Latest / Unspoken Career Truths / AI as a Coworker: What to Trust, What to Guard, and What Humans Still Own

Transcript

AI as a Coworker: What to Trust, What to Guard, and What Humans Still Own · Unspoken Career Truths

0:00
Welcome to unspoken career truths.
0:02
I'm Cara Wood. And I'm Lauren Levac Gilbert
0:05
and. We're here to be the Co workers
0:07
you can trust. Welcome back to Unspoken Career
0:12
Truth. We are so excited to be back and
0:15
pivoting our focus to the future of work because you know, we're
0:18
all trying to figure it out, right?
0:20
Hell yeah. So Speaking of the future of
0:24
work, something that has happened lately that's made
0:27
things a little bit tipsy turvy in the world, It has two letter
0:31
letters. What do you think it is, Cara?
0:34
I know what it is. I have the script in front of me
0:36
and it's what everybody's talking about.
0:38
Everyone's obsessed. AI Brain.
0:41
AIAIAI Artificial Intelligence. Yep, it's really changing
0:48
everything in so many different ways and so many different
0:51
aspects of our lives. We've already talked a little
0:54
bit about it last season and we talked about like how to think
1:00
about it, like how you should put it in perspective with
1:04
what's happening at work. But now we're really gonna talk
1:06
about kind of future looking and how AI can be a Co worker and
1:10
might be a Co worker for you in the future.
1:13
What do you think about having an aging Co worker Cara?
1:16
Honestly, AI as a Co worker has been the most frustrating than
1:21
like more frustrating than any human colleague that I've ever
1:24
had. And I think all of us have had
1:27
frustrating human, human colleagues.
1:28
That's just the nature of the world.
1:30
But I, you know, there's a few things that I look for when I
1:33
even think about like adding an agent to my Co worker roster or
1:39
working with one hiring 1, so to speak.
1:41
And it's like, is there transparency in the ways that
1:45
it's going to actually be making the decisions for me?
1:47
What are its motivations and like, expectations and
1:50
definitions for success? And these are questions that you
1:53
can ask software creators if you're looking at materials.
1:57
And they're also questions that you can bake into prompts that
1:59
you're writing. I'll just like say that another
2:02
big thing for me is like, are there ratchets in place for AI
2:05
to flag when it's confused or when it has gaps in the
2:09
information that I've provided or other people have provided?
2:13
Do I have transparency into what data feeds it's drawing from?
2:16
Like can I upload the data sources or is it pulling from
2:20
the world? And then what in the world is it
2:22
pulling from? And can you trust those sources,
2:24
right? Like, do you care if it's from
2:26
Wikipedia or do you care if it's?
2:28
Reddit. Reddit.
2:29
Exactly. Yes, exactly.
2:32
And if you're really thinking of it as like a trusted Co worker,
2:35
just like you would train a coworker to say this is the
2:38
standards I want you to uphold, Like that is exactly the kind of
2:42
feature that I want in a future AI coworker.
2:45
And then the other one is like, if I am releasing it as an agent
2:49
on my behalf and we're gonna talk about agency and a little
2:52
bit like, does the end user who's experiencing the AI have
2:56
an opportunity? How do they do they know that
2:58
they're like that they're being addressed with AI?
3:02
And is there a way for them to communicate confusion or
3:04
concerns back to me, the human? So yeah, I am.
3:09
I am open to seeing how AI evolves.
3:12
But so far, there's just so many ratchets and things you need to
3:15
put in place. It really is like hiring and
3:17
training a real human. And this is the dumbest it will
3:22
ever be. So hopefully it gets better.
3:26
One thing that we do want to talk about before we like really
3:29
dig into the meat of this is just to define things because
3:33
there's a lot of talk about AI and AI in general has been
3:36
around for a really long time. So when I talk to people like my
3:40
husband who's in science and I talk about AI, he's like, I've
3:43
been using that for years. Why is everybody talking about
3:46
so? Yeah, totally.
3:47
I I like, I'm like I took a class in AI when I first moved
3:51
to Boston like over a decade ago.
3:53
So exactly. So really the, the visible shift
3:56
here is really around generative AI or Gen.
3:59
AI, which is creating content, understanding natural language,
4:05
looking at data, interacting conversationally, and is now
4:08
widely accessible to the general public outside of things like in
4:13
science or in a lab or in in some of those other places.
