From Prediction to Creation: Rethinking Marketing in the Age of Generative AI

Generative AI is changing marketing fast — but what does that really mean for strategy, creativity, ethics and consumer trust? In this episode of AMS Illuminations, host Brad Carlson talks with Dr. Erik Hermann, Interim Professor of Marketing at European University Viadrina, about the shift from predictive AI to generative and agentic AI.

Together, they explore AI as both collaborator and competitor, why consumers form emotional attachments to machines, and how marketers can use AI responsibly to create value for customers, organizations and society.

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/.

Episode Transcript

SPEAKER_01 (00:07):
Welcome to AMS Illuminations, the podcast where
we explore the ideas, research,and trends shaping the future of
marketing in academia.
I'm your host, Brad Carlson, andeach episode gives us a chance
to have real conversations withscholars and industry leaders
who are helping redefine whereour field is headed.
Today we're talking about atechnology that seems to be

(00:28):
evolving faster than our abilityto create conference tracks,
special issues, LinkedIn thinkpieces, and emergency syllabus
updates about it, generative AI.
More specifically, we'reexploring what happens when AI
moves beyond prediction andoptimization and starts
creating, interacting,persuading, and maybe even

(00:49):
behaving a little more like aconsumer or a marketer itself.
Because the conversation aroundAI has shifted fast, we're no
longer just talking aboutsystems that analyze data or
automate repetitive tasks.
We're talking about AIgenerating campaigns,
interacting socially, actingautonomously, and increasingly

(01:09):
participating directly in valuecreation.
This raises some fascinatingquestions.
What happens when AI becomes acollaborator in marketing
strategy and creativity?
Why do consumers form emotionalattachments to machines?
Could consumers eventuallyprefer AI interactions in some
situations?
And how do we balanceinnovation, personalization,

(01:31):
ethics, autonomy, and trust in aworld increasingly shaped by
intelligent systems?
Today's episode is called FromPrediction to Creation:
Rethinking Marketing in the Ageof Generative AI.
And I'm really excited to bejoined by Dr.
Eric Herman.
Thanks for being here with us,Eric.

SPEAKER_00 (01:49):
Thanks a lot for having me.
A pleasure to be here.

SPEAKER_01 (01:51):
Well, for those who don't know, Eric is interim
professor of marketing at theEuropean University Via Drena in
Frankfurt.
He previously served aspermanent affiliate professor of
marketing at ESCP BusinessSchool and has also worked
extensively in industry,including roles involving
digital solutions, marketingstrategy, investor relations,

(02:13):
and advertising policy.
He currently serves as socialmedia editor for the Journal of
Marketing, and his researchfocuses on generative AI, AI
agents in marketing, AI ethics,consumer psychology, and
vulnerable consumers.
His work has appeared in leadingjournals, including Journal of
Marketing, Journal of theAcademy of Marketing Science,

(02:34):
International Journal ofResearch and Marketing, Journal
of Business Ethics, Journal ofPublic Policy and Marketing, and
Harvard Business Review.
Again, thank you for being herewith us today.
Really glad to have you.
I want to start with your pathinto this space for a minute.
Your work sits at this reallyinteresting intersection of AI,
ethics, consumer psychology, andsocial good.

(02:56):
How did you end up working inthis area?

SPEAKER_00 (02:58):
Yeah, that's a little bit of an unusual story,
as you mentioned.
I worked in industry since I hadto leave academia in 2019, and
one role was project manager forSME digitalization trading, and
I was focusing on informationand cybersecurity.
And I found an intriguingcomment on trust in AI and

(03:19):
cybersecurity as a double-edgedthought.
And one of the authors wasLuciano Floridi back then, a
professor in Oxford, now he's inYale.
And he had also several otherintriguing papers on this AI for
social good perspective, wherethey lay out five ethical
principles which should befollowed.

(03:41):
And then I was curious andchecked the marketing literature
and realized that, of course, insome papers, some principles
were addressed, but notcomprehensively.
And I decided to write a paperon it, which was eventually
published in Journal of BusinessEthics, quite well received.
That's how this whole researchpath started.

