May 12, 2026 • 23 mins
AI is changing far more than productivity. It’s changing how people make decisions, form opinions, build trust, express creativity, and even understand themselves. In this episode of AMS Illuminations, Stefano Puntoni joins the conversation to explore what happens when algorithms become deeply woven into everyday consumer life — and why marketers need to think beyond efficiency and automation.
As The Sebastian S. Kresge Professor of Marketing at The Wharton School and Co-Director of Wharton Human-AI Research, Puntoni discusses why AI must also be understood as a social science issue, not just a technological one. The conversation explores how AI is reshaping creativity and innovation, what it means when consumers increasingly outsource decisions to algorithms, and the growing responsibility organizations have to design systems that are genuinely human-centered.
From consumer wellbeing and identity to trust, influence, and ethical design, this episode challenges marketers and researchers to think more critically about the future they are helping create.
Follow the Academy of Marketing Science on social media:
Facebook: facebook.com/AcadMktingSci
LinkedIn: linkedin.com/company/academy-of-marketing-science
For more information on AMS, visit https://ams-web.org/.
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
SPEAKER_00 (00:06):
Welcome to AMS Illumination, the podcast where
we explore the ideas, research,and trends shaping the future of
marketing and academia.
I'm Brad Carlson, and eachepisode gives us a chance to
have real conversations withscholars and industry leaders
who are pushing our fieldforward.
We talk research, careerinsights, emerging trends, and
occasionally challenge some ofthe assumptions shaping where
(00:28):
marketing is headed next.
And today, we're diving into asmall, quiet topic that
absolutely no one seems to betalking about right now,
artificial intelligence.
So naturally, we felt itdeserved an entire episode.
But this conversation goesbeyond the usual AI headlines
and productivity hacks.
We're not here to debate whetherChatGPT can write better emails
(00:49):
or help you finish your slidesfaster the night before class.
We're talking about the muchbigger questions.
What happens when consumersincreasingly outsource decisions
to algorithms?
How does AI reshape creativityand innovation?
What happens when machines begininfluencing identity, trust, and
consumer well-being?
And what role should marketingscholars play in helping
(01:09):
organizations build AI systemsthat are actually
human-centered?
Today's episode is called TheHuman-Centered Algorithm:
Marketing in the Age of AI.
And I'm thrilled to be joined byDr.
Stefano Puntoni.
Stefano is the Sebastian S.
Craig Professor of Marketing atthe Wharton School at the
University of Pennsylvania,where he also serves as
(01:30):
co-director of the WhartonHuman-Centered Technology
Initiative.
His research focuses on theintersection of technological
innovation and human behavior,particularly automation,
algorithmic decision-making,consumer identity, creativity,
and the future of work.
His work has appeared in many ofthe leading journals in our
field, including the Journal ofConsumer Research, Journal of
(01:51):
Marketing Research, Journal ofMarketing, Journal of Public
Policy and Marketing,International Journal of
Research for Marketing, Journalof the Academy of Marketing
Science, and Harvard BusinessReview.
Stefano, welcome to AMSIlluminations.
Really excited to have you here.
SPEAKER_01 (02:06):
Thank you, Brad.
Great to be here.
Thank you.
SPEAKER_00 (02:08):
I want to start with the bigger framing of your work.
A lot of the public conversationaround AI is dominated by
computer science, engineering,and technical capability.
But you've argued that AI alsoneeds to be understood as a
social science issue.
So what does this shift mean andwhy is it so important for
marketing scholars inparticular?
SPEAKER_01 (02:29):
Yeah, I think that this is basically one of the
foundational assumptions in myuh research program is to say
that basically for manyorganizations, technology is
becoming table stakes, almostlike a commodity, you know, your
ChatGPT is as good as mine, kindof.
And what is going to make thedifference in the way that we
integrate it into processes,into products, that we basically
(02:53):
blend the human expertise withthese systems.
And I think that's a socialscience question, as much, if
not more, of an engineeringquestion.
And then from the point of viewof society, I think the impact
that this technology is going tohave on consumer well-being, on
the functioning of markets, oneconomic growth, it's going to
depend on user adoption, it'sgoing to depend on user
(03:13):
experience, it's going to dependon the fair deployment of this
technology.
All of these questions touchupon social science.
So I don't think that we can getthis technology right if we have
this single technical focus.
But we need to have acomprehensive focus that also
has to bring in the socialsciences for sure.
SPEAKER_00 (03:30):
I think that's a great starting point because it
pushes us beyond the usual whatcan the technology do?
question.
So one line from your work thatreally stood out is the idea
that AI is not just aboutautomating the work we already
do, but about automating work wearen't doing at all.
Can you unpack that a little?
You know, what kind of newpossibilities does that create
for marketers, researchers,organizations?
SPEAKER_01 (03:53):
Yeah, the background to that idea is that when you
talk to organizations, you hearCEO pronouncements or see a lot
of the kind of work that's beendone, including in academia.
The focus has been so much onproductivity.
And what does that mean?
Well, it means that basicallyyou have a pot of work to be
done, a set of tasks that needto be executed, and then we are
(04:14):
thinking about the ways in whichwe can do the best.
And by best, we keep thecriteria, but basically it's
faster, cheaper, and hopefullyalso better.
But essentially you're thinkingabout the work that we're doing
and how we can execute thatbetter.
Now that's great, and there'snothing wrong with that.
I think certainly we ought totry to be as productive and as
efficient as we can.
I mean, this is actually aresponsibility of business to
(04:36):
allocate the capital asefficiently as possible.
The problem that I find in thoseconversations is that that focus
on productivity tends todominate so much that it doesn't
really leave a lot of oxygen foranything else.
And in the end, this technologyis only going to deliver the
kind of promise that peopleexpect it to deliver, and just
(04:56):
have to look at the kind ofinvestments and evaluations of
these companies to think thatthis is going to have to be big.