4:16
And now generative AI is layered with traditional AI and it's now
4:22
enabling agentic behavior where AI can actually act on behalf of
4:28
us. So it can go and buy something
4:31
or in agent mode in chat to BT, it can book travel for you with
4:38
set parameters and things like that.
4:40
But this whole agentic shift is another thing that's I think
4:43
really important to define because agentic AI is referring
4:48
to AI systems that don't just respond to prompts.
4:52
They actually can plan, decide, take action.
4:55
And that action piece is a really big difference.
4:58
And that's what a lot of people are talking about now, because
5:01
if an, if an agentic or an AI agent is taking an action on
5:06
your behalf, that means they have some level of autonomy.
5:10
And so there's a little bit, I think caution around that.
5:15
And there's definitely a debate about what agentic actually
5:18
means. Yeah, I love that you call that
5:20
that distinction, Lauren. And like I actually start to
5:22
like picture Hollywood to keep these things straight.
5:24
Like, you know, you've got agents out there and they're
5:27
scheduling lunch and they're like negotiating contracts on
5:29
your behalf. Like, are you cool with that?
5:31
Do you trust that agent Versus like Jen AI, which is just like
5:34
a really speedy CGI compositor, right?
5:37
Like it could take the film that someone made and create its
5:40
owner, like layer on and be like, here's what the aliens
5:42
going to look like. Like, based on everything the
5:44
directors told me, I'm going to just create a bunch of stuff and
5:47
you got to sort of review it and decide.
5:49
So that's sort sort of where my mind goes of like keeping the
5:52
two distinctions straight. I love that.
5:55
I also think it's important to think about it as like,
5:58
regardless of if it's traditional AI, Gen.
6:01
AI, agentic AI, it's a tool, right?
6:04
Like we're using it as a tool. There's an opportunity for it to
6:08
help us with efficiency, with scale, make our lives easier.
6:12
But it's not like robots are going to take over the world
6:15
tomorrow, right? Like I think we just need to
6:16
kind of level set around that. But there is definitely a level
6:20
of fear and anxiety than we addressed in our last episode
6:23
about like losing your job or getting replaced or things like
6:26
that. But there's definitely ways to
6:28
mitigate that and think about what skills are really important
6:32
and valuable to focus on. Yeah.
6:34
I mean, I feel like that distinction is crucial.
6:36
And if you think about where AI started, it really is like, it's
6:40
embedded in your Google Maps. It's just something that you
6:43
use. And when Google Maps tells you
6:45
there's a traffic jam or there's a police officer, you know,
6:49
checking everyone's speed and you don't see that person,
6:52
you're not suddenly like, how did that mistake happen?
6:54
It's really frustrating when it like, takes you down the wrong
6:56
Rd. But we've all been in that
6:58
situation. It's just because it's based on
7:00
the data it has, it's made the wrong decision.
7:02
And every AI you use, no matter like how sophisticated feels, is
7:06
actually doing the same kind of thing.
7:08
It's making a prediction. And some of them aren't going to
7:10
be like, perfect. I think it's really important
7:13
when we're talking about something like this to really
7:15
look at the numbers because I think that helps kind of paint a
7:17
picture of what's actually happening.
7:19
So let's look at the numbers for a second.
7:21
In terms of statistics around AI, 70 to 75% of companies say
7:26
that they're using AI in at least one business function.
7:30
So there's definitely AI usage and in everyday life and up to
7:36
40% of current work activities could be automated by AI, which
7:41
I think is a really interesting step because there's a lot of
7:44
ways that AI can help, but you have to have the kind of basics
7:49
of like the data and understanding the processes
7:51
before you can layer it on top 100.
7:54
Percent like AI could speed up something that's already broken
7:58
and not help it at all if you don't have your foundation set.
8:02
Exactly like bad data in bad data out, right?
8:05
Like you can't just tack AI on it, it's going to solve all your
8:07
problems. And 50% of consumers say they're
8:10
open to using AI for product discovery or shopping help.
8:14
And I think that's really where it becomes more democratized
8:18
across everyone in the world, right?