(04:04):
Then I worked on more conceptualpapers on AI and marketing and
always took the social ethicalperspective on it.

SPEAKER_01 (04:12):
Very interesting backstory.
Now, before we dive deeper intoAI, let's start with some
terminology because AIresearchers seem to be committed
to inventing new categoriesfaster than the rest of us can
define them.
We hear predictive AI,generative AI, now agenic AI,
and we hear these things all thetime.
So for listeners trying to makesense of all this, what actually

(04:33):
matters in those distinctions?

SPEAKER_00 (04:35):
Predictive AI, as this term implies, is about
analyzing past data to predictcertain outcomes like the next
purchase.
It can be used for pricing,traditional examples,
recommendation systems likecollaborative filtering systems,
like we know from this quitewell-known big e-retailer or

(04:56):
e-commerce firm.
We have generative AI, which isable to create seemingly, and
that is quite important,seemingly novel content like
text, video, audio, code, andquite interestingly also data,
synthetic data.
I will come to that a little bitlater.

(05:17):
And then agentic AI or AIagents, which are advanced AI
systems, which are able toperceive, act, and reason on
behalf of so-called principles,which could be firms or
customers.
And that is currently discussedas maybe the future, agentic

(05:38):
commerce, and so on and soforth.

SPEAKER_01 (05:40):
I do note that you really emphasize seemingly
novel, and I'm sure we'll unpackthat a little bit.
But before we unpack that, doyou think marketers fully grasp
how significant this shiftreally is from predictive to
agenic AI?
Or are we still mostly treatinggen AI like a productivity tool
instead of something thatfundamentally changes how

(06:01):
marketing works?

SPEAKER_00 (06:02):
Depends a little bit of whom you ask.
There are so many reports outthere, uh, asking CEOs, asking
leaders.
Interestingly, a lot of them seethe potential, like there's this
Gardner report, where 65% ofCMOs say that it will
dramatically change their roleas the CMO chief marketing

(06:24):
officer, but only 32% say thatit also requires significant
changes to their own skills.
So there's a certain AI blindspot, or mostly the leaders
acknowledge the importance ofAI, Gen AI, but they also
express the feeling that theyare not really prepared for

(06:44):
that.
And some of them also say thatthey do not see the return on
investment now.
And what I also found uh veryinteresting was a report by PWC.
Their main finding was that uh74% of AI's economic impact is

(07:06):
just captured by 20% of theorganizations.
And these are basically largeorganizations.
So awareness is there, butimplementation seems to be still
an issue.

SPEAKER_01 (07:20):
That response naturally raises the anxiety
question.
Are we moving toward AI as acollaborator or as a competitor?
And where do you think the realpressure points are for
marketers right now?

SPEAKER_00 (07:34):
I think it could be both.
And this is actually related tothis debate about human
enhancement versus humanreplacement.
So it can be a competitor inregard to really basic
repetitive routine tasks wherehumans can be replaced, like
advertising copy or creatingvisuals.

(07:55):
It could be also a greatcollaborator for content
production or creativity taskswhere we have already studies
showing Gen AI can create quitegood ideas, but when it comes to
implementation, humans stillexcel.
For basic routine tasks, theremight be the risk of human
replacement and competition,whereas for higher level tasks,

(08:19):
higher order tasks, there couldbe the potential for Gen AI to
become a great collaborator.

SPEAKER_01 (08:26):
Do you think there's anything specific about those
higher order things?
Is it a creativity link?
Is there something that you seeas a clear advantage that we as
humans have over AI right now?

SPEAKER_00 (08:38):
Yeah, I would say so, definitely.
And it's also a criticalthinking and so on.
There's in the creativityliterature, there's this
interesting aspect of convergentthinking and divergent thinking.
So convergent thinking is adomain specific, a quite narrow
task where Gen AI could be quitegood since it's trained of tons

(08:58):
of data.
And divergent thinking is adomain general, which is at the
core of really creative tasks,where you need critical thinking
abilities to really judge anovelty of an idea.
And that is something where Ithink that humans still excel
and will excel.
We heard the term AI slot thatis getting more average and

(09:23):
average.
There are also uh discussionsand already opinion pieces that
Gen AI will homogenize thecontent, our attitudes, and so
on.
So if you take this intoaccount, then for those what I
called higher order orhigher-level tasks, um, humans
will excel.

