Episode 69: WPP's CTO spends over $300 million per year on AI

Summary

In this episode of the Architecture Podcast, AR Paparo engages in a deep dive with Eric Franchi and Stefan Pretorius, CTO of WPP, discussing the impact of AI on the advertising industry. They explore Innovid's Harmony initiative for open TV advertising, the evolution of adtech, and the substantial investments WPP is making in AI. The conversation covers AI's role in creative processes, media planning, and knowledge management, highlighting the practical applications and challenges of integrating AI into agency workflows.

Takeaways

• 🌐 Innovid's Harmony initiative aims to open up TV advertising and address challenges by optimizing at the infrastructure level, focusing on efficiency, transparency, ROI, carbon emissions reduction, and viewer experience enhancement.
• 📝 The podcast features an interview with Equative, discussing their investment in SSP, DSP, and curation, highlighting the company's growth and transformation in the adtech industry.
• 🤝 AR Paparo and Stefan Pretorius have known each other since the dot-com era, discussing their history in digital advertising, lead generation, and the evolution of their careers in the industry.
• 💡 Stefan shares his role as CTO of WPP, emphasizing the importance of centralized technology strategies and the integration of AI in content perception, labeling, and recommendation systems.
• 🚀 WPP invests heavily in AI, with $300 million annually, reflecting a significant commitment to transformational technology and its potential impact on the marketing services industry.
• 🎨 AI's role in creative processes is explored, discussing the shift from physical to virtual production, and the use of generative AI and virtual production technologies to create content efficiently.
• 📊 Media planning is identified as an area ripe for AI application, with a focus on finding more accurate theories of relevance and indicators through occasion-based targeting and contextual signals.
• 🤖 Knowledge management in large corporations is highlighted as a significant opportunity for AI, with the potential to build digital twins of organizations and enhance decision-making through reinforcement learning.
• 📉 Google's delay of cookie deprecation is discussed, noting the impact on industry certainty and the move towards alternative solutions like clean rooms and federated learning.
• 🛑 The potential implications of the TikTok ban in the US are considered, including the redistribution of ad revenue and the importance of the platform for media buyers and client engagement.
• 💸 Meta's financial report and stock market reaction are analyzed, reflecting on the company's investment in AI and the potential for open-source AI to create new monetization opportunities.

Q & A

What is Innovid's Harmony initiative about?

Innovid's Harmony initiative is an industrywide effort to address the major challenges in Connected TV (CTV) advertising by optimizing at the infrastructure level. It aims to improve efficiency, enhance transparency, increase ROI, reduce carbon emissions, and provide better viewing experiences.

What was the purpose of the Blink software platform mentioned in the podcast?

The Blink software platform was designed to allow users to upload and save their bookmarks across different browsers and computers. However, the business model shifted towards spam as the primary monetization strategy involved sending numerous promotional emails to users.

How does WPP's CTO view the role of AI in transforming the advertising industry?

WPP's CTO sees AI as a transformative technology that is fundamentally changing the marketing services industry. He emphasizes the importance of building an integrated platform and strategy that can service the current business while transitioning into the future business landscape influenced by AI.

What is the significance of the 'four core brains of marketing' concept introduced by WPP's CTO?

The 'four core brains of marketing' concept represents AI services that act as AI representations of a brand, audience, channel, and performance. These AI models are designed to provide insights and recommendations for marketing strategies, enhancing decision-making in the advertising process.

How does generative AI impact the creative process in advertising according to the podcast?

Generative AI is having a significant impact on the creative process by providing tools for ideation, testing ideas, and offering different perspectives from various audience groups or personas. It also aids in the production process, enabling the creation of content through virtual production technologies and 3D rendering, rather than physical production.

What is the current state of AI in media planning as discussed in the podcast?

AI in media planning is focused on developing more accurate theories of relevance and indicators of relevance. It emphasizes the importance of context, such as time, location, and situation, over demographic details in understanding consumer behavior and preferences.

How does WPP plan to leverage AI for knowledge management within the organization?

WPP aims to build a digital twin of the organization through the adoption of reinforcement learning, where interactions with software contribute to making the AI smarter over time. This approach is expected to improve knowledge sharing and decision-making across the organization.

What is the potential impact on the advertising industry if TikTok faces a ban in the US?

If TikTok faces a ban in the US, it could lead to a significant shift in advertising revenue, with potential gains for platforms like Instagram, YouTube Shorts, and Snapchat. The loss of TikTok could also reduce diversity and competition in the digital advertising space.

What was the reaction to Meta's announcement about doubling down on AI investments?

Meta's announcement to double down on AI investments was met with skepticism from investors, leading to a significant drop in the company's stock price. Concerns were raised about the potential overinvestment in AI similar to past concerns about investments in VR and the metaverse.

What is the potential buyer for TikTok in the US if it is forced to sell its operations?

Potential buyers for TikTok in the US could include companies like Walmart or Oracle, although both may face regulatory hurdles. Private equity firms or even tech companies like Microsoft or Apple could also be potential suitors, although the high price tag may limit the pool of interested buyers.

How does the podcast host view the future of AI in the context of the advertising industry?

The podcast host believes that while AI has the potential to automate many tasks in the advertising industry, it is essential to focus on building software that augments human creativity and intelligence rather than replacing it. He emphasizes the choice between building 'black boxes' that produce results without understanding or fostering a renaissance of human ingenuity supported by AI.

Outlines

00:00

🎙️ Podcast Introduction and Innovid Harmony Initiative

The podcast begins with an introduction to the Harmony initiative by Innovid, which aims to make TV advertising more open and efficient. The host, AR Paparo, welcomes guests Eric Franchi and Stefan Pretorius, the CTO of WPP. They discuss the early days of their careers in the adtech industry, including their involvement with a digital advertising company and the evolution of their work from lead generation to adtech. The conversation also touches on the challenges and innovations in the CTV advertising space.

05:02

🛠️ WPP's Centralized Technology Strategy and AI Investment

This paragraph delves into Stefan Pretorius's role as CTO of WPP and the company's centralized technology strategy. It highlights WPP's investment in AI, with an annual spend of $318 million, which is significantly more than other holding companies. The discussion covers the importance of building flexible platforms that can adapt to the evolving needs of clients and the industry, as well as the challenges of creating original technology in-house versus licensing existing solutions.

