Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
This episode is brought to you by Shopify.
0:03
Forget the frustration of picking commerce
0:05
platforms when you switch your business
0:08
to Shopify, the global commerce platform
0:10
that supercharges your selling wherever you
0:12
sell. With Shopify, you'll harness the
0:14
same intuitive features, trusted apps, and
0:16
powerful analytics used by the world's
0:18
leading brands. Sign up today for
0:20
your $1 per month
0:23
trial period at shopify.com/tech,
0:25
all lowercase. That's shopify.com
0:28
slash tech. Perhaps
0:33
you or someone you know has exclaimed
0:35
in dismay, AI is
0:37
going to take my job. Critics
0:40
fear the rapid growth of AI could threaten
0:42
jobs or be used for malicious purposes. Another
0:45
concern is could AI take jobs? Let's
0:47
focus now on AI's impact on the
0:49
way we work. One of the biggest
0:51
worries people have about AI is we'll
0:54
take our jobs. It turns
0:56
out some companies are already using or
0:59
considering using generative AI
1:01
internally. According to a Gartner
1:03
survey from June 2023, most
1:06
organizations appear open to adopting new
1:09
HR tech. Only 15% of
1:11
the HR leaders they surveyed said
1:13
they had no plans to add generative
1:16
AI to their HR processes. So
1:18
whether you're an AI optimist or pessimist, generative
1:22
AI is likely coming to your workplace if it's not
1:24
there already. A
1:27
lot of CEOs are going to their teams and saying,
1:30
okay, you work in marketing, how are you
1:32
going to use Gen AI to sell better,
1:34
to operate faster and cheaper? That's
1:36
WSJ reporter Chip Cutter. He
1:39
covers workplace, management, and leadership issues
1:42
for our corporate bureau. Some companies, for
1:44
example, say that they think they can give up doing
1:46
photo shoots for their products. Why
1:48
couldn't you just have various sort of AI
1:50
tools do that for you? In
1:56
The Wall Street Journal, this is the future
1:58
of everything. I'm Charlotte Carter. Today.
2:01
We're looking at our A I
2:03
Hr future, how close it really
2:06
is, and what sorts of generative
2:08
A I resources. Might become
2:10
commonplace for human resources.
2:12
Stay with us. She
2:19
I am on a mission. Y
2:22
C cost fifty four percent of
2:24
black Americans don't have enough savings
2:26
to retire. So in collaboration with
2:28
thick name artists like Wiper, John
2:31
C. I a released paper right
2:33
new music inspiring the new financial
2:35
future with one hundred percent of
2:37
streaming sales going through a nonprofit
2:39
that teaches students how to invest
2:42
stream paper right now and how
2:44
close. The Gap. Ai
2:54
in a chair has been a thing
2:56
for a while, but it's getting more
2:59
widespread. We. Reported in Twenty eighteen
3:01
that nearly all fortune five hundred
3:03
companies were already using some form
3:06
of automation in their hiring processes,
3:08
whether that's robot avatars, interview in
3:10
job candidates, or computers weeding out
3:13
potential employees by scanning keywords and
3:15
resumes. Last. Month Wsj reporter
3:18
tip Cutter went to the
3:20
World Economic Forum in Davos,
3:22
Switzerland. After talking to
3:24
Ceos and company leaders, he
3:26
says that we're likely to
3:28
see lots more Ai and
3:30
human resources very very soon.
3:33
Tip: What has been the pros and cons
3:35
of this kind of a I use in Hr.
3:37
Well, it's helped a lot of companies. Be.