8:20
I think when Google incorporated AI into their shopping feature,
8:25
then people that were using Google didn't realize they were
8:28
using AI. But now they're used to those AI
8:31
summaries. They're used to having more of
8:34
those conversational conversations and asking
8:36
questions. And so I think once that becomes
8:39
more of an ingrained behavior, we're going to see a lot more
8:42
change happen. It kind of already seeped into
8:45
search when folks were seeing that they, you know, it used to
8:48
be like you had to kind of hit, hit the right keyword and now
8:51
you can have now you can like type a whole paragraph in and
8:54
that first result in Google will sort of parse what you meant or
8:58
give you a story back based on your search phrase.
9:02
Yeah. And I, you know, I feel like,
9:05
hey, I can actually make us more human, right?
9:08
Like it can actually help take away some of the kind of hands
9:13
on keyboard, really kind of detailed stuff and enable us to
9:16
kind of think more broadly and, and be more strategic or more
9:21
creative. And so I definitely think
9:23
there's an opportunity there. And I think it's also
9:25
interesting to look at like generationally who's using AI,
9:30
30 to 40% of Gen. Z millennials already use AI or
9:35
chat tools to purchase products and to do research.
9:39
And I think when you're thinking about like the next generation
9:42
after that Gen. alpha, it's going to be way more ingrained
9:45
into how they operate and what they do.
9:47
So it'll be interesting what their behavior is as well.
9:51
Yeah, there's like a couple of things in my brain. 1 is when
9:54
you say like, makes us more human.
9:56
Just sort of depends on how you're defining humanity and
9:59
what you're expecting people to react to.
10:02
So like, is it going to free up time?
10:04
We're still discovering that. Is it going to change the way
10:07
that people interact with their digital screens?
10:11
Absolutely, 100%. And then I, I know there's a lot
10:16
of conversation already about how it's impacting Gen. alpha's
10:20
mental health as well as their logic and reasoning.
10:23
And so I think there's something to folks still knowing how to
10:28
have a really interesting conversation and generate
10:31
questions and natural conversation on their own.
10:34
And then what they bring to AI because of those skills they
10:38
already have. I think that's a good teaser
10:41
too. For some of our upcoming
10:43
episodes, we're going to talk about some of the skills that
10:45
will still be super valuable even in an AI world.
10:48
So big teaser. So in my day-to-day job, I
10:51
actually spend a lot of time thinking about how AI is going
10:54
to enable organizations. And I'm kind of toying with
10:58
calling it like the hybrid blend of work where there's really
11:01
this like shift in the workforce where work that was once done by
11:05
humans with the support of technology will now be done by
11:08
humans with the support of technology and an added element
11:12
of agents. So we've gone from human only to
11:16
human technology to human technology and agents.
11:20
Yeah, and how do you like assess what agents can actually do and
11:23
support? That's a great question and I
11:26
think there's certain strengths to each of those elements,
11:30
right. So the first thing to note is
11:32
that agents are constantly getting smarter.
11:34
Like I already said, like this is the demos that they're
11:36
they're ever going to be. And so they they might not be
11:41
able to do it now, but they might be able to do it in a few
11:43
months and they might grow and have much more intelligence and
11:46
capability. So the great way to think about
11:48
it is that an agent can help handle some of those kind of
11:51
straight forward zeros and ones kind of computing where humans
11:56
can handle more of the messy, imperfect parts of decision
11:59
making. So I kind of think about it on
12:01
the spectrum from work done by AI, work done with human and
12:06
agent orchestration, and then work that's done completely by
12:10
humans. Cool.
12:11
And so like let's start with work done by AI.
12:14
What are? What are some examples?
12:16
Yeah. So let's think about this.
12:18
With high volume tasks, repetitive data intensive AI can
12:24
look at so much data and analyze it so much faster where it may
12:28
have taken like 15 or 20 people to look at it previously.
12:33
Low ambiguity, just again, like zeros and ones.
12:36
Like I'm asking you analyze this data, what's coming out of it.
12:39
I kind of think about it as I'm like air quoting here, but like
12:42
the grunt work, right? Like hands on keyboard, keyboard
12:45
work that can really take a lot of time.
12:48
So creating content for a website or something like that,
12:52
it can write and create a ton of content for multiple different
12:57
websites and multiple different personas that can then be
13:00
reviewed by humans and then launched online.
13:03
Where? But Lauren, if it's being
13:05
reviewed by humans, then isn't that in your second bucket?
13:10
And if it's not, then isn't it something that is just going to
13:14
generate the same web page that we don't want to look at?
