recently.
• 14:40
  SPEAKER_01: This is the worst that AI is ever gonna be.
• 14:43
  SPEAKER_01: Yeah.
• 14:43
  SPEAKER_00: And it's worth grounding what the benchmark is actually testing.
• 14:47
  SPEAKER_00: 44 occupations across the nine industries that each contribute more than 5% to the US GDP.
• 14:54
  SPEAKER_00: 1,320 tasks.
• 14:57
  SPEAKER_01: Yeah, this is the GDP vow score.
• 15:00
  SPEAKER_01: So it's built from actual professional deliverables.
• 15:04
  SPEAKER_00: Yeah, designed by experts, averaging 14 years of experience, real legal briefs, real consulting reports, real financial analysis, then blind and reviewers compared AI output against human expert output.
• 15:17
  SPEAKER_00: Frontier models complete these tests roughly a hundred times faster and a hundred times cheaper than the human experts.
• 15:24
  SPEAKER_01: Yeah, it's wild.
• 15:25
  SPEAKER_01: So for 30 years, skills meant what humans can do that machines can't.
• 15:30
  SPEAKER_01: And that framing is now structurally broken across the 44 measured occupations.
• 15:36
  SPEAKER_00: So the question isn't what humans can do, it's what humans should be doing.
• 15:41
  SPEAKER_01: Yeah, I had interviewed uh Laura Reese at WPP, and she had a really interesting uh version of this um while I was doing some research on this PyShirps PieShape series I was writing for work.
• 15:54
  SPEAKER_01: And what did she land on?
• 15:55
  SPEAKER_01: Yeah, she had this line though that AI is elevating the average.
• 15:58
  SPEAKER_01: It lifts the floor, and and you really need people to raise the ceiling.
• 16:03
  SPEAKER_00: And that's a capability trap.
• 16:05
  SPEAKER_00: Organizations are trying to make every individual marketer faster at the tasks AI just disrupted.
• 16:11
  SPEAKER_01: Yeah, this is you know, back to the knowledge funnel thing.
• 16:13
  SPEAKER_01: It's the algorithmic part.
• 16:15
  SPEAKER_01: It's not the mystery part, it's the algorithmic part where we're, you know, we're we're doing things faster, faster briefs, faster decks, faster copy variants.
• 16:24
  SPEAKER_00: Mark, have you seen those like they're probably a generator, but the robots that are like doing track and field and they're running into wall.
• 16:32
  SPEAKER_00: So this is what this is like.
• 16:34
  SPEAKER_00: It's like that it's you're sprinting towards a wall.
• 16:36
  SPEAKER_00: The 5% MIT found getting actual return are doing something different.
• 16:41
  SPEAKER_00: They're reorganizing around decisions, not tasks.
• 16:44
  SPEAKER_01: Yeah, and that's super interesting.
• 16:45
  SPEAKER_01: And and the depth that matters is the depth the model can't reach alone.
• 16:49
  SPEAKER_01: And that's the thing that we often hear about judgment, stance, taste, distinctiveness, accountability.
• 16:55
  SPEAKER_00: This is the argument you've been developing in the PiShape marketer series for work as well.
• 17:00
  SPEAKER_00: Two areas of depth plus broad horizontal collaboration.
• 17:04
  SPEAKER_01: Yeah, and so the Pi shape, well, the T-shape model, um, I believe it was from the 2010s, with the GDP Val level capability off the shelf uh with these AI models.
• 17:16
  SPEAKER_01: Um, I think the single depth specialists are competing with models.
• 17:21
  SPEAKER_01: So you've got like rather than a single vertical, you've got you've got this tool that can do everything.
• 17:26
  SPEAKER_01: It's not a single vertical expert.
• 17:29
  SPEAKER_01: It's it's an expert in like 44 at least different roles.
• 17:34
  SPEAKER_01: And so if we're trying to compete against that, that's really a losing position.