3:39
faster that they but and will do
3:41
more with less for example of your
3:43
screen name millions of resumes a large
3:45
company probably need automation software to help
3:47
you do that or otherwise if he
3:49
really hard for one person to sift
3:51
through all that so that's been around
3:53
for years right of ways to figure
3:55
out the right candidates to interview ways
3:57
to for example look at who might
3:59
be at risk of quitting. Companies
4:01
oftentimes have built like complex software to see
4:03
who's been in a role for a given
4:05
amount of time, who maybe is in need
4:07
of a promotion and hasn't got one. There's
4:09
all these ways that companies can kind of
4:12
get a sense for whether someone's a flight
4:14
risk. So that's been one way that companies
4:16
have used some of this technology in the
4:18
past. But right now, I think we're on
4:20
the cusp of a lot of change. And
4:22
many companies, many CEOs are telling their HR
4:24
chiefs and all their department heads, you need
4:26
to think about sort of how Gen AI
4:28
is going to change how we work. You
4:30
were just in Davos, so I
4:32
want to talk specifically about generative
4:34
AI because it's being more integrated
4:37
into workplaces. What have you been hearing? You
4:39
could not walk down the promenade, the main street
4:41
in Davos, without just seeing one display after the
4:44
next about sort of how AI is going to
4:46
transform corporate America. Every
4:48
interview with CEOs would somehow come back to
4:50
generative AI and how they were trying to
4:52
use it within their companies. Some
4:55
organizations have a real plan. Others are sort
4:57
of earlier in their efforts. But a lot
4:59
of CEOs are going to their teams and
5:01
saying, okay, you work in marketing. How
5:03
are you going to use Gen AI to
5:05
sell better, to operate faster and cheaper? Some
5:08
companies, for example, say that they think they
5:10
can give up doing photo shoots for their products. Why
5:12
couldn't you just have various
5:14
sort of AI tools do that for
5:16
you? But then AI and HR was
5:18
a big discussion as well. So for
5:20
example, one HR chief at a large
5:22
tech company told me that she has
5:24
done an interesting experiment with Gen AI.
5:27
And there was a case in her company
5:29
where a manager and that person's direct report
5:31
were not getting along. And
5:34
so she said with both people's permission,
5:36
can we upload the chat logs between
5:38
them and see what happened?
5:40
Like why aren't they getting along? And
5:42
so they ended up uploading pages and
5:45
pages of logs. And it came up
5:47
with some sort of interesting answers. And
5:49
one was that the employee was asking
5:51
way too many questions. And then the
5:53
manager was getting frustrated by this. And
5:56
the manager also felt that the
5:58
manager wasn't being heard. So
6:00
just having that information, they were able
6:02
to go back to these people and
6:04
they said the ratio actually improved afterwards
6:07
because both sort of knew, okay, this
6:09
is what's bothering the other person. So
6:11
it's almost like AI as corporate psychologists.
6:14
That's wild. That
6:17
to me was the AI example that I
6:19
remembered the most and that stood out to me
6:21
as the most distinctive in a week full
6:23
of conversations on this. But to me, it
6:25
shows that HR chiefs are thinking, could Gen
6:27
AI really change how we operate? And
6:30
if you think about this example, it'd be really
6:32
boring for a person to go through pages and
6:34
pages of logs and try to see, wait a
6:36
minute, why do these people not like each other?
6:39
What's happening there? But a machine can do that pretty
6:41
easily. But who who might have
6:44
an objection to having AI play that
6:46
role in HR? Well, you could see workers being
6:48
frustrated if this is done in a case where
6:50
they don't know about it or they haven't given
6:52
their permission. And if we're using corporate chat tools,
6:54
we've given up our permission like that our company
6:57
is able to sift through that. But
6:59
in this case, everybody knew this this was going
7:01
on. But you could see down the road where
7:03
some might think this feels a little bit like
7:05
Big Brother. What have you been
7:07
seeing in terms of more high level attitudes
7:09
about generative AI in HR? Yeah,
7:12
so many HR executives are optimistic about it.