13:17
But that I think is also, that's a great question, Cara.
13:19
So it's still in the first bucket because I think it's the
13:22
work that's getting done. I think the orchestration piece
13:26
is where you're working in conjunction with the agent.
13:29
So when I think about like work that's just being done by AI,
13:32
it's like it's giving you an input or output.
13:35
And honestly, you don't have to have the checks and balance in
13:37
place, but I think that's best practice right now.
13:39
I think in the future you'll have an AI agent who does the
13:43
work and then you launch it into the world and you don't worry
13:45
about it. I think now there's that in the
13:47
loop like approval step. Can I make an adaption to how
13:51
you like what you described? Because I think what you're
13:54
saying or what the ideal scenario is, is you've got
13:57
someone, like you said, web pages.
13:59
So like you've got someone who made awesome sneakers and knows
14:05
those awesome sneakers and built all the product specs on them.
14:08
And then they put them into an AI agent and that AI agent
14:12
created the sale page for the mobile site and the website and
14:21
the website if you wanted the the sneaker to be green and that
14:25
didn't need any like oversight. But if you ask it instead and
14:32
you don't give it the source of like the product person that
14:34
really knows that shoe and you just say, just describe an
14:38
awesome sneaker to me. Like you're going to get a Nike
14:40
ad and you're not even selling Nikes.
14:42
And all of a sudden, the buyer on the other end isn't even
14:46
like, making the distinction between like, what makes that
14:49
shoe really unique? I think that's a great example.
14:52
Yeah. So like if you and I were to
14:54
have started a company about a product, like we would know so
14:57
much about it. We could use the AI and then it
14:59
could get launched. That's AI.
15:01
Think that's a really great designation where it's the level
15:04
of trust and also comfortability with risk that you have when it
15:08
comes to work and content that's done by AI, right?
15:13
Because if you have a level of comfort that says, hey, like I
15:15
trust what's going to come out of it, I'm willing to take the
15:17
risk, then it can go up without the guardrail or the checks in
15:20
place. And.
15:21
You know what's funny is one of the most exciting use cases and
15:24
I was hoping it was going to fit into this bucket of like, how do
15:27
you assess what agents can just do alone?
15:29
I'm curious if you think it fits in is 1 where you can just say
15:34
here, all my tests from, you know, my healthcare providers
15:38
here are all my scans now tell me what I have and what's wrong
15:40
with me. And like catch a cancer that no
15:42
doctor could like naturally see because you're brilliant and
15:46
you've scanned, you know, 10 years worth of like cancer
15:49
patients or something. And I, I, I don't, I know, I
15:52
think that is something that people want to see AI do.
15:55
And I'm curious if it falls into bucket one where it's just work
15:58
that's done by AI or if we're talking about something that's
16:01
human and agent together. I think it's really interesting
16:04
that you brought that up because ChatGPT just launched a health
16:07
feature, right? So you can actually do that.
16:09
You can connect all of your health records into ChatGPT and
16:13
it can analyze and assess and point to things that it's seen
16:18
or what it's what, what they think you should do for your
16:21
health. So I think that falls very much
16:23
in the work that's done by AI because that's like the analysis
16:26
piece, right? Like I'm analyzing something,
16:29
I'm putting in an input, I'm getting an output, and that's
16:33
the work that was done. So I actually think that fits
16:35
perfectly into this. That's an interesting 1 though,
16:38
right? Well, in people, I was going to
16:40
say another one that if you, if you don't want to take the risk
16:42
at with all of your health data, which I completely understand,
16:45
there's another shorter version of that, which is I've heard a
16:48
lot of people using it for their New Year's resolutions.
16:51
Like give me an exercise plan and then let me upload my data
16:55
of like the last month of exercising and you tell me what
16:57
I need to to change or modify based on my exercise plan.
17:00
And that one feels like you're still contributing to your
17:03
health, but it's maybe not as personal as you handing over all
17:07
your healthcare records. I love that and it can be like
17:10
an unbiased. Well, I'm air quoting unbiased
17:13
because that's challenging, but it can be another to provide a
17:18
perspective based on the data that is inputted.
17:21
Yeah. Well, so let's then truly talk
17:23
about like, what are some of the examples of when you do bring a
17:26
human agent together, What, what kind of work does that look
17:28
like? Yeah.