• 17:38
  SPEAKER_00: PieShape isn't aspirational anymore, though.
• 17:40
  SPEAKER_00: It's it feels like that's just become the floor.
• 17:43
  SPEAKER_01: Yeah, so the idea about PyShape is that you have your expertise plus, and you pair that with the AI.
• 17:49
  SPEAKER_01: Um and then the line across the top, the bar, that's the operational layer where good operational thinking can be programmed into there as well.
• 17:58
  SPEAKER_01: And so what does that do in the hiring conversation, though, V?
• 18:01
  SPEAKER_00: It really rewrites it.
• 18:02
  SPEAKER_00: The question stops being do you have this skill?
• 18:05
  SPEAKER_00: And starts from what can you do that the model cannot do?
• 18:09
  SPEAKER_01: Yeah, and the CMOs getting this right, and leaders in general that are getting this right, are hiring, I think, for judgment first, and then the executable skills second.
• 18:18
  SPEAKER_00: The one still stacking specialists are watching the team getting expensive while the output stays average.
• 18:24
  SPEAKER_01: Yeah.
• 18:25
  SPEAKER_01: And there's one other thing that's worth naming here.
• 18:27
  SPEAKER_01: I thought about this a lot.
• 18:29
  SPEAKER_01: That Marty Newmeyer had this great article from 2008 where he identified wicked problems, the most wicked problems in business.
• 18:37
  SPEAKER_01: And the wicked problems being things that are so challenging that they're they're almost impossible to solve.
• 18:42
  SPEAKER_01: And things like balancing short and long-term um priorities is one of them.
• 18:47
  SPEAKER_01: That might sound familiar.
• 18:49
  SPEAKER_01: Uh the thing that's crazy about the list is for the most part, the wicked problems have not changed.
• 18:54
  SPEAKER_00: No, and the dominant AI use case in marketing right now is copywriting.
• 18:59
  unknown: Right.
• 18:59
  SPEAKER_00: And so we have yeah.
• 19:02
  SPEAKER_01: We have one of the most powerful pattern matching technology tools ever built, and we're using it to make easy stuff faster instead of trying to solve the hard, wicked problems, the hard stuff that used to be impossible.
• 19:17
  SPEAKER_01: We could be applying it to that.
• 19:18
  SPEAKER_00: That's a capability trap at the industry level.
• 19:21
  SPEAKER_01: Yeah.
• 19:22
  SPEAKER_01: I wanted to talk about this other thing that I did about a year ago via this workshop.
• 19:26
  SPEAKER_01: Um, and it was a half day through the marketing association that I ran.
• 19:29
  SPEAKER_01: So we had 15 people in the room.
• 19:31
  SPEAKER_01: We had a mix of agency and brand side people, um, some senior, some junior.
• 19:36
  SPEAKER_01: It's pretty cool, like creatives, media account, planners, strategists.
• 19:40
  SPEAKER_01: And the workshop was built on that McCain when the chips are down case study that Les Bonette and DDB did.
• 19:46
  SPEAKER_01: Um it had one of, I think it was the most the best, best uh IPA effectiveness award winner.
• 19:53
  SPEAKER_00: Okay.
• 19:54
  SPEAKER_01: And so we swapped French fries for Progies um as a teaching device.
• 19:58
  SPEAKER_01: So it was a fictional brand, real problem though.
• 20:01
  SPEAKER_01: Private label that was eating shelf space for this for this uh fictitious brand that we created.
• 20:06
  SPEAKER_01: Um private label is cheaper and the brand needed to defend margins.
• 20:10
  SPEAKER_01: So we broke the group up into three.
• 20:12
  SPEAKER_01: And so one good group did pure vibe marketing.
• 20:15
  SPEAKER_01: They were just prompting the AI and just following what the AI told them to do.
• 20:20
  SPEAKER_01: Uh one group used familiar frameworks like the funnel Maslow's hierarchy, start with Y.
