00:00:00
Stefano Puntoni: Yeah, absolutely. And
so in the paper, first we start
00:00:03
by sketching this— psychological
threats can emerge from AI
00:00:07
deployment efforts. And like I
said, there are the three broad
00:00:10
categories of threats to
competence, threats to autonomy
00:00:13
and threats to relatedness. And
then in the second part of the
00:00:16
paper, what we're doing is
basically we are starting asking
00:00:20
that question. It's just asking,
saying, "What kind of reactions
00:00:24
can we expect people to engage
in if they feel threat?" And we
00:00:31
sketch, we're building,
basically on literature on
00:00:33
coping, and we argue that five
key reactions will be especially
00:00:41
common, and they vary in the
extent to which they are
00:00:44
positive or adaptive and the
extent to which they are
00:00:46
negative for the organization
and for the employee.
00:00:49
Welcome to <i>The Ripple Effect</i>,
the podcast that takes you on a
00:00:52
journey through the minds of
Wharton faculty. I'm your host,
00:00:55
Dan Loney, and in each episode,
we'll be diving deep into the
00:00:58
inspiration behind the
groundbreaking research that
00:01:01
Wharton professors have
conducted and exploring how
00:01:05
their findings resonate with the
world today.
00:01:07
Well, not only will
00:01:08
generative AI impact the work
that we do on a daily basis,
00:01:12
there is a belief that it could
impact how we think about work,
00:01:17
and could lead to providing
threats towards how we go about
00:01:20
work on a daily basis. This is
the genesis of some research
00:01:25
done by a group of professors,
including our next guest,
00:01:29
Stefano Puntoni, who is a
Professor of Marketing here at
00:01:32
the Wharton School. Stefano,
always great to talk with you.
00:01:35
How are you doing?
00:01:36
Thanks, Dan for having me.
Pleasure to be here.
00:01:38
So this is one
00:01:39
of the many areas that we
continue to see kind of evolving
00:01:44
right now, I guess, into how AI
is impacting so many different
00:01:48
aspects of our lives. In this
case, due to work.
00:01:53
Yeah, so I'm a consumer
researcher, so a lot of my
00:01:56
research is on the user of the
technology, the researcher, the
00:02:00
consumer. But I think one of the
biggest areas of interest for
00:02:06
companies thinking about AI
deployment is about the impact
00:02:09
on employees. Because obviously
this technology is going to be
00:02:12
useful only to the extent that
people are going to use it, and
00:02:14
so understanding adoption
patterns and psychological
00:02:18
reactions to AI tools is going
to be very important to
00:02:20
understand the impact of AI
programs.
00:02:23
So we're talking about this in
the scope of the employee. But
00:02:27
how important is it for the
employer to recognize this as
00:02:30
they're kind of putting a lot of
these processes in place?
00:02:33
Yeah. So if you look at this,
almost every company today has
00:02:37
some kind of AI deployment plan
where basically they are talking
00:02:40
to tech vendors or consultants
or doing in-house. They develop
00:02:44
some kind of tool. It might be
something like a generative AI
00:02:47
engine, like a ChatGPT type
corporate version of it. And
00:02:51
basically the idea will be that
this technology has a promise of
00:02:55
accelerating innovation,
accelerating productivity,
00:02:59
making firms, you know, faster
and better at what they do. But,
00:03:04
you know, this technology is
really only going to have an
00:03:07
impact to the extent that
employees, the people who do the
00:03:11
work in organizations, find a
way of using it, find a way to
00:03:16
integrate effectively these
tools into their workflows,
00:03:19
integrating it with their
competence and expertise. And so
00:03:22
that requires, you know, a lot
of changes to the way we think
00:03:26
about work, and oftentimes
companies are not thinking
00:03:30
enough about the psychological
aspect. And so I advise
00:03:33
companies who— that are
interested in deploying AI at
00:03:38
scale within their organization,
to have almost two parallel
00:03:41
tracks. One is the tech track,
where you're working with your
00:03:44
technology teams and outside
vendors to deploy solutions that
00:03:48
work. And so you have a lot of
concerns about data safety and
00:03:53
compliance and performance and
benchmarks and all of that. But
00:03:58
then at the same time, you also
need to marry that technical
00:04:01
effort with the management and
leadership effort, which is
00:04:05
targeted at employees to
understand and explain, what are
00:04:08
we doing? Why are we doing it?