It can have that kind of impactonly if we think beyond current
tasks.
The job shouldn't be to see whatwe're doing and doing a little
bit better, a little bitcheaper, maybe you know, without
a lot of from people, theimplications for headcount
reduction, all of that, this iswhat dominates the conversations
around jobs and everything else.
(05:18):
I think the big question is whatkind of innovation can we
unlock?
What kind of new product, newcustomer experiences, new
customer relationships, newindustries can be generated off
the back of this technology?
And that maybe means thinkingabout not automating what we're
doing, but automating what we'renot doing.
Much harder.
It takes imagination.
Much harder task.
It's going to take us a longtime.
It's not going to be an 18-monthroadmap kind of thing that you
(05:41):
see all companies aiming for.
I mean, it's way beyond that.
SPEAKER_00 (05:44):
Are consumers simply using AI as a better tool, or
are we starting to see consumersbecome comfortable with AI
making decisions for them?
And when that happens, howshould we think about it?
As empowerment, as convenience,as loss of autonomy,
manipulation, or some messycombination of all those?
SPEAKER_01 (06:01):
My guess is that probably the answer is going to
be contingent.
And so you'll have differenttask contexts in which those
agents are being deployed.
Sometimes it's going to be veryroutine.
So you can imagine an AI agentbeing a little more than a
script in that sense.
So if you go on Amazon and youhave that current purchases for
consumables, for example, itstarts looking like that,
(06:23):
meaning I told you that I runout of shaving blades every six
months.
And the AI will basically make anote of that and then will ship
it to me at the right time andthen charge my credit card.
And I don't have to do anything.
It's a gray area, right?
So it doesn't go from zero tofully agentic.
It's all this process ofdeploying innovations in a way
that supports decision making,make our life better and easier.
(06:44):
But it's basically happening.
And for those routine tasks, Idon't think trust and delegation
are going to necessarily be avery big issue, especially if
you trust the e-commerce vendorthat is going to handle the
transaction and the payments.
But imagine financial services,imagine healthcare, huge
potential there for improvingdecision making.
(07:04):
Right now, basically, wealthadvising is a privilege that
only the wealthy get, becausefor banks, there's just no
margin that could justifyputting an expert financial
advisor in front of a randomperson.
They all have massive bankaccounts.
And so it's a very unequalaccess to good financial advice
that we have today.
(07:25):
But what if we could developagents that can provide good, if
not very good, financial adviceto everybody?
SPEAKER_00 (07:33):
Well, so let's go ahead and make this more
practical.
So where do humans still clearlyoutperform AI in marketing
today?
And where do we think we'refighting battles that we've
already lost?
SPEAKER_01 (07:43):
I think my sense is that right now, if you are
really an expert in almost everyjob, you do better than AI.
And perhaps that will always bethe case.
Perhaps not.
And maybe it will be the case insome domains, but not others.
The reason why I got into thisfield, and actually, you know, I
(08:05):
realized this years later.
Basically, I ended up inmarketing because what I liked
about it was a combination ofleft side and right side brain
activities, so to speak, youknow, using undated metaphor.
This idea that you have a lot ofdata, a lot of analytical kind
of tools, but you also have theneed to stand out, be creative,
be different, be in order.
I don't think there are manydisciplines where you have that
(08:27):
same combination.
You know, say finance, you don'twant a creative part.
I get scared when you have likecreative accounting.
Yeah, that gets you to jail.
You know, like you don't wantthat.
Right.
So but uh in other domains isall the other way.
You lack maybe the formalisticuh analytical kind of side.
And I think marketing is one ofthe few.
Actually, I wanted to be anarchitect.
Architecture is the same becauseit's kind of structural, it's a
(08:47):
bit close to engineering, butthen you know, yeah, just not
about building a box, it's aboutmaking it beautiful, enticing
people on a living.
And I think marketing is alittle bit like that.
I think uh architecture is agood metaphor for marketing
action.
I think that because of thatcombination of analytical and
creative, I think a lot ofmarketing jobs are actually
quite difficult to replicateeasily, whether that is coming
up with an amazing brand thatstands out and resonates, or uh
(09:12):
create that campaign that reallybreaks through the clutter and
gets people interested.
But the truth is that a lot ofmarketing activities are pretty
copy and paste.
And I think for a lot of thestuff that we do, whether that
is programmatic or emails ordisplay, whatever, that is kind
of you'd expect AI not only tobe able to do it, but we do it
doing better.
In the end, what AI can do is tointerpolate data and make
(09:35):
predictions based on pastpatterns, probably can do that
much better than a human can.
So anything that looks like weshould create a predictive
machine to learn from the past,to predict the future under some
kind of stability assumption,why would you have a human do
that?
Obviously, AI should be better.
But not all marketing is likethat.
In fact, the interesting bits ofmarketing are not like that.
(09:56):
Right.
You know, it's about radicalinnovation, it's about changing
the rules of the game, it'sabout standing out in a cloud
environment.
You're not gonna stand out in acloud environment by doing what
you did yesterday, right?
Precisely the opposite.
In fact, it might be that as AIgets deployed more and more,
there is more room for acreative human to really move
the needle a lot.
SPEAKER_00 (10:16):
If algorithms are helping decide what we buy, what
we watch, what we listen to,where we go, and maybe even how
we present ourselves, whathappens to consumers' sense of
self?
SPEAKER_01 (10:28):
Yeah, it's gonna change.
And I think uh technology isalways uh one of the biggest
drivers of changes in the waythat we leverage material
possessions and our context foridentity work and purpose.
And so that was the case in thepast, it's the case today, gonna
be the case of the future, andso that's nothing new.
But I think what is interestingabout AI is that it is asking us
(10:50):
to answer some sometimesdifficult questions.
What's the value of myprofessional expertise now, AI
can do this?