SPEAKER_01 (09:43):
From a personal standpoint, this intrigues me.
I just wonder what your ownthought is.
Do you think creativeindividuals and critical
thinkers who use AI frequently,do you think it enhances their
creative abilities?
Or do you think it ispotentially reducing overall

(10:03):
creativity for those people?

SPEAKER_00 (10:05):
I think it really depends on how you use it.
And that is also a heateddebate, how the impact on
learning and so on, criticalthinking is yeah, if you start
using it frequently for allkinds of tasks and do not
reflect about the output whichis provided, and we do not use
your own brain to say it reallyin simple terms, then I think

(10:28):
there's more backlash thanbenefit.

SPEAKER_01 (10:31):
Another really interesting part of your work
involves anthropomorphized AI,if I said that correctly.

SPEAKER_00 (10:39):
So I do not know, I hate this word.

SPEAKER_01 (10:44):
It looks great in text.
Basically, it's consumersforming emotional attachments to
machines.
Why do you think this happens soeasily?

SPEAKER_00 (10:53):
We have this interesting phenomenon even
before Gen AI and so on, ofparasocial interaction, this
feeling that you are connectedand have a reciprocal
relationship between a mediatepersona, which was celebrities,
but also social media, people onsocial media, and now we have it
for Gen AI since those LLMchatbots, for instance, are

(11:16):
increasingly humanized, andthere are all kinds of apps, AI
companions.
They try to emulate those humancapabilities in regard to uh
conversation, empathy,personality, and therefore over
time, users, consumers might getthe feeling that they are
talking to real human.

(11:37):
Basically, humans are wired toanthropomorphize.
So we as humans naturallyascribe or try to ascribe human
traits to non-human entities,not just technology, but also to
pets, for instance, and somepeople even to objects.

SPEAKER_01 (11:55):
Do you think that we're entering into a world
where consumers may actuallyprefer AI interactions in
certain situations?
And if so, what kind ofsituations are those?

SPEAKER_00 (12:06):
We are now in a situation that we have enough
evidence to draw conclusionacross several years of
research, which is calledmeta-analysis for the people
which are interested to get moreinsights.
Basically, we already had thisphenomenon of AI aversion, that
people prefer human providers orhumans in marketing context over

(12:29):
AI.
One of those meta analyses hasshown that this AI aversion
became less and less over theyears and even turned into AI
appreciation.
And there was one point in time,and there was the release of
ChatGPT, where this reallychanged this picture.

(12:49):
So people are now accepting AIto a much higher extent than
some years ago.
Another meta analysis showedthat uh even when consumers are
skeptical, they still choose andbuy from AI and rely on AI
similarly to humans.

(13:10):
And this preference is, however,context-dependent.
So for certain tasks which arevery objective and more
utilitarian, not the hedonicpleasure context, they are more
willing to rely on AI.
And interestingly, also in anembarrassing consumption and

(13:31):
purchase context.
They are also more willing torely on AI than humans.

SPEAKER_01 (13:37):
Let's move into ethics for a minute.
So, where do you see the biggestethical blind spots right now
and how companies are deployingAI?

SPEAKER_00 (13:44):
Yeah, that brings me back to one of the points we
mentioned before.
Uh, this is uh over-reliance andso-called miscalibrated trust,
meaning that you trust AI,although the capabilities uh
should not justify this trust.
There are already studiesshowing that people using AI to
combat loneliness become evenmore lonely when using it, since

(14:08):
they're just interacting withGen AI and therefore do not
interact with humans.
So some studies have shown thereare short-term positive effects,
but in the long run, there couldbe huge risks.
I think there's also this riskof manipulation.
There was quite well-receivedand debated study in science
showing that AI conversation canreduce conspiracy beliefs.