10:04

🎨 AI's Impact on Creative and Media Planning in Advertising

The conversation shifts to the impact of AI on creative processes and media planning within WPP. It explores the use of generative AI and virtual production technologies to create content more efficiently, as well as the potential of AI to enhance ideation and concept development. The paragraph also discusses the application of AI in media planning, focusing on the importance of context and relevance in creating effective advertising strategies.

15:04

📊 AI's Role in Media Buying and the Future of Advertising

This section continues the discussion on AI's role in media buying, emphasizing the importance of understanding audience context to deliver relevant advertising. It also touches on the potential for AI to improve knowledge management within large corporations like WPP, suggesting that AI can help in building a 'digital twin' of an organization through reinforcement learning, thereby enhancing decision-making and strategy development.

20:06

🤖 The Debate Over AI's Potential to Disrupt Marketing and Advertising

The speakers engage in a debate about the potential of AI to disrupt the marketing and advertising industry. They discuss the possibility of AI replacing human strategists and creative professionals, as suggested by Sam Altman, CEO of OpenAI. While acknowledging the transformative potential of AI, they also express concerns about the loss of human creativity and ingenuity that could result from over-reliance on AI. The conversation highlights the need for a balanced approach that augments human capabilities rather than replacing them.

25:06

🚫 Google Delays Cookie Deprecation and TikTok's US Future

The paragraph discusses recent industry news, starting with Google's decision to delay the deprecation of third-party cookies for the third time. It also covers the potential implications for the ad industry if cookies are eventually phased out. Additionally, the conversation addresses the political and regulatory challenges facing TikTok in the US, including a bill signed by President Biden that could lead to a ban if the company does not comply with certain conditions.

30:07

📉 Meta's Earnings Report and Market Reaction

The speakers analyze Meta's earnings report, which showed positive results but was met with a significant drop in stock value. They discuss the market's reaction to Meta's continued investment in AI and the potential long-term benefits of these investments. The conversation also touches on the contrasting fortunes of Meta and Tesla, highlighting the unpredictability of market responses to corporate strategy and performance.

35:10

🌐 Palantir's Entry into Advertising and Data Analytics

The final paragraph covers Palantir's move into the advertising space, offering data analytics solutions for inventory planning and campaign optimization. While the details of Palantir's offerings are light, the conversation explores the potential impact of such a move on the industry. The speakers also share their perspectives on the suitability of Palantir's solutions for the fast-paced, innovative advertising sector.

Keywords

💡Harmony initiative

The Harmony initiative by Innovid is a key concept in the podcast, representing an industry-wide effort to tackle significant challenges within Connected TV (CTV) advertising. It aims to optimize the infrastructure level to improve efficiency, transparency, and ROI while reducing carbon emissions. The initiative is mentioned in the opening segment, setting the stage for discussions on innovation in advertising technology.

💡CTV advertising

CTV advertising refers to commercials delivered through connected television platforms. It is a central theme in the podcast as the Harmony initiative specifically addresses the challenges and advancements in this domain. The term is used to discuss the current state and future of television advertising in the context of Innovid's efforts.

💡ROI

ROI, or Return on Investment, is a financial metric used to measure the efficiency of an investment. In the podcast, it is mentioned in the context of the Harmony initiative's goals to increase ROI in CTV advertising, highlighting the importance of financial performance in advertising strategies.

💡carbon emissions

The term 'carbon emissions' is used to discuss the environmental impact of advertising technologies. The Harmony initiative aims to reduce carbon emissions, indicating a commitment to sustainability alongside efficiency and ROI improvements in the advertising industry.

💡AI

AI, or Artificial Intelligence, is a prevalent theme throughout the podcast. It is discussed in various contexts, including content creation, media planning, and knowledge management. The speakers highlight AI's transformative potential in the advertising industry, with specific examples such as generative AI and machine perception.

💡generative AI

Generative AI is a subset of AI that focuses on creating new content, such as images, videos, or text. In the podcast, it is mentioned as a significant opportunity in creative advertising, where it can help in ideation and content production, exemplified by the creation of car configurators and ads.

💡machine perception

Machine perception is the ability of machines to interpret and understand the world around them, often through AI. The podcast discusses its application in advertising, particularly in optimizing dynamic content optimization (DCO) by simulating and understanding the best-performing variants before going live.

💡

💡media planning

Media planning is a strategic process discussed in the podcast, which involves decisions about advertising placement and targeting. It is highlighted as an area where AI can have a significant impact, particularly in the context of data and optimization, as well as in developing new theories of relevance for advertising.

💡knowledge management

Knowledge management is the practice of capturing, distributing, and effectively using knowledge within an organization. In the podcast, it is presented as a key opportunity for AI to streamline and enhance the way organizations process information and make decisions, with a focus on reinforcement learning and digital twins.

💡federated learning

Federated learning is a machine learning approach that enables an AI model to learn from many decentralized devices or servers holding local data samples, without sharing the data itself. The podcast mentions this as one of the solutions that have been implemented to understand signal accuracy in the absence of third-party cookies.

💡third-party cookies

Third-party cookies are a topic of significant discussion in the podcast due to Google's delay in deprecating them. These cookies are used for tracking users across the web and serve as a foundation for personalized advertising. The delay in their deprecation has caused uncertainty and frustration in the industry.

💡TikTok

TikTok is mentioned in the context of potential threats due to political decisions that could force its sale or ban in the US. The podcast discusses the implications this would have on the advertising industry, particularly regarding the redistribution of its ad revenue and the impact on media buyers.

💡meta

Meta, formerly known as Facebook, is discussed in the context of its financial reports and investment in AI. The company's stock reaction to its commitment to AI investments is contrasted with other companies like Tesla, highlighting the market's perception of tech investments and their short-term impacts.

💡Palantir

Palantir is a company specializing in big data analytics platforms, known for its co-founders and its work with the US government. The podcast mentions its entry into advertising agencies, offering solutions for inventory planning and campaign optimization, which raises questions about vendor lock-in and innovation flexibility.

Highlights

Innovid's Harmony initiative aims to optimize CTV advertising by improving efficiency, transparency, and ROI while reducing carbon emissions.

Equative, formerly a smart ad server, is investing in SSP, DSP, and curation, indicating significant developments in the company.

AR Paparo and Stefan Pretorius discuss their history in adtech, highlighting the evolution from digital advertising to AI-driven content and recommendation systems.