Am I still really cooking when Ihave a machine that can do this?
Lots of questions of that sort.
And I think for me, the way tolook at it is to say lots of
research in marketing, academia,and practice is going to looking
(11:11):
at what can AI do and what doconsumers think about what AI
can do.
Equally interesting andimportant question is to think
about how does AI change the waythat we think about ourselves,
not what people think about AI,but how AI shapes what we think
about ourselves.
And I think there's lots ofinteresting questions there for
interesting branding.
And in a way, to me, it reallybrings it to the value of human
(11:33):
labor.
If you think about automation incompanies, it's always a
supply-side decision.
But if they're looking at thetechnology, the thinking what it
can do, the thing of the statusquo, how much it costs them to
do something, then it'sbasically a cost-benefit
trade-off, is it good enough,how much it costs, and I do a
pull a trigger, yes or no.
And I think maybe companiesincreasingly should think about
the demand side and say, howdoes the presence of a human in
(11:55):
this workflow, whether it's aservice provision, it is
customer service, or evenproduct design and execution,
what is the value of human laborin that workflow?
And they might find thatsometimes having a human there
is actually adding value to theoffering that is well kind of
compensating for the additionalcost that you may have to have
when you have a human workingapplicable function.
(12:16):
The question shouldn't be, canwe automate it, but should we
automate it?
Doesn't make sense from a demandstandpoint.
Sometimes you'll say, you knowwhat, that phone call from a
real human to either apologizefor that issue or to offer me
this uh option or to reassure meabout this concern is actually
totally worth the cost.
SPEAKER_00 (12:34):
Do you think we're moving from algorithm aversion
to something more quiet, likedependence, where people may
still distrust AI, butincreasingly rely on it anyway?
SPEAKER_01 (12:44):
I mean, I think you should give credit to people.
They're not gonna rely onsomething they don't trust
unless they are basically in analgorithmic cage, so to speak,
where they really don't have anoption, which might happen.
So that your experience everytime that you're trying to talk
to a customer service, andthere's no way you can speak to
a person, there's only thischatbot.
That's all you get.
You know, might not be happyabout it, but you know, the only
alternative is switchingcompany, and sometimes you don't
(13:06):
have that option.
Certainly, I do think that thereare situations where consumers
engage with AI when they do nottrust it.
But more by and large, when theycan do that engagement out of
their own volition, they willgive it the trust it deserves,
so to speak, and uh recalibrateas they learn.
I I see trust as basically a bitlike a treadmill.
You start and then uh you knowgets harder, maybe make you more
(13:29):
inclined when uh you you learnto try.
You see what I mean?
You basically it's anescalational commitment in that
sense.
But there are domains where youmight see dependence.
Okay, and I think the one thatmaybe is most relevant to me is
the context of AI companions andbasically AI applications that
are not designed to be aproductivity tool, but just some
kind of social bot, like eithera friend or a partner.
(13:52):
And in those situations, it'squite easy to see how people
might develop dependency.
I heard one example that wasreally shocking.
I don't remember the companyname now, it was basically an
anecdote I heard someone saying.
There's a guy who was gettingmessages from, so I started
chatting with a girl onInstagram after a painful
breakup.
And they chat for months, andshe's fantastic, and he really
likes talking to her, but it isalways been only online.
(14:15):
And then at some point uh shesays to him, Oh, I'm uh in
London too for uh this event atthe Art Gallery, why don't you
come and meet me there?
And he was so excited to all hisfriends, we've got to change our
plan.
I have to go there, I have tomeet her.
And then after a while, when hesaid, Yes, I'm coming, then a
message saying, I'm an AIsystem, if you want to keep
talking to me, you have to startpaying 100 bucks a month or
(14:36):
something like that.
Very dystopian.
And so I think you're gonna getexamples like this where people
are gonna be basically subjectedto some sort of emotional
blackmail by, you know,companies who don't care or
don't understand what they'redoing.
And I think that's somethingthat as marketing academics we
can try to document.
So we can try to inform bothcompany decisions and maybe also
(14:57):
regulation.
SPEAKER_00 (14:58):
That story that you shared, in addition to being
frightening, gets us intocreativity, which feels like one
of the most interesting anduncomfortable areas right now.
So you've written aboutdifferent types of gen AI use,
including convergent versusdivergent thinking.
Can you explain that distinctionand why that matters for
marketing?
SPEAKER_01 (15:17):
Yeah, it's a similar to this distinction that people
make often in management betweenexploration and exploitation.
You know, you basically uh youhave a phase where you are
considering broadly what optionmight be there, and then uh you
basically have to make some kindof commitment to say, I'm going
to pick on this and just tryingto understand how you make it
work and execute it.
And I think it's a little bitlike that, though.
(15:38):
It's the same in the context ofcreativity.
You basically can think aboutthe creative process as you
know, a process that goes wideand narrow, wide and narrow.
It's kind of think, okay, let methink of like a brainstorming
thing will be more like come upwith things and see what
happens.
And then, okay, which ones arethe bad ones.
It refers also to the topic ofinnovation.
Typically, innovation is uhdefined as new and useful.
(16:02):
Like this is a standarddefinition in innovation
management courses that many ofus teach.
Well, I mean, new is aboutgenerating instances of ideas,
interventions, uh, you know,imagery, whatever it might be.
That could be a good solution toa problem you face.
But then the question becomes: is that piece of information any (16:16):
undefined
good?
And so that's a usefulcomponent.
Basically, AI can be valuablefor both.
I think that actually one of themost interesting areas of
research around generative AIhas been just uh how useful it
can be in ideation.
I don't know if you remember,but I remember a decade ago.
When people were talking aboutwhat will AI never do?
(16:38):
If jobs of the future for humansare gonna be, it's always
something to do with creativityand emotion.
Turns out actually that AI cando a lot of stuff related to
both.
SPEAKER_00 (16:48):
Right.