• 20:25
  SPEAKER_01: I would call them the traditional marketers, and then one group used research-grounded prompts, so like the things like the 595 rule, mental and physical availability, distinctive brand assets, 6040 rule, those kinds of things.
• 20:37
  SPEAKER_01: So there were three distinct groups that were trying to solve this problem and come up with a campaign at the end.
• 20:42
  SPEAKER_01: Um, so the same brief, they had about 90 minutes of working time, and everyone presented at the end um and had some, and that's where the workshop really actually happened.
• 20:53
  SPEAKER_00: I I gotta know what what did you what did you see?
• 20:55
  SPEAKER_01: Yeah, so it was super interesting.
• 20:57
  SPEAKER_01: Everyone honestly produced something impressive.
• 21:00
  SPEAKER_01: Like it was the first thing that I had to say is like it was pretty awesome.
• 21:04
  SPEAKER_01: Like people making videos and jingles and characters and all kinds of stuff.
• 21:10
  SPEAKER_01: Um, and and and polished decks with report and research, and they did it all in 90 minutes, which was pretty insane.
• 21:17
  SPEAKER_01: Um, and so but nobody walked in that room, walked out thinking AI can't produce the work.
• 21:25
  SPEAKER_00: You've told me this story a few times, and the part that always gets me is that the vibe group's presentation, yeah, their output was polished.
• 21:34
  SPEAKER_01: It looked so good, it was so slick.
• 21:36
  SPEAKER_01: It looked like a real pitch deck.
• 21:38
  SPEAKER_01: Um and the only problem was that it didn't really look like a pitch deck for this brand.
• 21:43
  SPEAKER_01: It was completely off brand.
• 21:45
  SPEAKER_01: I'm sure they knew, right?
• 21:46
  SPEAKER_01: Yeah, well, they called it out, and that was the part it was interesting.
• 21:49
  SPEAKER_01: When they got up to present, they told the room, I don't know where this number came from.
• 21:53
  SPEAKER_01: I don't know where these characters came from.
• 21:56
  SPEAKER_01: I just did it and we went along with it.
• 21:57
  SPEAKER_01: And they it really hadn't involved any.
• 22:00
  SPEAKER_01: Any kind of critical thinking.
• 22:01
  SPEAKER_01: They just followed the prompts.
• 22:02
  SPEAKER_00: So it's funny because the failure wasn't the output.
• 22:06
  SPEAKER_00: It was that they couldn't defend it.
• 22:08
  SPEAKER_00: I've seen that in our own work.
• 22:09
  SPEAKER_00: You know, the the AI produces something, it looks right, it feels right.
• 22:13
  SPEAKER_00: And then somebody in the meeting asks, hey, what is that?
• 22:16
  SPEAKER_00: Let me look let's go a little bit deeper here.
• 22:19
  SPEAKER_00: And then there isn't an answer.
• 22:20
  SPEAKER_00: You can feel the room shift immediately.
• 22:22
  SPEAKER_01: Yeah.
• 22:23
  SPEAKER_01: And then the traditional framework group was more confident because we're familiar with some a lot of that stuff, like Maslow and Tofu and Star Wars V.
• 22:31
  SPEAKER_01: They sound familiar, but the familiar isn't the same as true.
• 22:34
  SPEAKER_00: The evidence-based group is where I always want to poke, though.
• 22:38
  SPEAKER_00: They had the frameworks, the ones that like you and I live and breathe religiously.
• 22:43
  SPEAKER_00: Did that actually produce better creative or just more defensible creative?
• 22:47
  SPEAKER_01: It's it's a great question.
• 22:49
  SPEAKER_01: The honest answer is that we didn't have to present to any CFOs and ask for actual money.
• 22:55
  SPEAKER_01: Um and really the the honest answer was that they had what they had was the ability to explain their choices.
• 23:02
  SPEAKER_01: Um why this over that, what the research says, what the business outcome should be.