Pretorius' role as CTO of WPP involves creating a centralized technology strategy for the agency holding company, emphasizing the importance of AI in content perception and prediction systems.

WPP invests $300 million annually in internal original technology, focusing on product data and AI research, including the acquisition of AI research lab Satalia.

The discussion explores the transformation of the advertising industry, emphasizing the need for flexible platforms that can adapt to future business needs.

Pretorius explains the concept of 'four core brains of marketing' in AI, which include brand, audience, channel, and performance brains, aiming to enhance marketing strategies.

AI's role in creative processes is discussed, with a focus on generative AI and virtual production technologies for content creation, reducing the need for physical production.

The potential of AI in media planning is highlighted, with Pretorius emphasizing the importance of context and relevance over demographic data in effective media strategies.

Knowledge management in large corporations like WPP is discussed, with AI seen as a tool for enhancing organizational learning and decision-making.

Pretorius addresses the potential impact of AI on jobs in the advertising industry, arguing for a balanced approach that augments human creativity rather than replacing it.

The conversation touches on the implications of Google delaying the deprecation of third-party cookies, suggesting a lack of progress and potential industry frustration.

The potential buyers for TikTok in the US are speculated upon, with the discussion focusing on the challenges of finding a suitable company that can handle the platform's algorithm.

The impact of TikTok's potential ban in the US on the advertising industry is considered, with Meta, Instagram, and YouTube Shorts highlighted as potential beneficiaries.

Meta's financial report and the market's reaction to their AI investments are discussed, with a focus on the potential benefits of open-source AI for the company's products.

Palantir's entry into advertising with its data analytics platform is noted, with Pretorius expressing skepticism about the suitability of such platforms for the agile advertising industry.

Transcripts

this podcast is brought to you by

innovid and their Harmony initiative

innovid believes the TV should be open

for everyone and controlled by no one

and the time to make that Vision a

reality is now that's why innovid

created Harmony an industrywide

initiative to address the biggest

challenges facing CTV advertising today

by optimizing at the infrastructure

level Harmony aims to improve efficiency

enhance transparenting the troll and

increase Roi all while reducing carbon

emissions and providing better View

experiences for all find out more and

get involved at innovid.com

[Music]

[Applause]