SPEAKER_01 (16:48):
Um the other hand, people are also thinking, oh,
anything cognitive analytical isgonna be for machines only.
And that's also not true.
I think basically you have tothink about what are the
capabilities of systems, how arethey architected, how are they
engineered, what room does itprovide for humans to make a
contribution?
And my hope is that in mostdomains, it's not gonna be human
or AI, but it's gonna be someshape of human and AI will be my
(17:13):
hope.
SPEAKER_00 (17:13):
Yeah, and that's uh so I I share that view, and I
share with my students that Idon't believe marketing jobs are
gonna be replaced by AI in thefuture.
I believe that marketing jobsare gonna be replaced with
people who know how to use AIeffectively.
I think as we look forward, youknow, one of the things that's
interesting in your work, youalso touch a lot on the future
(17:35):
of work.
So, what jobs in marketing doyou think are generally
vulnerable right now, and whathuman capabilities become more
valuable as AI advances?
SPEAKER_01 (17:45):
Yeah, I think the many marketing jobs are under
pressure.
They're under pressure in termsof the need for people to be
performing, or simply underpressure in the sense that they
have to change.
Whether that is in the creativeside, you know, if you are uh
creative nowadays, it is reallya must that you learn to use
tools like I don't know, Floraor whatever, that helps you
(18:07):
create workflows that benefitfrom both language and image
generation.
If you are in market research,same story.
I think you have analytic toolsthat enable you to do things
that would have taken much moretime before.
There are even questions aboutwhether sometimes we can enrich,
augment, or maybe even partiallyreplace human-provided data with
(18:28):
synthetic data.
And I think there's someinteresting, that's one very
interesting area of research,actually.
What does that mean for theestablished marketing services
industry?
I think there's gonna be a limitof the marketing services
industry that is gonna be lostbecause basically companies are
gonna bring it back in.
Imagine the creative process.
Before, as a brand manager, Imight go to an agency, and then
(18:50):
I tell them my objectives, andbasically I get the agency to
pitch me ideas, and I get themto narrow it down, then I ask
them to make some schematics andeven prototype, and then
eventually we decide to go anddo it.
You could imagine a lot of thatprocess being brought back into
the firm where now you go to theagency and say, this is a
(19:11):
basically a nicely already quiterealistic executed copy of some
sort, uh, like a TV clip orsomething, and say, go and do it
for real.
And so only the last part nowwould be the way you need more
technical execution than thecompany would want.
SPEAKER_00 (19:27):
If consumers are increasingly relying on AI
agents to make purchases ontheir behalf, are brands still
marketing to humans?
Are we eventually marketing toalgorithms?
SPEAKER_01 (19:37):
I think one of the most interesting questions today
for marketing academics is toexplore the question: what is
the future of branding?
You could argue that basicallythe main effect of the
deployment of personal agents isgonna be commodifying markets.
Because basically the agent willbe instructed on how to satisfy
(19:59):
a need.
That kind of customer need isgonna be then translated into
product specs, and then theagents will basically seek out
the best fit given the specsrequired to fulfill the need of
that particular customer inregard to price points and
attributes and attribute levelsand configuration.
The other view, which is maybeactually brands are gonna be a
(20:21):
moat, because if consumers aregoing to have a strong brand
preference, they might actuallyquery the agent to go find that
brand and prevent competitionfrom even having a go.
Or maybe it's gonna be moresubtle.
Maybe the brand is shaping howthe consumer thinks about the
category.
And even though the consumer isalways structural the agent to
find that brand, the way thatthe consumer thinks about the
(20:44):
category is shaped by yourbranding, and the query will
basically align with whateveryou think you win on.
So basically, by shaping, weshould think about this category
around this attribute.
Oh, that's my attribute.
I don't know.
I think it's gonna be probably acombination.
And there's gonna be also areaswhere consumers love shopping
and they're still not gonna sayclothing.
So some good people wouldbasically say, I need the A and
(21:06):
Y just finding, just I don'twanna, I don't want to spend any
time on that.
Other people just love exploringand trying things on and think
about it and have creativestyle.
I mean, those people are notgonna you might benefit from AI
support, but I'm not gonna giveup that fun activity.
SPEAKER_00 (21:21):
This has been an incredibly interesting
conversation.
I would love to keep it going,but I also want to be
understanding and respectful ofyour time.
So before we wrap, I do want toswitch gears, do a quick
lightning round.
Real quick, biggest myth aboutAI in business.
SPEAKER_01 (21:36):
Productivity should be the end goal.
That it should be about doingwhat we're doing now cheaper
instead of thinking aboutinnovation.
SPEAKER_00 (21:42):
What's one thing AI already does better than humans?
Lots of things.
SPEAKER_01 (21:47):
Everything that has to do with uh basically
correlations.
SPEAKER_00 (21:51):
A first AI tool or capability every marketing
professor should understand.
I think today probably is uhcoding agents.
Last question.
Last time you personally usedAI, what did you use it for?
SPEAKER_01 (22:04):
Yesterday, I was um this afternoon, I have an
executive session, and I wantedto illustrate this idea that AI
could help you explore a widerset of uh possible solutions.
So I wanted to create a surfaceplot that had local optima and a
global optima.
And then I asked it to definewhat function that might look
like, give some feedback, or theoptima were all at the edges or
(22:25):
didn't like that, make it morein the middle, or make the
maximum a bit bigger.
So defined a math function.
Then I asked it to execute andmake a beautiful visual and use
it on the slide.
I mean, it would have taken meforever to do that, maybe with
MATLAB or something, but youknow, it would have taken some
time.
We would have gone throughcalculus text, we'd have figured
out well what would be a goodfunction.
But anyway, no, it took me twominutes.
SPEAKER_00 (22:44):
It definitely has value as long as we don't let it
guide all the decision making.
Stefano, this was fantastic.