17:30
So characteristics of that would be guardrail setting, managing
17:33
expectations like interpreting signals, changing prompts and
17:37
defining intent. So an example I always think of
17:40
is like handling and review. So stepping in when an AI agent
17:44
flags a situation that might not be great, whether that is a like
17:51
a ratings and review that might have been listed online and it
17:54
might flag it and say like, hey, this is something that's being
17:57
said about your product. And then the human can come in,
18:00
provide that feedback to the R&D team or update the content.
18:04
It can be something like supply chain, right?
18:09
So you can, so that's talking about like packaging and
18:12
providing products to the retailers or all of the places
18:15
that you're selling your products.
18:17
So you could set up a system, you can set up an AI agent that
18:22
is telling you the capacity of your warehouse and you set up a
18:26
specific system. You say, this is how much space
18:28
I have, this is how much I want to sell.
18:30
And you can kind of collaborate with the AI agent on all of
18:33
that, but it can flag and say, hey, we're running out of space
18:36
here. And then the human could come in
18:38
and say, no, like, we want to use this space for this specific
18:42
product because we think it's going to be a big seller next
18:44
year or next week or during the holiday season.
18:47
And it can interpret those signals and then make a
18:50
recommendation based on what's actually happening in real life.
18:53
I want to throw another one that's like potentially a pie in
18:56
the sky use case of AI, but I'm, I know people are investigating
19:00
this and I'm super excited to see it happen.
19:02
And I think it falls into this category, which is weather
19:05
management prediction, climate changes and outcomes.
19:08
So again, there's so much. Data.
19:10
That's a good. One, yeah, just like so much
19:12
data that needs to be analyzed. So then you need a human on the
19:15
other end to say, yes, this is an alert that we should actually
19:18
send to like warn people that there's a flood coming.
19:20
Like, OK, no, this is something that we should change, like how
19:23
we're doing irrigation or how we're planting the season.
19:26
So it's giving you a ton of data analysis, but then you still
19:29
need someone on the other end to just to make the decision about
19:32
what's going to actually be done based on that data.
19:34
Yeah. And I think about like
19:36
executives at really big companies, so Microsoft or an
19:40
open AI or name insert name of big company, right?
19:43
Like they need to understand what's happening in their
19:47
business in every single aspect. Previously, you would have to
19:51
get all those teams to provide the data and then they would
19:55
need to prep their bosses, who would prep their bosses, and it
19:57
would eventually get up to the executive team.
20:00
Now executive teams can use AI to compile all of that
20:04
information and see it at a more real time basis and be able to
20:08
interact with an agent to ask questions or understand, like
20:12
why is this happening? And then make decisions based on
20:16
what's being recommended by the AI agent in just a faster.
20:20
Iterative way, where previously would take kind of weeks to
20:24
compile all that data and then by the time you look at it, it's
20:26
outdated. Well, and I can picture like
20:29
there's a devil advocate person and I think it's why we're
20:31
talking about AI plus a person who's like, if you just had, if
20:36
your boss was just monitoring every single like if there was
20:39
data collection across every single computer that was
20:41
attached to every single large company, there'd be so much
20:44
nuance and context into what's going on that like a lot of it
20:49
would need explanation. You know, like, oh, I was, there
20:51
was a sick day and my dog needed a walk.
20:53
Like that's why I wasn't like computer to key is this exact
20:56
moment. Like that's why I logged in at
20:58
10:00 PM at night. And there's some of that that AI
21:00
will like overtime learn how to interpret, but there's always
21:04
going to be that need for like the human to be able to like
21:06
both explain, provide context, and understand a. 100% and and
21:12
that applies to business decisions, right?
21:14
Like the AI might spit out and say like you should invest X
21:18
amount of dollars in this, but like as the leadership team, you
21:22
know that that doesn't make sense for you because you are
21:25
making a strategic decision to do something else in another
21:28
area that might make sense. That's where it's this kind of
21:31
combined collaboration between you and the AI.
21:35
The AI is providing you the data and information in a faster,
21:39
more iterative way, but the human, to your point car, is
21:41
still making the choice and still making the decision
21:43
because they have the context and the broader understanding
21:46
and expertise. Well, so then let's just talk
21:49
about the things that humans just do well on their own that
21:52
AI shouldn't touch. Yeah.