What's in it for the employee,
00:04:12
if it's going to be actually a
threat to their career and
00:04:16
livelihood, or is it going to be
benefiting them in some way, and
00:04:18
how? And how can you do that
with an authentic voice? So I
00:04:22
think it's important to have
both going at the same time. If
00:04:25
you do only the technical stuff,
but you drop the ball on the
00:04:29
communication and leadership
piece, I think you cannot expect
00:04:32
very good results.
00:04:33
What are some of these threats
that you believe are able to
00:04:38
to come forward here?
00:04:40
In our paper, we basically adopt
a very famous psychological
00:04:44
theory that we find useful to
help organize our thoughts in
00:04:49
this area. And basically we say
that psychological well-being is
00:04:54
really a function of
experiencing feelings of
00:04:57
competence, of autonomy and
of relatedness. These are the
00:05:01
components of this self-
determination theory, you know,
00:05:04
a theory going back to the '80s. So
it's been around for a long
00:05:07
time. But these are three
important antecedents of
00:05:10
psychological well-being. And
then we basically argue
00:05:13
that Gen AI can have important benefits
for both— for all of those. You
00:05:19
know, it can make you feel more
competent when all of a sudden
00:05:22
you're able to do things that
you couldn't do on your own
00:05:24
before. Because Gen AI makes it
possible, for example, to do
00:05:27
advanced analytics using natural
language. It can be empowering.
00:05:32
So it can give you feelings of
autonomy when you realize that
00:05:35
now you can do this. So there is
a sense of being independent and
00:05:39
not being able to rely on
others. And it can help
00:05:42
relatedness when, basically
these chatbots are creating the
00:05:46
seamless parasocial experiences
and can embed themselves into a
00:05:50
team or workflow. So there's
these benefits, but at the same
00:05:53
time, Gen AI can also be a
threat to all of this. Can be a
00:05:56
threat to competence, all these
discussions about jobs, and so
00:06:00
all of a sudden people are
wondering about the value of their
00:06:02
skills. Can be a threat to
autonomy, because now they feel
00:06:05
that they have to adopt these
tools and are no longer in control
00:06:08
of their workflows. They have to
delegate to these AI systems.
00:06:11
And then it can be a threat to
relatedness, when you feel
00:06:14
alienated from your team or from
the company, because you feel
00:06:17
this has been deployed in a way
that is threatening to you.
00:06:20
And seemingly, isn't it then
kind of a fine line between the
00:06:24
two, in terms of the impact that
that a— that an employee could
00:06:28
feel or could see play out?
00:06:31
Yeah. So I think the potential
is enormous for boosting
00:06:35
technological well being,
productivity and performance.
00:06:38
But the reality is that in many
organizations, the conversation
00:06:42
is not really oriented toward
the psychological well-being and
00:06:46
career advancements of the
employees who start to use this
00:06:49
technology. But a lot of the
conversations in business around
00:06:52
Gen AI are about cost cutting,
about productivity increases, to
00:06:56
the detrimental of headcount, and
those conversations are clearly
00:06:59
threatening to people. You cannot
expect people to hear this
00:07:01
stuff and thinking, "Yeah, that's
fine by me." And so it seems like
00:07:06
to me there's a lot of potential
for boosting psychological well-
00:07:10
being of employees, but in
practice, a way that lots of
00:07:13
conversations are going
00:07:14
are pointing in exactly
the opposite direction.
00:07:16
Is there an element—and I know
I've talked to you on different
00:07:19
topics about the generational
differences that are out there,
00:07:23
older people in the workplace
versus how younger people in the
00:07:26
workplace experience everything.
Is there an element of
00:07:29
generational understanding that
you have to have in this mix
00:07:33
here, because how the reaction
of somebody who might be in his
00:07:38
40s or 50s and dealing with AI
and seeing that impact for the
00:07:42
first time may very well be
different from somebody who's in
00:07:45
their 20s or 30s and is much
more digitally savvy.
00:07:48
Yeah, you need to be— one needs to
basically understand the
00:07:52
situation of the person, to be
able to make some predictions as
00:07:55
to what people are going to be
finding psychologically
00:07:58
threatening. Age is an obvious
dimension, function, maybe also
00:08:03
rollout tasks, whatever. There
might be more. With regard to
00:08:06
ages, what's interesting about
it is that, on the one hand, we
00:08:10
know, based on lots of research
on technology adoption, that
00:08:14
younger people tend to be faster
and more keen on new
00:08:16
technologies than older people.