• 23:07
  SPEAKER_01: When someone on the other side of the table could look at that and say, yeah, that makes sense from a business point of view.
• 23:12
  SPEAKER_01: Whether the creative itself was better is really a separate call, but it was defensible, I would argue.
• 23:19
  SPEAKER_00: Yeah, and unless you knew the frameworks, you probably couldn't tell which output was which.
• 23:23
  SPEAKER_00: All three decks looked polished.
• 23:25
  SPEAKER_00: The difference was invisible from the outside.
• 23:28
  SPEAKER_00: True.
• 23:28
  SPEAKER_01: And as I'm just thinking about this, the one thing that they didn't know that a lot of this stuff was new to them.
• 23:33
  SPEAKER_01: So they didn't have the confidence that they did in presenting the funnel, the Adafright model, the Maslow hierarchy, the start with why.
• 23:41
  SPEAKER_01: So it it's not just that it's there, it's also that you need to know it.
• 23:45
  SPEAKER_01: And so, you know, when you start um looking at what this means, uh the check has to happen inside, which means the check has to happen inside the room, V.
• 23:56
  SPEAKER_01: Before the deck gets pitched, someone has to be able to look at a piece of work and say there there are no distinctive brand assets in this, or you know, the target isn't the light buyer, and here's why that matters.
• 24:06
  SPEAKER_01: If nobody in the room can do that, it doesn't matter what the output is.
• 24:10
  SPEAKER_01: It just that's the critical thinking part.
• 24:12
  SPEAKER_01: If nobody in the room can do that, the output, the output of the AI just ships without any checks.
• 24:18
  SPEAKER_00: Yeah.
• 24:19
  SPEAKER_00: And what would that check catch on the vibe marketing deck, though?
• 24:22
  SPEAKER_01: Most likely all of it.
• 24:23
  SPEAKER_01: Um that's the point.
• 24:26
  SPEAKER_01: The polished but off-brand output is invisible to the room that doesn't know what on-brand looks like.
• 24:32
  SPEAKER_01: AI didn't cause that, but it's AI's speed that makes it much more expensive to find out later.
• 24:37
  SPEAKER_00: You didn't pitch any of it to a CFO either, though, right?
• 24:40
  SPEAKER_01: No, and and not an executive.
• 24:42
  SPEAKER_01: It was more just a workshop to see if we could do this.
• 24:45
  SPEAKER_01: Um, again, this is like a little over a year ago now.
• 24:48
  SPEAKER_01: It was just trying to figure out what we could do with with AI in a work setting.
• 24:52
  SPEAKER_01: And so the real test really is ultimately what happens when someone asks, you know, why did you choose this?
• 24:57
  SPEAKER_01: And the person presenting either says, I I don't know, or, you know, here's exactly my reason.
• 25:03
  SPEAKER_00: Yeah, producing creative is one thing.
• 25:04
  SPEAKER_00: Producing creative that's on brand, on strategy, and defensible is an entirely different thing as well.
• 25:10
  SPEAKER_01: Yeah, and AI won't get you the second one on its own without you guiding it, which is the leadership conversation, even with your own AI.
• 25:17
  SPEAKER_01: There's a bigger thing sitting underneath that though.
• 25:20
  SPEAKER_01: Yeah, there is.
• 25:20
  SPEAKER_01: I think we built modern marketing on a stack of assumptions that nobody seriously interrogates.
• 25:26
  SPEAKER_01: So we optimize to attribution models that measure what's measurable, not what matters, say like incrementality.
• 25:34
  SPEAKER_01: And so we chase digital metrics because they're everywhere.
• 25:38
  SPEAKER_01: They're digital, they're showing up on every dashboard, but not because they're the right matrix metrics.
• 25:44
  SPEAKER_01: Um, you know, we accept the frameworks because they're familiar, but not because they're true.