(14:32):
And now when you think aboutconspiracy beliefs, which are
really stable and conversationwith Gen AI can change it, then
think what Gen AI conversationscould change else.
I already mentioned sickerfancy.
There's also research showingthat it leads to a certain
confirmation bias andoverconfidence of users, even to

(14:54):
more extreme attitudes.
And the last one we have notmentioned yet is uh
sustainability.
There's Sasha Lucioni andcompany Hugging Face, they're
doing great work to bring inmore transparency here.
But of course, on the otherhand, the big tech and AI
providers try to hide the realamount of emissions,

(15:17):
environmental impacts.
Also, Google no figures abouttheir data centers, about the
emissions, and so on and soforth.

SPEAKER_01 (15:26):
So when when companies talk about ethical AI,
do you think that's becoming areal operational priority, or is
it still mostly aspirationallanguage for a lot of
organizations?

SPEAKER_00 (15:38):
Yeah, it depends a little bit uh how
comprehensively they reallyaccount for the different
ethical principles which are outthere.
Just take privacy.
That's such a huge issue.
Just to take privacy seriouslyis actually quite a big endeavor
then to guarantee autonomy, toguarantee justice and fairness,

(16:02):
so that really every consumer istreated equally, that every
consumer has equal access.
I think that's a huge task whichneeds uh self-regulation by AI
providers and also regulation.
But regulation, of course, giventhis task developments, lags
behind.

SPEAKER_01 (16:23):
So if I'm in an organization and I'm trying to
figure out how to navigate thisethical AI framework, it just so
happens that you've proposed theassurance framework as a way to
think more systematically aboutthis.
Can you walk us through thatframework in a practical way and
explain how marketers canactually apply it?

SPEAKER_00 (16:42):
Yeah, sure.
Assurance, by the way, I likeacronyms since therefore you
will find some acronyms in mypaper.
But since AI research isexploding, there's so many
frameworms using acronyms, sothat is not a Unix selling
reposition anymore.
But let's get back to assurance.
Uh the first one is autonomy.

(17:05):
So you should always keep humansin the loop on the side of the
firm, of course, but alsoregarding consumers, they should
have the possibility tointervene.
So both firms and consumers,particularly in high-stakes
contexts, just think of usingAI, Gen AI for financial advice.

(17:26):
Then you have security.
Yeah, a lot of personal privatedata is involved, so you have to
ensure the data protection thatAI is not misused or hacked,
something like that.
Then the second S and the U isabout sustainability.
So keep an eye on AI'sfootprint, and yeah, let's put

(17:47):
it simple, whether it's worththe environmental impact.
Then representativeness relatesto the input data, so that they
should be representative of thepopulation you are addressing.
There's this issue of theso-called weird samples that a
lot of LLMs have been trainedon, so-called weird samples and

(18:09):
underlying data, meaningWestern, educated,
industrialized, rich, anddemocratized.
So that's a bias towards uhWestern and industrialized
countries.
Then the A is aboutaccountability.
So who is held accountable ifsomething goes wrong?
That is, of course, an importantquestion.

(18:31):
Is it the AI provider?
Is it the firm?
Could it be the customer?
Here you really have to trace,audit, and take responsibility
for AI decisions.
Then non-biasedness andnon-discrimination is not about
the inputs but the outputs.
So, how are people treated byAI?
Are the recommendations andpredictions fair?

(18:55):
Then uh crediting, which wasreally big debate when
everything started, since LLMshave been trained on all kinds
of documents, also on art, onnewspapers, there are a lot of
lawsuits now going on.
Um, you should actually creditauthorship when you use AI.

(19:17):
Imagine as a firm you create alot of visuals which use the
work of uh artists.
Actually, you should credit the,although even it's one data
point, but still.
And the last one is empowerment.
So you really have to enable allstakeholders within the firm,
but also consumers to be able tounderstand AI and to use it in a

(19:42):
responsible way, which is oftenrelated to concepts like AI
literacy.
So you have to train people howto use Gen AI responsibly and
also in the future AI ages.