21:54
So think about this as things that have high ambiguity, things
21:58
that are more relationship driven, creative, problem
22:02
solving, anything that involves empathy and emotional
22:05
intelligence. So I always think about this as
22:08
like building community in connection.
22:10
So building a brand community. So getting all of the people who
22:14
love your brand together and having them have conversations
22:18
or provide feedback about new products that are coming out,
22:22
hosting events. But Lauren, wouldn't you want AI
22:25
in the room? Like wouldn't you want AI in the
22:28
room? Like I'm thinking of all the
22:29
times that we have AI in the room just to monitor meetings
22:33
and tell people what how that meeting went and what the
22:35
takeaways are. So Cara, I'm actually glad
22:38
you're bringing up all these examples because that's
22:40
demonstrates how much of this is a spectrum, right?
22:43
Like you can float in and out of work that's done by just AI,
22:47
work that's just done by humans and work that's done by both.
22:49
And I think we're realizing as we're talking through this,
22:52
there's a lot of Gray areas with this.
22:54
Like you definitely could have AIB in the room and take notes
22:57
and provide follow-ups and, and make clues and things like that.
23:01
I think where it's solely solely humans would be like an in
23:04
person event or planning a conference.
23:08
There's probably pieces of it that can be supported by AI, but
23:12
it still needs that kind of human element.
23:13
But I think it's an excellent point because this is very, very
23:16
much a spectrum. Yeah.
23:18
And I'm also thinking about, like, one of the first use cases
23:21
when Chachi PT first launched and people were sort of trying
23:23
to, like, wrap their heads around how other people were
23:26
choosing to use Chachi BT. And it was like people are using
23:29
it as a way to craft messages that they're sending to one
23:33
another. Like, there are people now who
23:34
use it as like ways to filter who they're meeting on dating
23:38
apps or how or like, let me bounce.
23:40
It makes me sad. Yeah, like how I bounce ideas
23:43
off of and it's not something it is something that's happening.
23:46
So we we can't just say, hey, don't like no one should do it.
23:49
And I think back to your like idea of is AI making us more
23:54
human? It's like there's a reason why
23:56
people are using it that way and is AI reflecting back to people
24:01
what they're saying or like interpreting what they're saying
24:05
in a new way? No, I I completely agree.
24:07
And I think that's why it's such an interesting time because as
24:13
you can even hear from our conversation, like we're
24:15
debating back and forth because it's not very clearly defined
24:18
right now. Like there it's we're just kind
24:20
of figuring it out and trying to fit something that's a little
24:23
bit new and evolving into like the boxes that make sense.
24:27
But I think the important part to take away from this is like
24:31
there are places it can help. There are places where it can't
24:34
help. There's places where like the
24:38
pieces that make us human are still super, super valuable, and
24:43
if you keep that perspective that it's a tool, I think it
24:46
makes it easier to understand how it's going to affect the
24:50
future of work. Yeah, one of the things I was
24:53
thinking about when I read through your awesome script and
24:56
I knew we were going to be talking about this was like your
24:58
3 tiers and how another way to think about them might be at
25:02
what upfront planning and scoping you actually need in
25:05
place before you let AI handle something.
25:08
So like for those high volume repetitive tasks, have you
25:11
actually scoped some of the outlier cases?
25:14
Have you brought in enough of your team members and like other
25:17
testers or users to understand all the different variants that
25:21
the system might need to handle before you like let it loose?
25:27
And then similarly with the human like oversight ones, like
25:31
it's kind of potentially the nice default.
25:33
And I think we started to talk about this like it might be the
25:36
way that you first launch a project to sort of see, wait, I
25:39
want AI to surface all these things to me to understand first
25:44
how much it can handle on its own.
25:46
You know, and I still think that when you're doing like creative
25:48
problem solving, AI can actually just act as like a thought
25:52
partner. I know a lot of folks are using
25:54
it right now as like a potential note taker to sort of reflect
25:58
back. I saw so many people at the end
26:00
of last year say my ChatGPT roundup, like taught me
26:04
something about myself because I had been talking to it for a
26:07
year. And now I get to see like my my
26:10
thoughts reflected back through a different filter than my own
26:13
brain. That's a great point.
26:15
Cara and I, I wanted to talk a bit about when you were talking
26:18
about training the model and the outlier cases.
26:21
A lot of the brands that I work with, this takes time.