• 25:48
  SPEAKER_00: Yeah, I think great examples here.
• 25:50
  SPEAKER_00: Take Maslow's hierarchy of needs.
• 25:52
  SPEAKER_00: Everyone knows it, everyone uses it.
• 25:54
  SPEAKER_00: It's in every introductory marketing textbook, every brand strategy deck, every customer journey workshop.
• 26:00
  SPEAKER_01: Yeah, and if you look at the actual academic record of it, the hierarchy as it's taught isn't what Maslow ever originally proposed.
• 26:07
  SPEAKER_01: The pyramid wasn't even his diagram.
• 26:09
  SPEAKER_01: The empirical evidence supporting that is really quite pretty thin.
• 26:14
  SPEAKER_00: Yeah, the fascinating thing though is nobody questions it because everyone just knows it.
• 26:18
  SPEAKER_01: Right, which is exactly how AI gets things wrong.
• 26:21
  SPEAKER_01: The model has to read every marketing textbook that copied the wrong version of Maslow, and it will hand you this the same wrong version with extraordinary fluency.
• 26:32
  SPEAKER_01: And that's the sourcing problem.
• 26:34
  SPEAKER_01: And it's everywhere in marketing.
• 26:36
  SPEAKER_00: Now let's let's get real concrete on measurement piece.
• 26:39
  SPEAKER_00: We measure what's easy to measure, not what's right to measure.
• 26:42
  SPEAKER_00: The streetlight effect.
• 26:44
  SPEAKER_00: Looking for the keys under the lamp because that's where the light is.
• 26:47
  SPEAKER_01: Yeah.
• 26:48
  SPEAKER_01: And in marketing, the streetlight is digital, I would say.
• 26:51
  SPEAKER_01: We can click or we can measure clicks, sessions, conversions, last click, attribution, paid social ROAS.
• 26:57
  SPEAKER_01: So we optimize against those metrics and treat the resulting picture as if it were the whole thing.
• 27:03
  SPEAKER_00: Here's the picture it actually is.
• 27:04
  SPEAKER_00: In Q4 of 2025, US retail sales totaled what$1.9 trillion.
• 27:11
  SPEAKER_00: E-commerce was$316 billion of that, roughly 17% of total retail.
• 27:17
  SPEAKER_00: That share is the highest it's ever been and has been climbing for two decades.
• 27:22
  SPEAKER_01: Yeah, which means that 83, the it's just roundup feed.
• 27:25
  SPEAKER_01: Let's call it 84% of US retail happened offline in stores through dealers in environments where digital attribution, the stack can't see them and can't measure them, and frequently doesn't even acknowledge them.
• 27:39
  SPEAKER_00: Exactly.
• 27:40
  SPEAKER_00: We've built marketing optimization regimes around that 17% because that's where the streetlight is.
• 27:45
  SPEAKER_00: And we treat the 84% as if it's noise.
• 27:47
  SPEAKER_00: AI is going to amplify whichever side of that ratio you're paying attention to.
• 27:51
  SPEAKER_01: Yeah.
• 27:52
  SPEAKER_01: And if your inputs are digital attribution data, your AI is going to make you very efficient at fishing in a pond that holds one out of every six fish in the lake.
• 28:03
  SPEAKER_00: There's research from Capero in Sweden on the future marketing departments that I know you and I have talked about quite a few times that puts numbers on this in practice.
• 28:13
  SPEAKER_00: 98% of B2C companies communicate in more than 11 channels.
• 28:17
  SPEAKER_00: Activities up, content production up, budgets are flat.
• 28:20
  SPEAKER_00: The Capero's team framing on this is the line that I I think you and I keep coming back to.
• 28:25
  SPEAKER_00: Busy is the new stupid.
• 28:28
  SPEAKER_00: Yeah.
• 28:28
  SPEAKER_00: We're really good at doing more of the wrong things faster.