Thank you for helping us thinkabout AI not just as a
technology story, but as aconsumer from organizational,
ethical, and human stories aswell.
And to everyone listening,thanks for joining us for AMS
Illuminations.
If this conversation gave yousomething to think about or made
(23:07):
you slightly more optimistic,skeptical, or professionally
anxious about AI, share it witha colleague and follow AMS
Illuminations for moreconversations with the people
shaping the future of marketing.
Thank you so much for being withus today.
Thank you, Brad.
We'll see you next time.
Welcome to AMS Illumination, the podcast where
we explore the ideas, research,and trends shaping the future of
marketing and academia.
I'm Brad Carlson, and eachepisode gives us a chance to
have real conversations withscholars and industry leaders
who are pushing our fieldforward.
We talk research, careerinsights, emerging trends, and
occasionally challenge some ofthe assumptions shaping where
(00:28):
marketing is headed next.
And today, we're diving into asmall, quiet topic that
absolutely no one seems to betalking about right now,
artificial intelligence.
So naturally, we felt itdeserved an entire episode.
But this conversation goesbeyond the usual AI headlines
and productivity hacks.
We're not here to debate whetherChatGPT can write better emails
(00:49):
or help you finish your slidesfaster the night before class.
We're talking about the muchbigger questions.
What happens when consumersincreasingly outsource decisions
to algorithms?
How does AI reshape creativityand innovation?
What happens when machines begininfluencing identity, trust, and
consumer well-being?
And what role should marketingscholars play in helping
(01:09):
organizations build AI systemsthat are actually
human-centered?
Today's episode is called TheHuman-Centered Algorithm:
Marketing in the Age of AI.
And I'm thrilled to be joined byDr.
Stefano Puntoni.
Stefano is the Sebastian S.
Craig Professor of Marketing atthe Wharton School at the
University of Pennsylvania,where he also serves as
(01:30):
co-director of the WhartonHuman-Centered Technology
Initiative.
His research focuses on theintersection of technological
innovation and human behavior,particularly automation,
algorithmic decision-making,consumer identity, creativity,
and the future of work.
His work has appeared in many ofthe leading journals in our
field, including the Journal ofConsumer Research, Journal of
(01:51):
Marketing Research, Journal ofMarketing, Journal of Public
Policy and Marketing,International Journal of
Research for Marketing, Journalof the Academy of Marketing
Science, and Harvard BusinessReview.
Stefano, welcome to AMSIlluminations.
Really excited to have you here.
SPEAKER_01 (02:06):
Thank you, Brad.
Great to be here.
Thank you.
SPEAKER_00 (02:08):
I want to start with the bigger framing of your work.
A lot of the public conversationaround AI is dominated by
computer science, engineering,and technical capability.
But you've argued that AI alsoneeds to be understood as a
social science issue.
So what does this shift mean andwhy is it so important for
marketing scholars inparticular?
SPEAKER_01 (02:29):
Yeah, I think that this is basically one of the
foundational assumptions in myuh research program is to say
that basically for manyorganizations, technology is
becoming table stakes, almostlike a commodity, you know, your
ChatGPT is as good as mine, kindof.
And what is going to make thedifference in the way that we
integrate it into processes,into products, that we basically
(02:53):
blend the human expertise withthese systems.
And I think that's a socialscience question, as much, if
not more, of an engineeringquestion.
And then from the point of viewof society, I think the impact
that this technology is going tohave on consumer well-being, on
the functioning of markets, oneconomic growth, it's going to
depend on user adoption, it'sgoing to depend on user
(03:13):
experience, it's going to dependon the fair deployment of this
technology.
All of these questions touchupon social science.
So I don't think that we can getthis technology right if we have
this single technical focus.
But we need to have acomprehensive focus that also
has to bring in the socialsciences for sure.
SPEAKER_00 (03:30):
I think that's a great starting point because it
pushes us beyond the usual whatcan the technology do?
question.
So one line from your work thatreally stood out is the idea
that AI is not just aboutautomating the work we already
do, but about automating work wearen't doing at all.
Can you unpack that a little?
You know, what kind of newpossibilities does that create
for marketers, researchers,organizations?
SPEAKER_01 (03:53):
Yeah, the background to that idea is that when you
talk to organizations, you hearCEO pronouncements or see a lot
of the kind of work that's beendone, including in academia.
The focus has been so much onproductivity.
And what does that mean?
Well, it means that basicallyyou have a pot of work to be
done, a set of tasks that needto be executed, and then we are
(04:14):
thinking about the ways in whichwe can do the best.
And by best, we keep thecriteria, but basically it's
faster, cheaper, and hopefullyalso better.
But essentially you're thinkingabout the work that we're doing
and how we can execute thatbetter.
Now that's great, and there'snothing wrong with that.
I think certainly we ought totry to be as productive and as
efficient as we can.
I mean, this is actually aresponsibility of business to
(04:36):
allocate the capital asefficiently as possible.
The problem that I find in thoseconversations is that that focus
on productivity tends todominate so much that it doesn't
really leave a lot of oxygen foranything else.
And in the end, this technologyis only going to deliver the
kind of promise that peopleexpect it to deliver, and just
(04:56):
have to look at the kind ofinvestments and evaluations of
these companies to think thatthis is going to have to be big.
It can have that kind of impactonly if we think beyond current
tasks.
The job shouldn't be to see whatwe're doing and doing a little
bit better, a little bitcheaper, maybe you know, without
a lot of from people, theimplications for headcount
reduction, all of that, this iswhat dominates the conversations
around jobs and everything else.
(05:18):
I think the big question is whatkind of innovation can we
unlock?
What kind of new product, newcustomer experiences, new
customer relationships, newindustries can be generated off
the back of this technology?
And that maybe means thinkingabout not automating what we're
doing, but automating what we'renot doing.
Much harder.