Harmony welcome to the architecture

podcast I'm AR paparo I am joined today

by Eric franchi and Stefan pretorius who

is the CTO of wpp and whom I have

personally known for We're Going on 30

years adtech Stefan thank you for being

here it's wonderful to be here nice to

see you gar uh so we'll talk about how

we know each other in a moment uh just a

quick housekeeping so this week's vendor

interview is with equative which is the

new name for a smart ad server it's an

interesting interview with their cro

about how they're investing in SSP DSP

and curation um so there's a lot going

on with that company you may not have

heard of it's definitely worth listening

to and it's uh free from behind the pay

wall until Monday all right let's jump

in so Stefan you and I know each other

from around the year 2000 and that is

about four or five years before I was an

adtec so many people do not think I had

a life before atch I was just born in

the double click headquarters but

actually I'm a lot older than that and

during the dotc com we work together do

you want to you want to tell the

audience you know how we know each other

I'm not sure it's gonna it's going to

bathe either us and Glory um so um I was

starting I was starting a a digital

advertis in a lead generation business

called acceleration at the time and uh

you remember were trying to flog a

software platform called blink I think

it was called right that where you could

upload your bookmarks and as you move

between browsers and computers you you'd

be able to save your bookmarks yeah

obviously Google solved that problem

with you later and that you know whole

business model disappeared but um but it

was it was fun times because we figured

out how to do large scale lead

generation track results do attribution

modeling and so in many ways I mean it's

like even though they might sound like

errand schemes now they teach you a lot

about the fundamentals and the basics of

of our industry right yeah uh I was in

The Bookmark business I was fogging

bookmarks but the secret to our bookmark

business is that it was really a Spam

business uh we we would basically get

people sign up for bookmarks and the

only way we really can monetize them was

by sending them like six emails a week

with different offers about hey add this

you know money saving scheme to your

bookmarks or golf.com you to golf add

that to your bookmark it was it was

really uh desperate times early signals

business I would say you know I mean

today would be a signal driven AI

business you know with a back by an LM

you know would have scraped all all the

content created a personal GPT for

everyone based on those bookmarks you

know it had legs you just didn't know

before we move on what was uh our our

co-host AR like as a 20-some year old

that would embarrass him but but quite a

lot skinnier yeah it's true I was pretty

skinny he was the cool guy he was the he

was the cool kid of the um of the

digital you know early.com days um he

had this great um Campbell Soup tattoo

on his arm he was like you know he had

he had Swagger um I

mean he was he was pretty he was pretty

intimidating to a lots of people even in

like a rented you know sort of Regis

office in in the B STS building right

yeah that is back here one also side

many people know this but my co-founders

at link um they were from the finance

World they went on to found to Sigma and

are now billionaires anyway I'm here as

a podcast host all right let's get to

the real subject we want to talk about

I've bet if a lot of our listeners don't

even know that wppb has a CTO like then

it's sort of like wpb CTO huh that's

interesting I didn't know they had that

what is what's your job my god um so

I've been I've been C for five years so

sold I sold that same business I I you

know started when I worked with ur2 wppp

in 2012 it became an attech modtech SI

eventually you know we implemented

everything from you know double click to

omniture to you know exact target and

then all the all the sort of subsequent

you know parents of those companies so

we sold that business to wpp in 2012 ran

it um here for a couple of years and

then in 2018 um Mark Reed rcoo asked me

to become CEO for the group you know so

it's been 5 years but you're right

historically agency holding companies

didn't really have centralized

technology strategies you know they they

effectively allowed all the agency

Brands individually to decide what they

build what they buy you know how they

innovate and I think you know it was

sort of not a not a moment too soon

because you know even back in 2018 you

know we already had the early sort of

showings of you know what AI is going to

do to content at least at that time

content perception labeling things like

that a lot of you know AI driven

prediction systems and recommendation

systems coming on board and so you know

I sort of started working really

furiously five years ago on on what does

an integrated platform and strategy look

like for an agency holding company you

know what what's amazing about the past

five years experience is that it's

really an exercise in mapping the

transformation of our industry in our

category to a technology solution so how

do you build software that actually

allows you to service both the business

you have today but also allows you to

transition into the business you're

going to have 5 years from now

you know all of you've built you know

software yourself so you know how

difficult it is just to get like one set

of requirements out and get people to

use it and so on but to build a platform

that's effectively flexible enough to

accommodate that transition while at the

same time supporting multiple different

versions or configurations of our

service offering to clients was actually

remarkably difficult intellectually I

mean technically not that hard but

intellectually very hard you know so

some clients only buy one thing from us

they only buy PR services from us or

they only buy Production Services from

us or only CRM but then other clients

like you know Coca-Cola buy everything

that we do from us end to end and you

know so you have like 6,000 people in

wppp working on the the coke account

doing everything from audience research

to Media buying and activation through

to creative production CRM and

everything in between that's what I've

been doing and um maybe because you

haven't if you haven't heard that much

from me is because I've been super busy

executing that um hope I with know with

all my my AI Keynotes at BTC and next

and things like that I'm I maybe you

know getting out a little bit more but

well I think there's there's this

feeling that when agencies invest in

Tech it's around orchestration Tech

where you're saying well you know we're

going to license an ad server from this

company and an analytics server from

this company and then the tech we're

going to build is coordination piping

you know overlays on top of that how

true is that versus building original

Tech no I I think that's slightly Legacy

view of the world I think what's

happened in the last you know

particularly last few years but but

certainly you know tracing back even

further is that all the the large

holding companies have realized that

they need to have tools and platforms

that differentiate themselves not only

in terms of how they deliver for clients

you know and the impact that that work

has but also how they work with clients

in other words how they engage with them

intake work track work etc so you know

if you look across the category all our

peers have invested in something like

this um you know or some form of this

kind of strategy in the last few years

so we spend about $300 million a year on

on internal original technology and

that's a combination of product data and

research so and and the research is

mostly on the AI side and we you know

we've invested very seriously in in

capabilities to do that so we we we

bought an AI research lab here in London

called satalia there're about um 200

people now but you know resarch grade

sort of AI researchers from UCL you know

like sort of friends of Demus you know

sort of vintage right and um you know

that's enabled us to do kind of really

really original research in in applied

AI but also you know obviously in the

product development um of you know for

what we do so our platform you know it's

built on on hyperscale of Technology but

our platform is entirely you know built

from scratch you know by ourselves and

then all the the studios that we used

for Creative production media CRM Etc is

effectively tools that we've built

ourselves to not only suit the way that

we work with our clients but also to

effectively go further than the the

regular sassin martic that's out there

um at the moment yeah let's talk numbers

first so I think correct me if I'm wrong

the publicly disclosed numers that

you're investing $318 million annually

in Ai and in in contrast the only other

holding company I was able to find that

gave a number was puess who says that

they're doing 325 million over three

years so that's like onethird the

investment and meanwhile the other the

other ones like densu really were

putting out press releases like hey

we're working with Nvidia

congratulations there was nothing really

to it those numers accurate yeah they're

accurate they're completely accurate I

mean you know we're a publicly listed