• 28:32
  SPEAKER_00: AI accelerates that pattern unless someone or a leader intervenes.
• 28:37
  SPEAKER_01: Yeah, and the job here isn't to fix attribution.
• 28:40
  SPEAKER_01: The job is to admit that the measurement models are wrong to begin with.
• 28:45
  SPEAKER_01: And then make decisions that account for what the model can't see.
• 28:49
  SPEAKER_01: So AI doesn't fix any of this, really.
• 28:51
  SPEAKER_01: Leadership does.
• 28:52
  SPEAKER_01: Or it doesn't.
• 28:53
  SPEAKER_01: And the 5% keeps looking like a mystery.
• 28:56
  SPEAKER_00: This is where the argument lands.
• 28:58
  SPEAKER_00: There's a book we both came into, uh Liz Wiseman's Multipliers.
• 29:02
  SPEAKER_00: Originally, what, 2010, revised edition since her thesis is that leaders come in two kinds diminishers and multipliers.
• 29:10
  SPEAKER_00: Now, multipliers get more out of their teams without working them harder.
• 29:14
  SPEAKER_00: They extract intelligence rather than imposing it.
• 29:17
  SPEAKER_01: Yeah, it's one of my favorite books for sure.
• 29:19
  SPEAKER_01: And what strikes me every time I I've read or reread um her book is that how little of the AI conversation touches multiplication.
• 29:28
  SPEAKER_01: And maybe not surprisingly, because people don't always read the same books, but every vendor pitches about productivity.
• 29:36
  SPEAKER_01: Productivity is an individual game where multiplication is an organizational level game.
• 29:41
  SPEAKER_00: Productivity tools make individuals slightly faster.
• 29:44
  SPEAKER_00: Multipliers make organizations dramatically more capable because the leader's thinking propagates through the team.
• 29:51
  SPEAKER_01: Yeah, which is what AI should be.
• 29:53
  SPEAKER_01: It should be multiplication.
• 29:54
  SPEAKER_01: It should take what a leader knows and let it scale to everyone in the org who needs it.
• 29:59
  SPEAKER_00: And it only happens if the leader puts their thinking into the system, not the tool, the system, the the standards, the training, the rules for success, the description of what good actually looks like.
• 30:12
  SPEAKER_01: Yeah, and there's a song that stuck, uh the title of it stuck with me for years.
• 30:16
  SPEAKER_01: It's so it came from a brand strategist I work with named Ross Smith.
• 30:19
  SPEAKER_01: He's a super awesome guy.
• 30:21
  SPEAKER_01: Uh he used to walk into the room with this theme song playing.
• 30:25
  SPEAKER_01: Um was it Ain't What You Do, It's the Way That You Do It by Banana Ram and Funboy 3, not the Elephant's Gerald one.
• 30:31
  SPEAKER_01: Though it's also very good.
• 30:33
  SPEAKER_01: That's the AI argument in one line.
• 30:35
  SPEAKER_01: Yeah, everyone has the tools, everyone has the same models.
• 30:38
  SPEAKER_01: The differentiation is no longer in the what.
• 30:41
  SPEAKER_01: It's in the how.
• 30:42
  SPEAKER_01: It's in the way that the leader shows up, the way the team executes, the way the standards are encoded.
• 30:47
  SPEAKER_01: The the widgets themselves are pretty much commoditized.
• 30:51
  SPEAKER_01: The way you make the widgets isn't, though.
• 30:53
  SPEAKER_00: Yeah, you also built a two by two around that framing for the Pi Shape series that you did.
• 30:57
  SPEAKER_01: Yeah, it's the it's really like what you do on one vertical and the way you do it on another.
• 31:02
  SPEAKER_01: Um but I think the frame itself carries here too.
• 31:06
  SPEAKER_00: Yeah, the other thing that you've been circling around for a while is where the leader actually stands, behind a team or in front of it.