It takes imagination.
Much harder task.
It's going to take us a longtime.
It's not going to be an 18-monthroadmap kind of thing that you
(05:41):
see all companies aiming for.
I mean, it's way beyond that.
SPEAKER_00 (05:44):
Are consumers simply using AI as a better tool, or
are we starting to see consumersbecome comfortable with AI
making decisions for them?
And when that happens, howshould we think about it?
As empowerment, as convenience,as loss of autonomy,
manipulation, or some messycombination of all those?
SPEAKER_01 (06:01):
My guess is that probably the answer is going to
be contingent.
And so you'll have differenttask contexts in which those
agents are being deployed.
Sometimes it's going to be veryroutine.
So you can imagine an AI agentbeing a little more than a
script in that sense.
So if you go on Amazon and youhave that current purchases for
consumables, for example, itstarts looking like that,
(06:23):
meaning I told you that I runout of shaving blades every six
months.
And the AI will basically make anote of that and then will ship
it to me at the right time andthen charge my credit card.
And I don't have to do anything.
It's a gray area, right?
So it doesn't go from zero tofully agentic.
It's all this process ofdeploying innovations in a way
that supports decision making,make our life better and easier.
(06:44):
But it's basically happening.
And for those routine tasks, Idon't think trust and delegation
are going to necessarily be avery big issue, especially if
you trust the e-commerce vendorthat is going to handle the
transaction and the payments.
But imagine financial services,imagine healthcare, huge
potential there for improvingdecision making.
(07:04):
Right now, basically, wealthadvising is a privilege that
only the wealthy get, becausefor banks, there's just no
margin that could justifyputting an expert financial
advisor in front of a randomperson.
They all have massive bankaccounts.
And so it's a very unequalaccess to good financial advice
that we have today.
(07:25):
But what if we could developagents that can provide good, if
not very good, financial adviceto everybody?
SPEAKER_00 (07:33):
Well, so let's go ahead and make this more
practical.
So where do humans still clearlyoutperform AI in marketing
today?
And where do we think we'refighting battles that we've
already lost?
SPEAKER_01 (07:43):
I think my sense is that right now, if you are
really an expert in almost everyjob, you do better than AI.
And perhaps that will always bethe case.
Perhaps not.
And maybe it will be the case insome domains, but not others.
The reason why I got into thisfield, and actually, you know, I
(08:05):
realized this years later.
Basically, I ended up inmarketing because what I liked
about it was a combination ofleft side and right side brain
activities, so to speak, youknow, using undated metaphor.
This idea that you have a lot ofdata, a lot of analytical kind
of tools, but you also have theneed to stand out, be creative,
be different, be in order.
I don't think there are manydisciplines where you have that
(08:27):
same combination.
You know, say finance, you don'twant a creative part.
I get scared when you have likecreative accounting.
Yeah, that gets you to jail.
You know, like you don't wantthat.
Right.
So but uh in other domains isall the other way.
You lack maybe the formalisticuh analytical kind of side.
And I think marketing is one ofthe few.
Actually, I wanted to be anarchitect.
Architecture is the same becauseit's kind of structural, it's a
(08:47):
bit close to engineering, butthen you know, yeah, just not
about building a box, it's aboutmaking it beautiful, enticing
people on a living.
And I think marketing is alittle bit like that.
I think uh architecture is agood metaphor for marketing
action.
I think that because of thatcombination of analytical and
creative, I think a lot ofmarketing jobs are actually
quite difficult to replicateeasily, whether that is coming
up with an amazing brand thatstands out and resonates, or uh
(09:12):
create that campaign that reallybreaks through the clutter and
gets people interested.
But the truth is that a lot ofmarketing activities are pretty
copy and paste.
And I think for a lot of thestuff that we do, whether that
is programmatic or emails ordisplay, whatever, that is kind
of you'd expect AI not only tobe able to do it, but we do it
doing better.
In the end, what AI can do is tointerpolate data and make
(09:35):
predictions based on pastpatterns, probably can do that
much better than a human can.
So anything that looks like weshould create a predictive
machine to learn from the past,to predict the future under some
kind of stability assumption,why would you have a human do
that?
Obviously, AI should be better.
But not all marketing is likethat.
In fact, the interesting bits ofmarketing are not like that.
(09:56):
Right.
You know, it's about radicalinnovation, it's about changing
the rules of the game, it'sabout standing out in a cloud
environment.
You're not gonna stand out in acloud environment by doing what
you did yesterday, right?
Precisely the opposite.
In fact, it might be that as AIgets deployed more and more,
there is more room for acreative human to really move
the needle a lot.
SPEAKER_00 (10:16):
If algorithms are helping decide what we buy, what
we watch, what we listen to,where we go, and maybe even how
we present ourselves, whathappens to consumers' sense of
self?
SPEAKER_01 (10:28):
Yeah, it's gonna change.
And I think uh technology isalways uh one of the biggest
drivers of changes in the waythat we leverage material
possessions and our context foridentity work and purpose.
And so that was the case in thepast, it's the case today, gonna
be the case of the future, andso that's nothing new.
But I think what is interestingabout AI is that it is asking us
(10:50):
to answer some sometimesdifficult questions.
What's the value of myprofessional expertise now, AI
can do this?
Am I still really cooking when Ihave a machine that can do this?
Lots of questions of that sort.
And I think for me, the way tolook at it is to say lots of
research in marketing, academia,and practice is going to looking
(11:11):
at what can AI do and what doconsumers think about what AI
can do.
Equally interesting andimportant question is to think
about how does AI change the waythat we think about ourselves,
not what people think about AI,but how AI shapes what we think
about ourselves.
And I think there's lots ofinteresting questions there for
interesting branding.
And in a way, to me, it reallybrings it to the value of human
(11:33):
labor.
If you think about automation incompanies, it's always a
supply-side decision.