company so we can't make these things up

we have to you

know okay

sure not a venture back company you can

say whatever they want no no it it's

they are real and you know the I mean to

but to put it in context right I mean

that's 2% of net sales right you know

and and you you could argue that that is

still probably under investing in a

transformational technology if you say

that that AI is going to fundamentally

transform the entire you know marketing

Services industry and you know one of

the leading companies in the space

spending two 2% of net sales on that

impact you know the question is is

that's enough you know we are delivering

like absolute Gang Busters right now and

going super fast I I mean I personally

can't really you know basically consume

or or or manage much a much faster Pace

right now to be honest but time will

tell and I think it'll be interesting to

look back in a few years time to see did

we invest enough right well let's talk

about the use cases and what you're

investing in uh there was a very good

article that featured you in ad age this

week um that talked through a lot of

things but let's just let's kind of go

through it let's start with creative uh

what do you see as the opportunity in

Creative obviously that's kind of the

first thing that comes to mind for a lot

of people with AI yeah and and and I

think what what's what's kind of

confusing about about this stuff is that

you know if you think about the the

basic kind of image generation Gena of

AI use case people think that you know

in an agency like wpp we have you know

thousands of people who sit around you

know taking photographs and doing video

shoots and things like that all day

nothing could be further from the truth

right I mean most of that is

subcontracted to production companies or

to specialist you know Studios and

things like that you know we we do have

a production business Hogarth which is

about um 6,000 people it's a it's a

large large kind of you know tier one

production business largest commercial

production business in the world and so

we use um AI a lot but it's a

combination of generative Ai and virtual

Production Technologies to create

content so instead of instead of you

know building a car taking it to the

desert filming it and then making an ad

you can just take the cad model you know

render it in 3D and Omniverse add a

generative background you know make it

move and then create the ad and put it

out so there's a shift in in production

from physical to viral production so

that's let me just interrupt you there

how much of that is real right now

versus like that's the vision for where

it's going it's completely real I mean

we by the middle of the Year our largest

Auto client will have all their car

configurators run through that that

pipeline in live consumer facing content

we already have a a car configurator for

buid that was made exactly like that we

have you know several other deals you

know or projects in in flow you know

right now for for delivering that so

this is I mean that's completely real I

mean this is not a not a fictitious kind

of five post- production in terms of you

know maybe the stuff that the the adtech

audience is excited about in terms of

optimization personalization all of that

stuff how much of that is you know real

versus uh versus pipeline or you know

sort of like actual versus pipeline I

think when people think about when they

say creative you know they mean

different things if you ask agency

creative directors what they mean by

creative they they will say it's ideas

it's Concepts it's an inspiration it's a

way to position a product in the

marketplace product you know how to

relate a product to an audience how to

position it in culture and then

ultimately There's a Hero piece of

content there's like a hero you know

sort of you know this is the kind of

style we're looking for this but then

production is the process of taking that

PO concept and hero you know sort of

asset and then you know trans creating

it into 100 languages and adapting it to

you know a thousand different channels

and formats and doing vertical and

horizontal and doing you know things for

billboard and things for print ads and

you know I mean you know all that work

the whole content purification from that

fans out from the core idea that's the

production business now production is

about it's less than 10% of our total

revenue you know but it's an important

part where generative AI is having a big

impact but geni is also having a massive

impact in ideation right so tools to

make you think differently about things

tools to test ideas tools to have

different perspectives you know from

different audience groups or for from

different personas you know that's all

being done through generative AI so we

last year we we rolled out internally a

a creative Studio product which is like

a a gen sandbox with about 50 different

Tools in four marketing functions from

you know reinterpreting the brief to

coming up with you know shower thoughts

you know you know net new ideas about

things to building personas to you know

evaluating summarizing documents you all

these kind of things the key thing is

right and this is the I think we are

really differentiating is that using our

satalia research um Team we've built

this Con concept of the the four core

brains of marketing and brains are

really just AI Services right and the

four are a brand brain an audience brain

a channel brain and a performance brain

and what these are are effectively AI

representations of a brand an audience a

Channel or you know how that will work

all together in actual performance and

so you know simplistically you know if

you know if I was speaking to AR over

appli he would say oh well that's just a

model right said yes of course it's a

model but it's a model with lots of data

that's been validated that's been

exposed as a service that can be

consumed in software right and so let me

give you an adte example just to kind of

bring it back to the core the core

audience here one of the most impactful

applications in attic is to do is not to

do dco live but to simulate dco and

understand the best performing variance

before you go live right so you you

basically take a performance model that

says here are 100 content features you

know from the text to the logo to the

composition to the color schemes Etc

that link to some kind of performance

metric so brand lift studies for

instance that allows you to build a

model that says that creates both a

recommendation as well as explainability

so this variant will work better than

that variant because 1 2 3 4 5 if you

want this variant to work better expose

the logo at the end for another 5

Seconds you know change the position of

the product make the the brand bigger

you know whatever the case is now when

you've got a normal dco process right

you would spend the first x million

impressions optimizing your million

variants down to the high performing

ones right which means you've got

underperforming media Impressions but if

you run that entire process on the model

beforehand you only go live with the

with a Content that you know is going to

perform well right that solves a big

problem for dco solves a massive problem

for dco and I mean that that's what we

what we launched at um at Google Max

last week with um in the in Thomas's um

keynote and it's the thing you know

there's nothing generative in it it's

perception right it's the opposite there

the flip side of generative AI which is

machine perception but it's a massively

useful you know use case in in in

advertising yeah that's fascinating

let's move on from creative to media

planning sir Martin cell spoke at um at

the adtech investor conference um that

both Eric and I were at yesterday and he

specifically called out wpp just I just

want to mention that he said wpp has

45,000 people in media planning and AI

could currently do their job better than

they could it's that's a paraphrase what

he said I'm not g to ask you to respond

to that then he gave some anecdotes

about hanging out with Vladimir Putin so

that's how he

rolles

that's

anyway media planning does seem like an

area where it's all numbers it's all

data you have a ton of historical data

seems pretty ripe to have fewer human

beings doing it and more machines so

what are you doing in that area yeah I

mean it's a huge area right I mean I

think I think what's um look I think the

the first thing I would just want to say

is that that not all AI is Gena of AI

and and not all useful not not all

useful things you do with numbers and

formulas you know is necessarily an

algorithms is necessarily AI right so

there's lots of problems in media which

are just like basically discrete

mathematics you know optimization

problem they they're not they're not

necessarily AI problems right so we have

a little bit of an industry definitional

problem that everyone wants to call

everything Gena of AI or they want to

solve every problem through generative

AI which is you know not always you know

the the case but but what I would say is

in on the media side which is which I

find the most exciting is that besides

the the kind of creative optimization

and and and content intelligence is this

whole notion of finding more accurate

and better theories of relevance or

indicators of relevance right so this

