Untertitel (81)
0:00I'm Martin Gonzalez. I'm a principal of
organization and leadership development at
0:04Google and I'm the author of The Bonfire Moment.
The book explores this idea that teams are harder
0:10than tech. In the process of innovation, it's
so important for leaders and CEOs and founders
0:18to pay attention to the people's side of the
business because that could easily derail your
0:23best laidout plans. We know a lot of employees
and organizations are starting to use AI for
0:30their work. We also know that we flip-flop between
these really intense narratives of substitution.
0:37Our jobs are going to go away. Um my role will
get replaced. Um there will be less of people
0:43playing my kind of work, my kind of role because
of AI. And a narrative of augmentation which is
0:50these tools give me superpowers that allow me to
do more within my role. and and I will succeed and
0:56do well in the future if I can only adapt these
new um technologies. There's a lot we need to
1:02think about when we think about the augmentation
model because as early research is showing as we
1:09bring these tools into the workplace, we're
not quite seeing the kind of transformative
1:14potential that AI has been talked about by
by its inventors. So I've started to think
1:20about three puzzles we need to solve for as we
bring these technologies into our organizations.
1:31One of the challenges in bringing AI into an
organization is what I've started to call the
1:37selective upgrade puzzle. This is when these
tools endow its users with superpowers but
1:44not all users and somehow there's a selective
upgrade that happens um when these tools get
1:50shared in an organization. This one randomized
control experiment that was run by researchers
1:56from places like Harvard and MIT engaged the
Boston consulting group and set up their junior
2:04consultants in control groups and in a couple
of experimental groups. What they did was they
2:09gave them access to a um to a large language model
and they were asked to do two kinds of tasks. The
2:17first task was a creative ideation task. They had
to help a fictitious client come up with different
2:24product ideas that they can go to market with.
The second was a business analytics task where
2:29they had to analyze why a business was struggling
and create recommendations. What the study went
2:36on to discover was that when they looked at the
top performers, they tended to do much better
2:41and when they looked at the lower performers,
they tended to do much worse. When you think
2:45of this selective upgrade effect um spread across
thousands of employees over a span of time, what
2:52we might see is an ever growing gap between your
best and worst performers. And this variability
2:59would have been attributed to the use of these
AI tools where that where that gap didn't exist,
3:04you know, before you deployed these tools. There
are a couple of things that leaders can think
3:08about as they deploy AI in their organization.
And the first is to create really clear guard
3:14rails around what these tools should be used for
and what they shouldn't be. And those guardrails
3:20will possibly diminish over time as these tools
become much more um much more effective. But it's
3:27important to go through this experimental period
understanding you know where it actually augments
3:33the work and where it actually takes away from the
work. Another thing to consider is it's important
3:39for the users as they leverage these tools for
certain domains that they have a certain basic
3:46level of of expertise in these domains. It allows
the users to apply good judgment when a tool is
3:54actually leading them in the in a worse direction
um and when it actually is augmenting the work.
4:00Using a tool when you have zero knowledge of
that domain is a very very dangerous proposition.
4:14As we think about bringing AI into our
organizations, we need to think about this
4:18agentic preference puzzle. We as humans have a
tendency towards control. And when these tools
4:26take away control from the work, we see that
adoption rates drop. There are some fascinating
4:33studies done out of Wharton that explore this idea
that they called um algorithmic aversion bias. For
4:39example, when was the last time you decided to
override what Google Maps or Ways told you was
4:46the right way home? We'll sometimes believe
that we actually have better or have a lower
4:51error rate than these machines. And what
this branch of study had had looked into
4:56was when individuals actually purs or see a a an
algorithm commit an error, even if that error rate
5:05is still lower than the human error rate, we
we would much rather trust our human judgment
5:10over the algorithm. It goes on to explain that
perhaps one way to think about this is when we
5:16think of algorithms and these AI bots, their
error rates are knowable and they're static,
5:24but human intuition and human intelligence is
perfectable and perhaps we therefore trust that we
5:31can perfect our own judgment in certain tasks. The
research goes on to then try to figure out what's
5:36the right antidote to this and they allow in one
study they allow users of these algorithms to
5:43tweak ever so slightly um different parameters of
these algorithms. When these people are given that
5:51um that leeway to control the algorithm um what
you find is that the error rates will increase as
5:57a result as you would expect. But then you
also see the adoption rates significantly
6:02um increase because people can control it. And
this drives home a really valuable point around
6:09adoption of these AI tools. As a leader, you
might think about what is an error rate that
6:15is acceptable if only it means that you then
create a lot more adoption in the workplace.
6:21The ideal scenario is that people adopt these
tools fully without tweaking them. But we know
6:27that that comes at a cost of lower adoption.
Are we willing to to sacrifice some amount of
6:32um precision in the use of these tools in
exchange for an improved level of adoption?
6:46The final puzzle is this self-sufficiency
spiral. If you think about all the work we do
6:51in an organization, you can categorize them into
solo work and interdependent work. And you might
6:58say that in the future, these tools will allow us
to do a lot more solo work. And a lot of the solo
7:06work will colonize parts of the interdependent
work. And then what gets left behind as
7:12interdependent work, whether it's writing emails
or doing presentations or conducting meetings,
7:19a lot of these tasks will then get intermediated
by these AI tools. When you think about what
7:24it takes to create culture in an organization
or the role of the leader in kind of bringing
7:31people together around a shared mission, a lot
of that is about interactive tasks. A lot of
7:36that is about not being in solitude and doing
isolated work, but actually coming together as
7:42a group. And if the future of the workplace
is a lot more solo and a lot more isolated,
7:49I worry a little bit about what this means for the
future of organizations and our ability to create
7:55cultures and create a sense of identity with the
organization. We've seen other technologies in
8:02the past kind of deliver to us a future that we
didn't quite want. You take for example social
8:08media where it had this promise to create
a more connected world but instead what it
8:13gave us is possibly a more fragmented polarized
world where we perhaps have expected less from
8:19each other and as MIT ethnographer um once said
we are alone together through these tools. So
8:27we don't want this future for the workplace
and we need to think about ways that we can
8:31bring people together through perhaps different
means and and and different approaches so we can
8:37continue to create you know thriving environments
for for people as they engage with these tools.