7:14
They do feel that they want to use
7:16
it to some extent. And it's oftentimes cases
7:19
where companies might roll out, for example, a
7:21
chatbot that helps people say an employee could
7:23
more easily ask, like, how many vacation days
7:25
do I have left? Or
7:27
what is our policy on X? The type
7:29
of questions where someone might go to an
7:31
HR person and ask them, some HR people
7:33
think like, why can't we just build a
7:35
tool that answers them more easily? That's one
7:38
area where a lot of looking uploading policy
7:40
manuals, uploading benefits guides, all of this stuff
7:42
where it just might be easier just to have
7:44
Gen AI sort of be an aid that helps
7:46
employees navigate the company. Overall, HR,
7:48
like a lot of like a lot
7:50
of functions in corporate America is being asked to do more
7:53
with less. So Companies, This is
7:55
this as we're continuing to see sort of
7:57
an era of layoffs and kind of white
7:59
collar job. Cut in particular so it's
8:01
not a H R Teams are and get tons
8:03
of additional people there. She earth the so if
8:05
if you're in an Hr executive you're trying to
8:07
figure out how do we do more with less
8:09
and many are thinking jenny I could be helpful
8:11
here. So. What stands in a
8:14
way. Of there being more generative
8:16
ai in hr in the future. Part
8:18
of it as cost. I mean companies might
8:20
think they'll save money with this eventually, but
8:23
not in a short term. It's gonna cost
8:25
money to set this up and to you
8:27
know that to be able to do that.
8:29
So that's that's one immediate barrier. And the
8:31
others as technical, know how companies are at
8:33
all different stages on decimate. Some executives are
8:36
really immersed in it. Some companies, for example,
8:38
have had rolled out sort of. Mandatory
8:40
Jenny I training where for example by March
8:42
everybody the company has a complete you know
8:44
this multi our course on Jenny Ice. you're
8:46
up to speed on at your experiment with
8:48
it you're thinking about how could him a
8:50
better jobs And then there's a lot of
8:52
other companies that are saying this is still
8:54
hype, it still unproven. We don't want to
8:56
be an early move on Nes, let's hold
8:58
back. So a brew etti depends on the
9:00
attitudes are the executives have towards of technology
9:02
to. Yeah, Ike. I got that
9:05
impression a bit from this Gartner
9:07
survey from this past summer that
9:09
found that only five percent of
9:11
H leaders were reporting that they
9:14
were already implementing generative ai. And.
9:16
Only nine percent of those surveyed.
9:19
We're. Currently conducting generative ai
9:21
pilots which seems low
9:23
to. Me Very love. Is
9:26
this gonna take off soon? Or. You
9:28
keep hearing that while Twenty Twenty Three was the
9:30
year that we experimented with it and Twenty Twenty
9:32
Four as the year that Jenny eyes rolled out
9:34
at scale at companies and and I mean that
9:37
that's what executive some executives are say, but any
9:39
look at numbers like you just eat have shared.
9:41
It's clear a lot of companies are still very
9:43
early on all of us. We
9:48
may be in the early days of generative
9:51
ai been. Used in Hr but according
9:53
to the founder of girls. First,
9:56
the focus of the and innovation
9:58
and mitigating. Rest. Rushmore
10:00
Some johnny tell us about her organizations
10:03
new chat bot and why she's optimistic
10:05
about how Ai will be used in
10:07
the future. look at work and beyond.
10:12
This episode is brought to you by
10:14
Carnegie Mellon's Tapper School appears s was
10:16
advance your career as which feel and
10:19
and be a so Carnegie Mellon pepper
10:21
school a business can help own year
10:23
degree from a half races assault with
10:25
a thought provoking curriculum. One on
10:28
one leadership Wilson support from
10:30
experience Three, a counsellor and
10:32
full time online hybrid. An
10:34
accelerated. And the a former
10:36
Soviet solicit Fisher is is
10:38
Cmu Edu sticker to. Then
10:40
there's. Wsj
10:49
reporter Chip Cutter told us were
10:51
already seen. Chat bots been developed
10:53
to help employees navigate company policy
10:56
and answer questions. Rush Muscle Johnny,
10:58
founder of Girls Who Code, recently
11:00
helped launch a similar chat bots
11:02
to do just that. Her. Goal
11:05
was to reach further than just
11:07
one company. paid leave.ai attack bought
11:09
designed to help people parse the
11:11
patchwork of laws affecting paid family
11:13
leave. New York State went online
11:15
in December of last year. The.
11:17
Tools Lawn Spicer, Johnny's Other
11:19
Organization Mom First in partnership
11:21
with Movie and Craig Newmark
11:23
Philanthropies. When. She first
11:26
came up with the idea for Danny
11:28
Says. She contacted Open A I founder
11:30
and Ceo Sam Altman. She.