whole idea of you know you can only be

effective in media planning and buying

if you know who someone is I think is

you know has to be completely debanked

right so this this notion of if I know

who you are then I can be relevant to

you is actually a fallacy because very

often who you are is a far weak weaker

indication of what your needs and and

desires are than where you are what time

of day it is what's the context that

you're in right so you might behave

entirely differently when you're at the

football on the Saturday afternoon in

the stadium then when you're sitting at

home you know with your wife watching

series you know 11:00 versus when you're

in a Superstore 10:00 on a Saturday

morning right same person same

demographic details same identifiers

right same person in the in the CRM

database but your context and what you

open to is entirely different I mean for

us things that that I'm the most excited

about is the the areas where we've

managed to use things like occasion

based targeting right or a combination

of time space and and other contextual

you know signals to understand what will

resonate with people and what I love

about that is that all that work is done

at kind of a cohort or aggregate level

right I mean sometimes in these sort of

you know these sort of cells of maybe

like half a mile right it's they they

they're not they're not sort of small

cells but for many of our clients one to

one is a complete fantasy right I mean

if he's selling fizzy drinks or he's

selling you know kind of washing powder

I mean you know no one wants to receive

you know a direct email from you know

you you're washing up Liquid right I

mean it's like not the case you know

this this super exciting stuff happening

on the onetoone side as well and all the

kind of the clean room stuff we've

always been doing and all the you know

the kind of the data matching and all

the probabilistic matching and all the

Federated learning I mean all those

things are great but for me it's like

the it's the new theories of how do you

how can you be relevant to someone that

I that I find the most interesting sure

sure no I get that um so the other area

that sir Martin called out not to just

toot his one but I thought it was pretty

interesting he said one of the Bigg

opportunities for AI in agencies is

knowledge knowledge transfer and that

it's a huge inefficiency to get everyone

on the team pointing the same direction

knowing the same things Etc and I think

your your competitor um omnom or pist

that came out with Claude uh and I think

people laughed a little bit because it's

such a silly name but that was their

first you mean Marcel Marcel was that

which one was that CLA is the anthropic

AI model yeah Claus anthropic yeah

Marcel I'm sorry U there was like a chat

bot That was supposed to uh get everyone

on the same page so how do you how do

you think about that about Knowledge

Management in a giant corporation like

wpv I think that's that's that's a

correct observation I think the in any

knowledge um based organization whether

it's a law firm or a you know a

consultancy or a you know Investment

Company I mean you know we all trade on

on knowledge and and sort of experience

and and expertise right now you know

what what I think not not to get too

philosophical but I mean AI is really a

technology about intellig

and so it's one of the first

technologies that that makes the way we

think and the way the way we process

information and and the way that our

Chain of Thought works and the way that

you know different perspectives

influence decision- making is sort of

made overt right and so I completely

agree with with the with the the notion

that it's one of the biggest

opportunities and the question is is how

do you make that kind of organizational

change so that not only is your are your

knowledge stores available but you are

building a reinforcement learning into

your software so that people people are

constantly making it smarter right so in

our creative Studio we have this this

sort of almost like a creative pinboard

you know for projects where the way that

people are interact with those things

and move things up and down and like

them and dislike them and pin them is a

reinforcement learning process to

actually tell the machine what is good

and what is bad now the longer you do

that the B it becomes I mean you know

the reason why chbt was was so uncannily

good so fast is because had 100 million

people training it day and night for the

first few weeks right so then anything

that gets launched feels a little bit

like kind of like awkward and a bit like

the the sh kid you know who doesn't

quite know the the social codes right

until it gets trained and then it

becomes better and so I think that's

that's that's the the the trick right is

is to basically build a a digital twin

of your organization through the

adoption of reinforcement learning you

know in the way that people interact

with the software yeah you can imagine

uh a big impact on account management

where the client calls up their account

person and says hey we're thinking about

doing digital ad in Supermarket

checkouts what do you think and

currently the account manager may or may

not have any ability to answer that

question but I'm sure there's a lot of

people at wppp who really could answer

that question extremely well and I

already have so that was a huge

opportunity there well and and and I

mean just to give you one one example

practical example of how how augmented

you know people can become through these

these tools right so if I say to you um

okay you work on the Fant account right

and I want to I want to know what the

best ways to to execute Fanta on Tik Tok

What are you going to say but if I say

to you here's the the Fanta brand brand

platform Playbook right it's a 100 page

PDF it's been done by strategist you

know it tells you exactly how to execute

Fant it tells you exactly what the proof

points are what the kind of the the cut

throughs are etc etc and I then tell you

here's a 200 Page Tik Tok best practice

execution guide right now you can go and

read those and then make a whole bunch

of notes and you know ponder about it

the whole weekend or what you can do is

you can upload both those documents to

our creative studio and say at the

intersection of these two things how do

I execute the Phantom brand platform on

Tik Tok it will give you an amazing set

of starting ideas an amazing set of

starting ideas none of them will be

wrong some of them will be very good you

could argue none of them are going to be

truly inventive because they are within

the it's maybe interpolation and

extrapolation within the the boundaries

of what they have but it's going to be

ne new invention because it hasn't

doesn't have the you know the ability to

do that yet but for as you said I mean

you're you know if you're an account

person the client says to you hey um you

know I feel we're a little bit

underweight on Tik Tok what's the best

way to execute our new brand platform on

Tik Tok you can come back to the to to

the client within five minutes with a

pretty good answer that's the the real

practical example of of where good input

information right so crisp articulations

of Brands good understanding of

platforms markets Etc combined with

you know well structured AI tools that

are safe secure Etc can get enormous

augmentation of of of outputs right and

clients hate paying for strategy right

so effectively you're you're boiling

down a lot of strategy for them I'd

argue clients hate paying for account

managment more than

anything Stefan um are you familiar with

the Sam Alman quote about uh Ai and

agencies I I am yes it did it did make

it

round yeah I'd imagine so so for for the

listeners I I'll blow through a fast and

then I'd love your reaction and I quote

from Sam Alman uh CEO of openai 95% of

what marketers use agency strategists

and creative Professionals for today

will easily nearly instantly and at

almost no cost be handled by the AI and

the AI will be will likely be able to

test the creative against real or

synthetic customer focus groups for

predicting results and optimizing again

all free instant and nearly perfect

images videos campaign ideas no problem

that's um that is the thought experiment

right it's exactly the thought

experiment and you know when we we

spoken publicly about this um you know a

capital Market day early in the year

when we set out on this journey I mean

we effectively had the sort of three

layer pyramid for our AI strategy you

know the first one was just getting into

productivity software get everyone you

know working faster and better second

layer is put AI into all our products

and systems and make sure that we build

better tools that differentiators the

third one is total disruption basically

put money in the one end get results out

the other end and no humans in the in

the process right and what you've just

described what what Sam probably

extrapolated in his mind there is

effectively that right where where there

are no Brands side marketers there's

only a CFO that says I'm going to spend

12% of my gross profit on marketing and

I'm going to feed it into the marketing

Magic Machine and I'm going to get you

know results out there then now I think

we we are you know there's another

another AI entrepreneur who famously

said you know people