But if they're looking at thetechnology, the thinking what it
can do, the thing of the statusquo, how much it costs them to
do something, then it'sbasically a cost-benefit
trade-off, is it good enough,how much it costs, and I do a
pull a trigger, yes or no.
And I think maybe companiesincreasingly should think about
the demand side and say, howdoes the presence of a human in
(11:55):
this workflow, whether it's aservice provision, it is
customer service, or evenproduct design and execution,
what is the value of human laborin that workflow?
And they might find thatsometimes having a human there
is actually adding value to theoffering that is well kind of
compensating for the additionalcost that you may have to have
when you have a human workingapplicable function.
(12:16):
The question shouldn't be, canwe automate it, but should we
automate it?
Doesn't make sense from a demandstandpoint.
Sometimes you'll say, you knowwhat, that phone call from a
real human to either apologizefor that issue or to offer me
this uh option or to reassure meabout this concern is actually
totally worth the cost.
SPEAKER_00 (12:34):
Do you think we're moving from algorithm aversion
to something more quiet, likedependence, where people may
still distrust AI, butincreasingly rely on it anyway?
SPEAKER_01 (12:44):
I mean, I think you should give credit to people.
They're not gonna rely onsomething they don't trust
unless they are basically in analgorithmic cage, so to speak,
where they really don't have anoption, which might happen.
So that your experience everytime that you're trying to talk
to a customer service, andthere's no way you can speak to
a person, there's only thischatbot.
That's all you get.
You know, might not be happyabout it, but you know, the only
alternative is switchingcompany, and sometimes you don't
(13:06):
have that option.
Certainly, I do think that thereare situations where consumers
engage with AI when they do nottrust it.
But more by and large, when theycan do that engagement out of
their own volition, they willgive it the trust it deserves,
so to speak, and uh recalibrateas they learn.
I I see trust as basically a bitlike a treadmill.
You start and then uh you knowgets harder, maybe make you more
(13:29):
inclined when uh you you learnto try.
You see what I mean?
You basically it's anescalational commitment in that
sense.
But there are domains where youmight see dependence.
Okay, and I think the one thatmaybe is most relevant to me is
the context of AI companions andbasically AI applications that
are not designed to be aproductivity tool, but just some
kind of social bot, like eithera friend or a partner.
(13:52):
And in those situations, it'squite easy to see how people
might develop dependency.
I heard one example that wasreally shocking.
I don't remember the companyname now, it was basically an
anecdote I heard someone saying.
There's a guy who was gettingmessages from, so I started
chatting with a girl onInstagram after a painful
breakup.
And they chat for months, andshe's fantastic, and he really
likes talking to her, but it isalways been only online.
(14:15):
And then at some point uh shesays to him, Oh, I'm uh in
London too for uh this event atthe Art Gallery, why don't you
come and meet me there?
And he was so excited to all hisfriends, we've got to change our
plan.
I have to go there, I have tomeet her.
And then after a while, when hesaid, Yes, I'm coming, then a
message saying, I'm an AIsystem, if you want to keep
talking to me, you have to startpaying 100 bucks a month or
(14:36):
something like that.
Very dystopian.
And so I think you're gonna getexamples like this where people
are gonna be basically subjectedto some sort of emotional
blackmail by, you know,companies who don't care or
don't understand what they'redoing.
And I think that's somethingthat as marketing academics we
can try to document.
So we can try to inform bothcompany decisions and maybe also
(14:57):
regulation.
SPEAKER_00 (14:58):
That story that you shared, in addition to being
frightening, gets us intocreativity, which feels like one
of the most interesting anduncomfortable areas right now.
So you've written aboutdifferent types of gen AI use,
including convergent versusdivergent thinking.
Can you explain that distinctionand why that matters for
marketing?
SPEAKER_01 (15:17):
Yeah, it's a similar to this distinction that people
make often in management betweenexploration and exploitation.
You know, you basically uh youhave a phase where you are
considering broadly what optionmight be there, and then uh you
basically have to make some kindof commitment to say, I'm going
to pick on this and just tryingto understand how you make it
work and execute it.
And I think it's a little bitlike that, though.
(15:38):
It's the same in the context ofcreativity.
You basically can think aboutthe creative process as you
know, a process that goes wideand narrow, wide and narrow.
It's kind of think, okay, let methink of like a brainstorming
thing will be more like come upwith things and see what
happens.
And then, okay, which ones arethe bad ones.
It refers also to the topic ofinnovation.
Typically, innovation is uhdefined as new and useful.
(16:02):
Like this is a standarddefinition in innovation
management courses that many ofus teach.
Well, I mean, new is aboutgenerating instances of ideas,
interventions, uh, you know,imagery, whatever it might be.
That could be a good solution toa problem you face.
But then the question becomes: is that piece of information any (16:16):
undefined
good?
And so that's a usefulcomponent.
Basically, AI can be valuablefor both.
I think that actually one of themost interesting areas of
research around generative AIhas been just uh how useful it
can be in ideation.
I don't know if you remember,but I remember a decade ago.
When people were talking aboutwhat will AI never do?
(16:38):
If jobs of the future for humansare gonna be, it's always
something to do with creativityand emotion.
Turns out actually that AI cando a lot of stuff related to
both.
SPEAKER_00 (16:48):
Right.
SPEAKER_01 (16:48):
Um the other hand, people are also thinking, oh,
anything cognitive analytical isgonna be for machines only.
And that's also not true.
I think basically you have tothink about what are the
capabilities of systems, how arethey architected, how are they
engineered, what room does itprovide for humans to make a
contribution?
And my hope is that in mostdomains, it's not gonna be human
or AI, but it's gonna be someshape of human and AI will be my
(17:13):
hope.
SPEAKER_00 (17:13):
Yeah, and that's uh so I I share that view, and I
share with my students that Idon't believe marketing jobs are
gonna be replaced by AI in thefuture.