11:32
Says he put her in touch with the
11:34
team behind Techy Be T, who provided early
11:36
technical advice and support. The. Connected her
11:38
with Novi, a developer that is part
11:40
of the Open A I Service Alliance
11:42
who developed Paid Leave That Ayaan Open
11:44
A I Tools. The. Goal of paid
11:46
leave that ai. To. Answer questions
11:49
about leave policies in plain
11:51
language. I begin by asking
11:53
restless oh Johnny to take me through
11:55
how the website works. So.
11:57
Let's say you're pregnant and you go to paid leave
11:59
Today I. I am will can you is
12:01
am I eligible? How
12:04
much money? Miles before? And. It's
12:06
gonna give you an action plan when you
12:08
go to the government website. they don't really
12:10
do that for you on in. One of
12:12
the things I love about the site is
12:15
it gives you a bunch of kind of
12:17
easy prompts because maybe don't know what you're
12:19
supposed to ask for. It doesn't just answer
12:21
your questions, it will draft to an email
12:23
to Hr if you need one. It will
12:25
let you email yourself in action plan or
12:28
checklist. You can save it and Senate's yourself
12:30
into to do to work on it. So
12:32
do you see this as a model for
12:34
other kinds of chat bots? Being used in
12:36
Hr. ah it's very interesting. it in a
12:38
because you can tell him pay we did
12:40
a i wears a traffic coming from most
12:42
of the traffic's coming from linked in and
12:44
if you look and see it's like it's
12:46
literally it or professional sending it to to
12:49
one another But I find it super interesting
12:51
that the you talk to each are people
12:53
that are like yeah bring on the chat
12:55
bots. We've. Been operating in a
12:57
state of fear. That. We need
12:59
to move through that. That doesn't mean
13:01
that we should be like there's no
13:03
risk so good you are, but that
13:05
mean so that we should be mitigating
13:07
the risk. But. Since think he
13:10
that the innovation. right? So let's look
13:12
the risks in the face. Hear what are
13:14
the risks? Of creating an hr
13:16
Tough but. I met
13:18
one of them. Is that it's gonna replace. Workers.
13:22
And. So I think that there's a lot
13:24
of fear that like a tool like
13:26
this will make it like you don't
13:28
really need a human. To. Help you
13:30
navigate this process. Good night out chap. Ah, They'll
13:33
do it either. Second thing is it will
13:35
give you the wrong answers. This.
13:37
Is terrifying way when it comes to
13:40
these laws. Momentous. Events
13:42
that you really need to know
13:44
how much money my going to get
13:46
on what I need to save
13:48
for and and sell these risks are
13:51
these Fears are real. Ah
13:53
I'm If you're really feeling them and are in
13:55
a it's funny when I started this project with
13:57
first muscle were hallucinations. Do I have. To hire
13:59
a. Human Yeah, I mean to
14:01
make sure that they're looking through
14:03
all the answers. And as I
14:05
dug in, I learned so much
14:08
right about how this particular use
14:10
case of paid leave the doesn't
14:12
have the same safety risks because
14:14
the wings bottle full limited. This.
14:16
Particular Large Language Model
14:18
or Llm helps mitigate
14:21
potential hallucinations. So. What
14:23
data went into the training of this chap?
14:25
Ah, Are. Ai is trained on
14:27
the New York State paid leave law. I'm
14:29
like like chatty Bt if not pulling from
14:32
the internet and collecting set of data from
14:34
a bunch of different sources. That's when it
14:36
that way, butter and for a sexy make
14:38
things up. To. Do way hallucination
14:41
against how much information you could possibly
14:43
feel. Absolutely listen. I mean like I
14:45
said this is one of the first
14:47
use cases of doing this so I
14:49
did not want set us up to
14:51
fail and so I what's what's interesting
14:53
is because there is like again l
14:55
a limited L am on there's there's
14:57
less safety risk issues so with the
14:59
large language model your training at all
15:01
any to limited to New York state
15:03
law. Yup that helps get rid of
15:05
some or get around some of the
15:07
possible quote. Unquote Hallucination Correct Yeah. But
15:10
then we do have the the risks of like
15:12
is this gonna replace humans the ones that you
15:14
mentioned How do we deal with those things. I
15:18
don't know. I don't know.