But in this case, there's also a
00:08:20
lot to be worried about in terms
of junior positions, because
00:08:24
what we see is that Gen AI is being
adopted in a way that oftentimes
00:08:27
looks like an AI intern. So a lot
of entry-level positions are
00:08:31
really being now kind of
prioritized as something that
00:08:35
you can do with AI, which
actually might provide a greater
00:08:39
actual threat to the employment
of younger people more than
00:08:42
older people. There is even some
evidence that AI investments are
00:08:46
creating— slowing down career
trajectory for other people
00:08:49
while accelerating those for
older people, already more senior
00:08:52
in the organization. So it's not
obvious which way things are
00:08:55
going to go.
00:08:56
The other thing I
wanted to ask you about is also the
00:08:59
fact that not only is there the
potential age component here,
00:09:02
but you also have to think about
the persona of the individual,
00:09:05
and the fact that, you know,
everybody's persona, in many
00:09:08
cases, is different. So each
person is going to react
00:09:11
differently to a lot of these
components.
00:09:14
Yeah, absolutely. So in the paper,
first we start by sketching
00:09:18
this— psychological threats can
emerge from AI deployment
00:09:22
efforts. And like I said, there
are the three broad categories
00:09:25
of threats to competence,
threats to autonomy and threats
00:09:27
to relatedness. And then in the
second part of the paper, what
00:09:31
we're doing is that basically we
are starting asking that
00:09:34
question. Is just asking, saying,
what kind of reactions can we
00:09:38
expect people to engage in if
they feel threat? And we sketch—
00:09:46
we're building, basically on
literature on coping, and we
00:09:49
argue that five key reactions
will be especially common, and
00:09:56
they vary in the extent to which
they are positive or adaptive,
00:09:59
and the extent to which they
are negative for the
00:10:01
organization and for the
employee. And so we— just to
00:10:04
summarize them, there is one we
call direct resolution. It will be
00:10:07
basically, you feel a threat to,
for example, your competence,
00:10:10
and you decide to upskill
yourself, or you sign up to a
00:10:13
prompt engineering course to be
able to be a proficient user of Gen
00:10:16
AI. That is tackling the threat
directly to solve it and become,
00:10:20
then, you know, a proficient
user and benefit from it. The second
00:10:24
one we talk about is symbolic
self completion, and that
00:10:27
strategy is one where, basically,
the employee is reminding
00:10:31
themselves and others and
underlying the role of human
00:10:34
judgment. For example, you can
imagine a consultant who, in the
00:10:37
course of a presentation,
underlines the human insights
00:10:41
that are brought in. Then there
is dissociation, where basically
00:10:45
what you do is that you are
trying to move away from Gen AI
00:10:50
tools or Gen AI jobs. So for
example, a graphic designer might
00:10:53
rediscover old fashioned
techniques. And as an element of
00:10:57
this, there might be a component
of sabotaging, trying to behave
00:11:00
in a way that makes AI fail, in a
way. And so that's obviously not
00:11:04
good for the company's effort to
benefit from Gen AI. Then you
00:11:08
have escapism, which is
basically disengagement. I'm
00:11:12
now going to have Gen AI doing all
this work and I'm spending all
00:11:15
my time scrolling on the phone.
Clearly not good either. And
00:11:19
then you have one called fluid
compensation, which is trying to
00:11:22
assess, what's AI doing well and
what it's not doing so well, and
00:11:26
then pivot a little bit
recalibrate your activity and
00:11:30
your skills towards the areas
where you feel AI is falling
00:11:33
short. And so that's a more
adaptive one again.
00:11:35
And so that's an interesting
component that we'll touch on now, is
00:11:38
that there is a bit of fluidity
to AI right now in terms of how
00:11:43
it's being implemented, and as
you just alluded to, how we may
00:11:47
see changes occur to better
adapt AI to specific businesses
00:11:53
as we move forward in the
future. So this is still all
00:11:56
very much a process in motion,
isn't it?