• 31:12
  SPEAKER_01: Yeah, and I think the honest answer is to that both really have trade-offs, real trade-offs.
• 31:18
  SPEAKER_01: Leading from the front is what most arg org charts assume the traditional pyramid, the CEOs in the front, the team behind, the leader pulls, um, and the whole chain of command thing.
• 31:28
  SPEAKER_01: It works when the road is straight and when the terrain is predictable.
• 31:32
  SPEAKER_00: Leading from behind is the opposite.
• 31:34
  SPEAKER_00: Leader holds the compass, team fans out ahead, works when the terrain is complex, and the team needs to make the calls in real time.
• 31:42
  SPEAKER_01: Yeah, neither one is universally right.
• 31:45
  SPEAKER_01: The choice needs to ultimately match the terrain.
• 31:47
  SPEAKER_00: But AI changes the math on something like this because your judgment can be embedded in the system.
• 31:52
  SPEAKER_01: Yeah, which means your standards are at ultimately every decision point, guiding the team while they execute.
• 31:58
  SPEAKER_01: And you can also be at the back watching and watching the bearing and adjusting course.
• 32:02
  SPEAKER_00: You don't have to trade one for the other, essentially.
• 32:05
  SPEAKER_01: Right.
• 32:05
  SPEAKER_01: And this is a technology that I think finally lets a leader be in two places at once if they do the work.
• 32:12
  SPEAKER_00: And this isn't a new idea.
• 32:13
  SPEAKER_00: That's the piece that surprised me most when you sent it.
• 32:17
  SPEAKER_00: Someone worked this out in 1855.
• 32:20
  SPEAKER_00: It's crazy.
• 32:21
  SPEAKER_00: I know.
• 32:23
  SPEAKER_01: So God.
• 32:24
  SPEAKER_01: When I was doing the research for that article, I I've I stumbled across this work from Daniel McCallum, who's an engineer that at the time was running the Newark and Erie Railroads, uh building of them.
• 32:36
  SPEAKER_01: And he drew what's generally recognized as the first modern organizational chart, the first work chart.
• 32:42
  SPEAKER_01: Um, most people have heard of it, um, but probably have never actually seen it.
• 32:47
  SPEAKER_01: And the shape is the reveal.
• 32:49
  SPEAKER_01: It's not a pyramid.
• 32:50
  SPEAKER_01: No, it's it's so interesting.
• 32:52
  SPEAKER_01: He designed it around a tree.
• 32:54
  SPEAKER_01: Um, so the board of the directors and the superintendent and really in the executive team are at the roots, at the bottom.
• 33:02
  SPEAKER_01: The branches grow upward.
• 33:04
  SPEAKER_01: Uh, leaves at the top are the workers running the trains, repairing the bridges, working the stations.
• 33:09
  SPEAKER_01: And McCallum actually called it uh a fountain of power.
• 33:13
  SPEAKER_01: Leaders are at the back, the frontline workers at what he called the sharp edge of work.
• 33:17
  SPEAKER_00: Sharp edge is the phrase that really gets me because that's exactly where AI augmented workers are operating right now.
• 33:24
  SPEAKER_01: Yeah, and McCallum designed it that way because the railroad had new technology at its time.
• 33:30
  SPEAKER_01: It sounds crazy for us now, but um, that was breaking their operational model.
• 33:35
  SPEAKER_01: And so they had this thing called the the fancy new tool called the telegraphy.
• 33:39
  SPEAKER_01: And for the first time, information could move faster than the trains and and faster than people.
• 33:46
  SPEAKER_01: And so the old command structure couldn't keep up with what the telegraph was surfacing in real time.
• 33:52
  SPEAKER_00: And here's a line that ties us all together.
• 33:54
  SPEAKER_00: McKellum said this back in 1855.
• 33:56
  SPEAKER_00: Any system which might be applicable to the business and extent of a short road would be found entirely inadequate to the wants of a long one.