people both underestimate the the the

pace of change but also overestimate the

the the the pace of impact right and to

some degree you know this is the case

with with that that Sam Alon quote

theoretically he is entirely correct it

will take us years to get there

practically because the the process of

marketing and the process of of you know

Commerce and running large brand

organizations is way more messy than

that and you know I also think that it's

not a particularly optimistic or

humanistic view of the world right

because if if that is true of marketing

advertising that's also true of law

Finance Consulting coding right and so

what do we all do you know we sit around

eating you know eggs and pesta all day

um which is you know sounds kind of nice

but it's not going to you know make us

very happy um so so I think I think

where he's wrong is that he's sort of

analytically right and and practically

wrong technically it's possible to build

systems like that they will not be free

I massively disagree with him you know

the the capital expenditure of of

building these systems and the gpus that

support them you know if you went to the

Jensen Jensen Wong you know rocks Rock

coner the other day you will know these

systems are incredibly expensive so

they're going to be monetized and they

will they will be monetized you know um

through customers so they won't be free

but could you theoretically automate a

lot of what What's Done in our industry

today um I think the answer is yes but I

think you will destroy an entire

category of human Ingenuity and

creativity in the process and so you

know I think we have a choice right to

say we either want to build black boxes

that make that get good results but

basically make us all stupid or we can

build good software that makes us a lot

more effective augments our creativity

and makes everyone smarter so it's like

go back to the Middle Ages you know and

the Inquisition is run by you know by

the AI entrepreneur

um or do the Renaissance that's the

choice we have and and what a mic drop

yeah let's let's call it let's let's

call about that that quote um all right

we're gonna take a quick break come back

with news the week around Tik Tok BR

meta paler a lot of exciting news this

week oh yeah and also

cookies this podcast is sponsored by

Adelaide you've probably heard of

adelade by now this month they announced

a new product adelaid for Publishers for

the first time Publishers can leverage

attention me s to curate inventory

deliver attention insights and drive

higher cpms join the weather company NPR

westw one cumulus media and other

leading Publishers taking advantage of

the attention economy's most popular

metric Au maybe one day Mark will

convince me to sell guaranteed Au slot

onm architecture but probably not check

out Adel metrics.com

AFP for more information and we're back

um I'm still reeling from that L quote

to fun so uh so so work with me folks

we've got a lot of uh news this week um

so we'll we'll prioritize on the ones

that are you know probably affecting the

largest markets uh I trillion dollar

companies so first is maybe a surprise

maybe no surprise Google has delayed the

deprecation of cookies for the third

time okay question number one surprise

or no surprise I was on the record

saying I thought they would do it on

time just because of the embarrassment

Factor uh but most other people I know

who are smart were predicting it would

be delayed um and they were right and I

was wrong and it is embarrassing so I

was right about it being embarrassing

what's most embarrassing about this

Google announced cookies going away now

I think we're threeyear anniversary and

the progress has just not been very

impressive I think that it feels to

everyone involved like it's a process

that's out of control that Google

probably shouldn't be running the

process uh because their conflicts of

interest and their antitrust issues uh

that the solutions being offered are are

overly complicated and not necessarily

as private as some simpler Solutions and

people are just frustrated it's been

going on for so long what do you think

happens from here do you think they kick

the can down the road again or is Q2 of

next year which is going to come fast um

is Q2 of next year like actually going

to happen so and and what what happens

what what impacts the industry if if

anything yeah I I think that at this

stage of the process it comes down to

nitty-gritty details in the spec and in

the apis and in the testing and I'm not

really a practitioner anymore so I'm not

doing that but if you talk to people who

are very smart like Paul banister uh who

are working on this they are finding

problems in terms of latency there's

some issue where it only really works

with gam as an ad server which is

obviously a problem antitrust wise uh

and the CMA the UK competition group is

going to look at these very fine grain

detail to see if it gets launched or not

so as a result I would say that whether

launches or not is not so much a

political issue as it's now at the level

of a sort of technical issue Stefan any

take on this yeah I mean look and just

to be precise right it's the dication of

third party cookies right right since we

all adtech nerds here or reformed adtech

nerds not having certainty for the

industry is not great but I think

practically um you know most most large

Brands and and you know agencies have

actually virtually absorbed the the the

impact in the news and have you know

have have many several years ago started

Building Solutions to to understand kind

of you know signal accuracy and so on

you know in different ways and so I

think the market is actually moved

significantly on I mean there are more

implementations of clean rooms and

Federated learning and things like that

now than than there probably still are

largely scale third party you know um

cookie Solutions or you know yeah in

Practical terms um I mean I'll I'll let

aric comment on the the sort of the the

industry perception and everything else

but I mean in Practical terms I don't

think it's it's going to have much of an

impact on us I I think people are

working on it there is a lot of activity

going on to improve the situation and um

and giving people another three to six

months is probably helpful the other

thing was that no one was happy at all

about this idea that the implementation

of the deprecation would be during Q4

for the technical reasons so that was an

issue second announcement and a bigger

announcement Maybe the biggest

announcement R to uh to paraphrase your

Tweet is um is this Tik Tock bill so

Congress passed and Biden signs a bill

that could threaten Tik Tock in the US

bite Dan has until January 19th 2025

again right around the corner basically

uh subject to a 90-day extension if a

deal is in process otherwise uh they

face a ban so this is this is crazy this

is moving forward the company bite dance

uh they plan to sue because they don't

want to sell largely because they don't

want to relinquish the uh the algorithm

and and for you know Tik Tok the

algorithm is is everything so two

questions top of mind for for me outside

of the fact that this is you know pretty

wild that this is actually moving

forward so um number one who is the

potential buyer for this this will

probably be like a hundred billion

dollar price tag and then number two if

there is no buyer and this disappears

and it's a little tough to figure out

exactly how how much the approximately

$15 billion of AD Revenue you where

where does that go so let's talk about

question number one who who is the

potential buyer for Tik Tok in the US I

think the standard buyers meta Google

Etc won't be allowed to buy it uh right

so you have to look at who's big enough

outside that I think it was I think it

was Alex krawitz um who suggested

Walmart was a good potential Target um

Oracle comes up with the conversation

because they're the hosting provider and

but I don't think Oracle has the

consumer chops to do it uh those are

some ideas it's a big chunk Microsoft to

Apple yeah Microsoft would be an

opportunity I mean they have a pretty

diverse set of businesses under their

umbrella right now GitHub and Linkedin

and all these interesting companies

could happen it could be private Equity

maybe I maybe just to make it public

that's a very big private Equity deal I

think the biggest one in the history of

private

Equity yeah yeah Stefan what's your

what's your bet no I I I have no idea I

mean honestly it's a it's an

unprecedented kind of um kind of move

you know I I don't think it's great for

geopolitics I mean I think it's sort of

a it sets us back right and and so from

that perspective I I'm I'm sort of you

know disappointed by the whole the whole

thing going through I mean the more

interesting question for us is is where

does the where does the ad Revenue go

right and I I think if there's a

prolonged period of uncertainty um you

know that that would obviously be be the

worst thing for everyone because then

you know you simply don't know what kind

of longer term commitments to make right

I mean I think I think the you know

whatever you think about the this the

sort of consumer engagement um you know

impact I do think the growth of Tik Tok

has been very good for for media buyers

and for clients you know looking for new

channels um and for more diversity more