15:20
I mean. It's so funny.
15:22
Again, memorize. Spend my life. Teaching
15:25
girls to an avid mom is like wheat
15:27
said they saw learn how to code. you
15:29
know because there there's this perception might have
15:31
the skills that you're really gonna need and
15:33
as artificial intelligence moment his creativity like you
15:35
would actually me to lean more into the
15:37
liberal arts that because. the bottle
15:40
code for you so i think
15:42
that we have to see in
15:44
many ways how this kind of
15:46
plays out but i still worry
15:48
about it in the future i
15:50
mean if we see more chat
15:52
bots replace hr employees that i
15:54
i won't be able to interact
15:56
with a human being during his
15:58
high that's pretty sensitive it It
16:01
feels potentially dehumanizing,
16:03
like I'm just data relating
16:05
with data. It
16:07
might, but what are we going to do? It's here. It's
16:10
here. We're stuck in this
16:12
conversation of like doomsday, and it's
16:14
preventing us from really capturing all
16:16
the opportunities and use cases of
16:18
AI. Every conversation that we're
16:21
engaging on in AI, is it good or is it
16:23
bad? Is it going to destroy us or not? Is
16:25
it going to like replace workers or not? It's here.
16:28
It is here. And so now the question
16:30
is, is how do we use the technology
16:32
in a way to help
16:35
people, to preserve jobs, right?
16:38
To solve COVID, cancer, climate,
16:40
right? To do good things. Not
16:44
having access to paid leave is a
16:46
major driver of women leaving the workforce
16:48
or downshifting. And so like
16:50
if I can actually navigate having time
16:52
off, money in my pocket, the resources
16:54
I need maybe to hire a care
16:57
worker or to get sorted, right?
16:59
I'm going to stay in the workforce. So
17:01
what's the bigger picture here? Where
17:03
do you hope paid leave.ai goes from here?
17:05
We're going to expand. The goal now is
17:08
I want to prove that
17:11
generative AI can actually increase the
17:13
uptake of benefits and figure
17:15
out exactly how to do that. So
17:17
that means I got to have a handful of governors
17:19
that are partners so we could work
17:21
with the Department of Labor to understand what are
17:24
the pain points, is our tool solving for those
17:26
pain points, and now what's the uptake of benefits
17:28
now that we have this tool? I'm
17:31
noticing this trend in how you're talking about
17:33
things. I think one of the things people
17:35
think about AI is like that it is
17:37
the solution in itself. The
17:39
product, the AI, the chatbot is
17:41
the endpoint. You seem to
17:43
be talking about it in that way. You seem to
17:46
be talking about it as this is
17:48
the next step in a journey towards other
17:51
things. Absolutely. I think it's like a
17:54
tool, a friend, a helper. I see
17:56
this from an activist perspective. My sole
17:58
slave. our focus is to get paid
18:00
leave in child care passed. And
18:03
so I'm thinking about all the different ways, all
18:05
tools that we can have to do that. Part
18:09
of why I think this got so much attention, it was unexpected.
18:11
It was an unexpected use
18:13
case of generative AI. And so
18:16
I think as we're looking at what
18:19
are the use cases of generative AI, go
18:22
to the unexpected, go to the most
18:24
vulnerable. Like the next place I want to go
18:26
is like, how am I helping military vets access
18:29
their benefits? How am I helping
18:31
people who need food stamps access their
18:33
benefits? How am I helping people on
18:35
Medicare access their benefits? Who's the most
18:37
vulnerable? And how do we build, you
18:39
know, AIs for
18:42
those communities to
18:44
help them change the
18:46
quality of their lives? That's
18:48
where we need to begin. The
18:51
Future of Everything is a production of
18:53
the Wall Street Journal. This episode was
18:55
produced by me, Charlotte Gartenburg. Thanks for
18:57
listening.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More