• 34:04
  SPEAKER_01: Yeah, in other words, the operating operating models built for one scale of complexity.
• 34:09
  SPEAKER_01: So that's the short road part, um, break when the scale changes, the long road.
• 34:14
  SPEAKER_01: Exactly.
• 34:14
  SPEAKER_01: And the AI moment is the scale change that we're talking about here.
• 34:19
  SPEAKER_01: What worked on the short roads doesn't actually work on the long ones.
• 34:22
  SPEAKER_00: Yeah, Caitlin Rosinolf, the Berkeley historian who did the primary work on McCallum, wrote about this for McKinsey.
• 34:30
  SPEAKER_00: Her thesis is that every major information technology revolution forces a corresponding revolution in how organizations are structured.
• 34:37
  SPEAKER_00: The companies that figure that structural change capture that value.
• 34:41
  SPEAKER_00: The ones that bolt the new technology onto an old org structure spend a fortune and get nothing.
• 34:47
  SPEAKER_00: Sound familiar?
• 34:48
  SPEAKER_00: Which sounds a little bit familiar to us.
• 34:52
  SPEAKER_00: So what does the leader actually do with this now?
• 34:54
  SPEAKER_00: Because we've argued the case.
• 34:55
  SPEAKER_00: What's that move on a Monday morning?
• 34:57
  SPEAKER_01: Yeah, Laura Wees had this line uh when I interviewed her for that article I was writing that AI is great and it gives us speed, but much more importantly, it gives us space.
• 35:06
  SPEAKER_01: And the big million-dollar question for organizations now is what do they do with that space?
• 35:12
  SPEAKER_00: That's a question.
• 35:13
  SPEAKER_00: And it points at three concrete moves.
• 35:16
  SPEAKER_01: First, I think write down the standard, not the strategy deck, the actual standards.
• 35:21
  SPEAKER_01: What does good look like for your company in your category, for your brand?
• 35:25
  SPEAKER_01: And it's sort of those unwritten rules that's just you know pass from person to person.
• 35:30
  SPEAKER_01: The things that you would never accept.
• 35:32
  SPEAKER_01: Put it into a knowledge base that AI can use.
• 35:34
  SPEAKER_01: And this is the leadership work that the tool can't do for you.
• 35:38
  SPEAKER_01: Everything else really kind of depends on it.
• 35:39
  SPEAKER_00: The second, stop optimizing individuals, start multiplying teams.
• 35:44
  SPEAKER_00: The instinct is to send people to AI training.
• 35:47
  SPEAKER_00: The multipliers move is the opposite.
• 35:50
  SPEAKER_00: Build the operating system once and let the team execute against it.
• 35:54
  SPEAKER_00: AI without leadership extracts the mode.
• 35:57
  SPEAKER_00: AI with leadership extracts the multiplier.
• 36:00
  SPEAKER_00: It's a good one.
• 36:00
  SPEAKER_01: And third, I think is that you can lead from both ends.
• 36:03
  SPEAKER_01: If you put your judgment into the system so it's present at every decision point that the team makes, um, then the playbook holds for the whole line.
• 36:13
  SPEAKER_01: And then you can stay at the back and watch the bearing.
• 36:15
  SPEAKER_01: So you're both at the front and at the back at the same time, and the team moves between you.
• 36:21
  SPEAKER_00: The 5% are building operating models, the 95% are buying tools.
• 36:25
  SPEAKER_01: And the difference is whether the leader has done the hard work of putting their thinking into the system and then had the discipline to step back and let the team multiply it.
• 36:34
  SPEAKER_00: AI won't save you, leadership will.
• 36:37
  SPEAKER_01: Or it won't.
• 36:38
  SPEAKER_01: And we'll all be back here in 18 months reading the next study that tells us why.
• 36:42
  SPEAKER_01: Thanks for keeping us in yours, everyone.
• 36:44
  SPEAKER_01: Stay sharp.