00:11:58
Yeah, and the technology is changing
really fast, so it's very
00:12:01
difficult for people to feel
sure footing. In fact, I believe
00:12:06
that one source of threat for
employees is precisely the pace
00:12:10
of change. Where people feel
things are moving so fast I can
00:12:13
never catch up, and everybody's
lagging behind. And you know, I
00:12:16
hear many organizations saying
we are one year behind, but
00:12:18
obviously, if everybody's one
year behind, and— you know,
00:12:20
nobody's behind. But, you know,
it's— it's this feeling of a bit
00:12:24
of FOMO, you know. No— never
knowing what's next and the next
00:12:27
new gadget or whatever. So that
is kind of destabilizing by
00:12:31
definition, almost. But then you
have also the fact that as these
00:12:35
capabilities change, it is
difficult to say, "I should be
00:12:39
investing in this." You know,
like two years ago, everybody
00:12:42
was talking about prompt
engineering. And increasingly,
00:12:44
prompt engineering is kind of
being embedded within these
00:12:47
systems that are getting more
and more sophisticated, for
00:12:50
example, with the reasoning
models, and some of these
00:12:52
principles might not be worth
all that much already. And so to
00:12:56
what extent can we count, for
example, on the technology not
00:13:00
acquiring certain capabilities
that right now seem "safe" from the
00:13:04
point of view of an employee, is
difficult to say. And so I think
00:13:07
there's a lot of uncertainty,
and that uncertainty is actually
00:13:10
part of the problem.
00:13:11
What do you think you and your
colleagues take from this
00:13:14
research that's most important
for both companies and employees
00:13:18
to truly understand about this
moving forward?
00:13:20
Yeah. To me, the bigger picture here is
that this technology is quite
00:13:23
different from other waves,
previous waves of digital
00:13:27
transformation. So if you look
way back, maybe 10 years
00:13:30
ago, maybe to the cloud
computing revolution, what
00:13:33
companies were doing was like
major investments and big risk
00:13:36
taking and saying, "We're
shifting everything away from
00:13:39
our service onto the cloud. We
transform our IT functioning to
00:13:43
basically pay-per-use
service, and we're going to get
00:13:46
our IT needs certified that way."
Now that it was a big change,
00:13:50
and it's a big deal, but if you
think about the user of
00:13:53
computers, imagine working in a
company and writing an email.
00:13:57
Whether the email sits on your
desktop or is up in the cloud,
00:14:00
you write the same email. So it
doesn't necessarily require the
00:14:03
organization to change the way
that people work. It's a
00:14:08
decision that will be made by
the CTO and the CFO, together
00:14:12
with the leadership, and say, you
know, is it good or bad? You
00:14:16
know, go, no go decision. You
pull the trigger and you do it.
00:14:18
And so hopefully it works out.
But then once it's deployed,
00:14:21
then you don't have to teach
anything almost to anybody else.
00:14:24
A little bit, but not much. While
with this technology, because
00:14:29
it's basically— we are using it
to outsource cognitive labor,
00:14:32
this technology is only going to
be productive to the extent that
00:14:36
people find clever ways of
bending it into a workflow. So
00:14:39
integrating AI into a function.
And for that, you need the
00:14:43
people in the function to do the
work of integrating it. And so
00:14:47
it requires now the whole
organization to get on board.
00:14:51
And so it's a much bigger,
harder change management
00:14:54
process. And as an academic
outside business, what I think,
00:14:58
what I'm lamenting a little bit,
is that so many of the
00:15:01
conversations in the media
you know, proclaim
00:15:04
from, you know, proclamation
from CEOs for investor
00:15:09
relations or trying to boost the
share price. You heard Klarna
00:15:12
recently, or Duolingo, or
whatever, that basically they
00:15:16
emphasize headcount reduction.
And of course, you know,
00:15:20
companies operating in a
competitive environment, they
00:15:22
ought to find the efficiencies
that we can find. But if the
00:15:25
only thing we can find around AI
is, what do we do in order to
00:15:29
fire people? I mean, that's not
inspiring to anybody in the
00:15:32
organization. It's only a
threat. And so I think we need
00:15:34
to have a conversation that is
also trying to bring employees
00:15:37
into the picture and say, what's
in it for them? How do you help
00:15:41
use Gen AI to make them
more successful, to capitalize
00:15:45
on their expertise, to, you
know, elevate the status,
00:15:48
accelerate the career, or maybe
even simply give them an
00:15:52
afternoon off, if they can be
more productive. You know, do
00:15:54
something for them.
00:15:56
And I think that
conversation is often missing.
00:15:58
Stefano, always great to talk
with you and get your insight.
00:16:01
Thanks very much.
00:16:02
Thanks, Dan,
great talking to you.
00:16:03
You got it. Stefano Puntoni, Marketing
Professor
00:16:06
here at the Wharton School.
00:16:08
Thank you for listening
to <i>The Ripple Effect</i>. We hope
00:16:10
you found this episode
informative and engaging. Don't
00:16:13
forget to subscribe and leave us
a review so that we can continue
00:16:17
to bring you the best insight
from the Wharton School.