competition in the in the space I mean

there's you know apart from from Tik Tok

there there's only a handful of really

large scale Global digital platforms

these days right so it would be a you

know a negative impact on the industry

if if went away as a um as a significant

player I saw a chart in a semaphor uh

this morning that showed that when Tik

Tok was banned in India virtually all of

the revenue went to meta to Instagram

some went to some local providers but it

was pretty much a flip 100% to Instagram

I don't see that being very different in

the US maybe snap could pick up some

yeah so the thought experiment is it's

you know Meta Meta properties um you

know specifically reals on Instagram

it's potentially YouTube shorts which

has a you know similar experience a lot

of Brands just Port them directly to

YouTube shorts and it's and it's snap um

what's interesting about about this you

know from a timing perspective and we

can talk about the next one now um is

this had uh no positive impact on meta

who reported after the Bell last night

so we we'll jump right into into meta So

Meta reported uh 36.467853

year-over year and they beat

expectations they now reach across all

of the the various properties uh the

family I think they they call it

3.24 billion users and other metrics

were were similarly positive and the

stock dropped at one point last night it

was down 19% I don't know what it opened

at um but like I think the largest

potentially one day drop of any company

from a market cap perspective because of

two things um number one soft guidance

uh for the year and then number two

people didn't like the fact that uh you

know Zuck said they were you know

doubling down on AI Investments which is

a little hard to you know kind of wrap

your head around because on the one hand

I understand that you know people felt a

little burned by the um metaverse uh

Investments but then all of these great

results they're posting are largely

because of the Investments they've

already made in AI so what am I missing

here and how did the Tik Tock you know

rumor not help them at all positively

I'm not an expert in public markets but

seems like Madness to me because if you

such short-term memories just two years

ago or even less 18 months ago The Meta

stock was getting hammered because of

overinvestment in in AV um virtual VR

sorry and it was all the same thing it

was like too much capital investment

it's going to take down the stock and

Zuck pulled this amazing turnaround and

the stock is skyrocketed since and all

the metrics are great so why don't give

this guy a little a little slack here

you know I think he knows what he's

doing right yeah I think so they were

getting hammered for overinvestment in

VR and you know uh the impact of uh

Apple at which again through investments

in AI they turned around everything any

cut costs and and and revenue is is

outperforming so and it in contrast we

have Tesla which offers a horrendous

quarter everything bad is happening to

Tesla no one's buying their cars

everyone hates Elon they're getting

kicked out of China he has nothing to

offer the street and then he just says

oh and by the way we're making a cheaper

car and the stock skyrockets he has no

plans to make a cheaper car there is no

cheaper car underway it's not launching

this year he's so full of the guy

is a known liar and he he just somehow

convinced all these people that their

stop should be at a you know multi

hundred PE ratio Madness I love this

it's almost like it's not a podcast um

the um I thing I would say is is kind of

I mean I had a similar response to you

Eric I I was like I was convinced that

of the the back of the no you know news

that um that meta would actually rally

and um what he's spending money on is

basically infrastructure to build open

source AI right and there's something

you know there's going to be a direct

benefit of all that open open source AI

on the The Meta products right the

family of of of products but there's

also a larger you know future

monetization opportunity now maybe the

again like I I'm not an expert in public

markets so you know this is not

investment advice or or financial

analysis but you know certainly it's

strikes me that from a technology

perspective there's a massive

monetization opportunity around the open

source AI that is bringing to Market I

mean Lama 3 is very impressive extremely

capable they've got a fantastic team

there you know there's nothing hindering

them from building a very significant um

you know sort of ancillary Commercial

Business around the the core open source

models so and and ironically you know

the the company that's bulying them you

know all the all the gpus to build this

infrastructure is doing very well off

the the back of this of these

Investments right so not entirely

consist and we're recording this too

early to hear the Google announcement

but it'll be very interesting to hear

because unlike meta Google has

everything to lose for AI and uh their

search business could be in Decline we

don't know yet and it's a much more

interesting story there whereas meta

pretty much is all upside from AI That's

right um I I was playing with the uh The

Meta AI just the the interface um to you

know just you know kind of create images

and everything like that and it's great

it's like fast it's accurate I think

it's on par with some of the other ones

um so uh I think I think you're right

we'll see what happens with uh with

Google one more to uh to close it out

this been been awesome so um paler which

is a company that specializes in uh you

know software platforms for big data

analytics noteworthy because it was um

co-founded by Peter teal um Alex karp

Joe ondale a bunch of guys that have

done some pretty incredible stuff and

you know noteworthy and perhaps you know

somewhat notorious because uh the US

government from a defense perspective is

a big customer of paler there was a a

story about paler now uh selling into

agencies and selling in the paler

Solutions uh as a way to have various

use cases on things like inventory

planning and programmatic sales campaign

optimization it was It was kind of light

in terms of details but interesting in

that you know this company is you know

packaging up some of these data services

to to agencies um and Stefan you have to

comment directly on you know if if paler

has been knocking on the doors of of wtp

but since since we have you on what's

what's your reaction or perspective on

this and if it's a no comment totally

cool no no I I'll I'll definitely can

I'm happy to talk about it no I mean

they they're clearly a capable business

right and um I don't actually know the

details of of who they're selling to and

where they're actually deploying their

Solutions in the um in the wild and and

exactly for what use cases um but I

think the you know certainly my my

strategy has always been that very

strong platform and vendor lock in is

not a good thing in a in a fast moving

you know Innovative space

and so you know we've taken a far more

kind of platform and and you know vendor

agnostic approach in terms of building

our own Solutions we obviously work with

the large hyperscalers we work with all

the large Foundation model companies and

many the the software you know SAS

companies in the space but you know if

you really want to innovate I think you

need a lot more flexibility you know

there's an interesting notion one of my

my product managers has this um riff at

the moment he you know when he when when

he's thinking about um generative AI

based Innovation he says you have to

lean into the emerging capabilities of

the new models what it means is that

you've got to explore the kind of the

full extent of the impact in

applications of these new model

capabilities rather than think about

historical and negative ways of doing

stuff so if you can have a million token

Contex window for a large language model

which Gemini you know 1.5 gives gives

you now you can basically upload two

volumes of war in piece um before you

ask it to to give you an answer right

part of your prompt is two volumes of in

piece now that means effectively large

parts of what we used to do through ETL

and things like that just go away right

you don't need it in fact you don't even

need robotic process automation because

you can just take screenshots of of

screens and have gptv or you know some

version of of you know chin Vision read

the text in the screen and then input it

into the next process right so I mean

slightly diverging but my point is I I

don't think models like that or

platforms like palente are are ideal for

very fast moving very Nimble very agile

kind of um categories and industries

probably great for defense probably

great for security maybe great for for

you know Health Sciences or or kind of

medical practitioners but you know I I'm

I don't think it's ideal in my opinion

for our C that's awesome perspective all

right anything on that no that was I

learned something learned something new

today same year I learned several things

today all right let's call it uh Stefan

thank you so much for being here

wonderful AR it's great to see you guys

again this is a pleasure thank you yeah

you too thank you thank you for

subscribing to marketecture new

interviews are added every week at

marketecture dotv and your favorite

podcasting app

[Music]