SPEAKER_01 (19:54):
Before we wrap up, I want to switch gears and do a
quick lightning round.
So here's how this works.
I'm gonna throw out a handful ofkind of rapid-fire questions,
and you've got about 60 secondsfor each answer.
So it's long enough to saysomething insightful, but short
enough to prevent either of usfrom accidentally turning this
into a conference symposium.
Here we go.
First one.

(20:14):
How much of the current Gen AIbuzz is real utility versus pure
fear of missing out?

SPEAKER_00 (20:23):
I would say it's both.
It's real utility at thespecific task level, like uh
content creation, analytics, andso on.
But at this overarchingstrategic level, it might be
more FOMO of so firms areexperimenting, but they are not
really redesigning theirworkflows to really implement it

(20:44):
effectively and efficiently.

SPEAKER_01 (20:46):
All right, number two, one AI hype trend you
completely don't buy into.

SPEAKER_00 (20:53):
That AI in the future will replace human
creativity.
So really original vision,taste, and so on, cultural
walls.
I do not think that AI ispossible to do that.

SPEAKER_01 (21:09):
One comment uh about that issue.
I don't know if you've listened.
There's some AI tools that takeexisting songs and recreate
them.
So for instance, there's somethat take rap songs and then
they recreate them in like a 50sblues.
I don't know if you've seen orheard any of these, but some of

(21:29):
these are incredible.
They're incredible because theoriginal songs are great, but
the voices, the instrumentation,it's a little scary how good
some of this is.

SPEAKER_00 (21:39):
Yeah, that's true.
But that is again then thisquestion of creativity.
What is creativity?
Is it creative to more or lessrecombine and put it in another
context?
But yes, for some things they'reincredibly good.

SPEAKER_01 (21:54):
Yeah.

Okay, number three (21:55):
one human trait AI will never truly
replicate in a marketingcontext.

SPEAKER_00 (22:02):
I think that's really about genuine empathy,
which is really grounded inlived experiences.
AI can simulate it, but it can'tfeel, and you have to feel in
order to show real empathytowards others.

SPEAKER_01 (22:20):
All right, number four.
Ten years from now, will westill call it AI marketing or
will it simply be calledmarketing?

SPEAKER_00 (22:28):
I would say it's a little bit like digital.
So digital is now the default.
Sometimes we still use this termdigital marketing, although a
lot of stuff related tomarketing or almost everything
is digital.
I I would say the same couldhappen for AI.

SPEAKER_01 (22:47):
All right, last one.
One underappreciated opportunityin AI and marketing.

SPEAKER_00 (22:54):
To come back to the introduction, uh, doing good by
using AI.
So, like this AMA definition,providing value not just for
firms, but also for society atlarge and of course customers.

SPEAKER_01 (23:10):
Yeah, that seems to be one area that we continually
need more effort doing good.
And it seems like AI gives anamplified opportunity to stray
from that path or move closer tothat path.

SPEAKER_00 (23:25):
Yeah, actually, one problem, well, not problem, but
it's often neglected.
When we think about AI, we thinkabout commercial interests and
context and firms, but you couldalso leverage for charities,
NGOs to optimize the campaignsand so on.
And that is also doing good withAI.

SPEAKER_01 (23:46):
Absolutely.
Eric, this was fantastic.
I really appreciate how grandthis conversation was, not just
in the technology itself, but inthe human side of marketing,
ethics, trust, vulnerability,and empowerment.
And to everyone listening,thanks for joining us for
another episode of AMSIlluminations.
If this conversation sparked anew idea, challenged your

(24:08):
assumptions, or made youslightly rethink your
relationship with AI, share itwith a colleague and follow AMS
Illuminations for moreconversations with people
shaping the future of marketing.
Thanks so much for being with ustoday, Eric.

SPEAKER_00 (24:20):
Thank you.
Thanks for having me.
What's great?

SPEAKER_01 (24:23):
All right, we'll see everyone next time.