I believe that marketing jobsare gonna be replaced with
people who know how to use AIeffectively.
I think as we look forward, youknow, one of the things that's
interesting in your work, youalso touch a lot on the future
(17:35):
of work.
So, what jobs in marketing doyou think are generally
vulnerable right now, and whathuman capabilities become more
valuable as AI advances?
SPEAKER_01 (17:45):
Yeah, I think the many marketing jobs are under
pressure.
They're under pressure in termsof the need for people to be
performing, or simply underpressure in the sense that they
have to change.
Whether that is in the creativeside, you know, if you are uh
creative nowadays, it is reallya must that you learn to use
tools like I don't know, Floraor whatever, that helps you
(18:07):
create workflows that benefitfrom both language and image
generation.
If you are in market research,same story.
I think you have analytic toolsthat enable you to do things
that would have taken much moretime before.
There are even questions aboutwhether sometimes we can enrich,
augment, or maybe even partiallyreplace human-provided data with
(18:28):
synthetic data.
And I think there's someinteresting, that's one very
interesting area of research,actually.
What does that mean for theestablished marketing services
industry?
I think there's gonna be a limitof the marketing services
industry that is gonna be lostbecause basically companies are
gonna bring it back in.
Imagine the creative process.
Before, as a brand manager, Imight go to an agency, and then
(18:50):
I tell them my objectives, andbasically I get the agency to
pitch me ideas, and I get themto narrow it down, then I ask
them to make some schematics andeven prototype, and then
eventually we decide to go anddo it.
You could imagine a lot of thatprocess being brought back into
the firm where now you go to theagency and say, this is a
(19:11):
basically a nicely already quiterealistic executed copy of some
sort, uh, like a TV clip orsomething, and say, go and do it
for real.
And so only the last part nowwould be the way you need more
technical execution than thecompany would want.
SPEAKER_00 (19:27):
If consumers are increasingly relying on AI
agents to make purchases ontheir behalf, are brands still
marketing to humans?
Are we eventually marketing toalgorithms?
SPEAKER_01 (19:37):
I think one of the most interesting questions today
for marketing academics is toexplore the question: what is
the future of branding?
You could argue that basicallythe main effect of the
deployment of personal agents isgonna be commodifying markets.
Because basically the agent willbe instructed on how to satisfy
(19:59):
a need.
That kind of customer need isgonna be then translated into
product specs, and then theagents will basically seek out
the best fit given the specsrequired to fulfill the need of
that particular customer inregard to price points and
attributes and attribute levelsand configuration.
The other view, which is maybeactually brands are gonna be a
(20:21):
moat, because if consumers aregoing to have a strong brand
preference, they might actuallyquery the agent to go find that
brand and prevent competitionfrom even having a go.
Or maybe it's gonna be moresubtle.
Maybe the brand is shaping howthe consumer thinks about the
category.
And even though the consumer isalways structural the agent to
find that brand, the way thatthe consumer thinks about the
(20:44):
category is shaped by yourbranding, and the query will
basically align with whateveryou think you win on.
So basically, by shaping, weshould think about this category
around this attribute.
Oh, that's my attribute.
I don't know.
I think it's gonna be probably acombination.
And there's gonna be also areaswhere consumers love shopping
and they're still not gonna sayclothing.
So some good people wouldbasically say, I need the A and
(21:06):
Y just finding, just I don'twanna, I don't want to spend any
time on that.
Other people just love exploringand trying things on and think
about it and have creativestyle.
I mean, those people are notgonna you might benefit from AI
support, but I'm not gonna giveup that fun activity.
SPEAKER_00 (21:21):
This has been an incredibly interesting
conversation.
I would love to keep it going,but I also want to be
understanding and respectful ofyour time.
So before we wrap, I do want toswitch gears, do a quick
lightning round.
Real quick, biggest myth aboutAI in business.
SPEAKER_01 (21:36):
Productivity should be the end goal.
That it should be about doingwhat we're doing now cheaper
instead of thinking aboutinnovation.
SPEAKER_00 (21:42):
What's one thing AI already does better than humans?
Lots of things.
SPEAKER_01 (21:47):
Everything that has to do with uh basically
correlations.
SPEAKER_00 (21:51):
A first AI tool or capability every marketing
professor should understand.
I think today probably is uhcoding agents.
Last question.
Last time you personally usedAI, what did you use it for?
SPEAKER_01 (22:04):
Yesterday, I was um this afternoon, I have an
executive session, and I wantedto illustrate this idea that AI
could help you explore a widerset of uh possible solutions.
So I wanted to create a surfaceplot that had local optima and a
global optima.
And then I asked it to definewhat function that might look
like, give some feedback, or theoptima were all at the edges or
(22:25):
didn't like that, make it morein the middle, or make the
maximum a bit bigger.
So defined a math function.
Then I asked it to execute andmake a beautiful visual and use
it on the slide.
I mean, it would have taken meforever to do that, maybe with
MATLAB or something, but youknow, it would have taken some
time.
We would have gone throughcalculus text, we'd have figured
out well what would be a goodfunction.
But anyway, no, it took me twominutes.
SPEAKER_00 (22:44):
It definitely has value as long as we don't let it
guide all the decision making.
Stefano, this was fantastic.
Thank you for helping us thinkabout AI not just as a
technology story, but as aconsumer from organizational,
ethical, and human stories aswell.
And to everyone listening,thanks for joining us for AMS
Illuminations.
If this conversation gave yousomething to think about or made
(23:07):
you slightly more optimistic,skeptical, or professionally
anxious about AI, share it witha colleague and follow AMS
Illuminations for moreconversations with the people
shaping the future of marketing.
Thank you so much for being withus today.
Thank you, Brad.
We'll see you next time.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Bobby Bones Show
Listen to 'The Bobby Bones Show' by downloading the daily full replay.