26:25
Like we're not saying overnight you're going to have an AI agent
26:28
that is like able to understand your business and your problems
26:32
and like solve all of these things.
26:33
The brands that I'm working with, this takes like 3 or 4
26:36
months, right? You have to train it on internal
26:39
conversations, on documents, on ways of thinking, ways of
26:42
working. So this is definitely a process
26:44
and it's iterative. And I think you'll see that
26:48
when, when or if you start using it and you start having it as
26:51
one of your Co worker. It's more of something that's
26:54
going to evolve over time. I think you're right and I, and
26:59
actually, I feel like it's a great point for anyone listening
27:02
who's either being told that they're taking on AI projects or
27:06
being tasked with something. Because a lot of times one of
27:09
the values that leadership is attaching to AI is speed and
27:14
efficiency. But to kind of push back on that
27:17
initially and in a realistic way to say these pilots take a long
27:22
time. So let's really set goals around
27:24
when it's going to speed up. Because if you start the project
27:28
hoping to gain efficiency, you might move too quickly before
27:32
the system that you're building is ready.
27:35
I think, well, that piece about incentives, I think is really,
27:37
really important. Like how are you measuring the
27:38
success of your employees and how are you measuring the
27:41
success of implementing AI? That's another space that I
27:44
think is really, really evolving because you can't just use
27:46
efficiency as a measure, right? Because to your point, it's,
27:49
it's providing a different, it's providing the wrong angle,
27:52
right? Like just don't get it done
27:53
faster. Like you want to get it done in
27:56
a way that gets you to a better outcome and that it provides
27:59
more value to you. So that's something from an HR
28:02
perspective, like how do you measure the success of your
28:05
talent? Like what are the new metrics
28:08
that you need to be measured on? If you have AI agents as Co
28:11
workers or you're managing agents?
28:14
And I think this is also something that I'm really
28:17
passionate about it and very interested in because it feels
28:20
to me like you're moving from more of a like individual
28:25
incentive and an individual assessment to more of a
28:29
collective assessment. Because it kind of puts the
28:33
entire organization on the same playing field with the level of
28:36
knowledge. If everybody has the same agent
28:38
that they can use and you can all work towards one goal versus
28:41
trying to be siloed. Yeah, that makes sense.
28:46
So as you can tell, there's a lot of questions and there's a
28:49
lot of things that we're trying to figure out.
28:50
If you have questions about AI, like please let us know comments
28:55
so that we can address them because we're all really
28:57
figuring this out together. So just to wrap it up fact, AI
29:03
is not going away anytime soon. But let me interject anyone
29:10
who's telling you that like they know exactly what role AI is
29:14
going to play long term or which AI company is actually going to
29:17
be dominant, I swear they're selling you something.
29:20
It's probably shares our futures in some platform and they don't
29:24
know yet because no one does. We're still in the experimental
29:27
phase of this new technology. Yes, no one is an expert.
29:31
We're all figuring it out. Another fact, this is the
29:34
dumbest AI will ever be. And I'm going to chime in again
29:39
here, Lauren, and just say that I think there's two factors that
29:44
you can judge AI on. And one of them, I agree, it's
29:48
getting smarter because there's a capacity for data analysis,
29:52
for sensory recognition, for speed of decision making.
29:55
And that piece is getting faster and better all the time.
29:59
But then there's another component that we need to judge
30:01
AI on that we've been talking about, which is like that output
30:04
and end results. And it's like, wait, OK, is this
30:07
saving you time? Because it is calling your
30:10
Internet company and negotiating your rate even before you
30:13
realize your old promo rate was expiring, Which is something I
30:16
want AI to do for me because it, I love that.
30:19
Yes, it's annoying. Or is it undressing 14 year olds
30:23
on Grok, which is also happening on AI right now.
30:26
And it is the stupidest and criminalist thing to be doing on
30:29
my eye. So I think that, you know, this
30:31
is intelligent tech, but how intelligent it's going to be is
30:35
really like what we're tasking it to do and what how it's
30:38
interpreting people's incentives and like appropriate behavior.
30:42
And so there's no guarantee that it's actually going to be doing
30:44
smarter things over time. That's up to us.
30:47
I think that's a fairpoint because we are prompting it and
30:49
we are asking the question. So that is an excellent point.
30:52
So to close out, what should you be doing?
30:55
Try to embrace it, learn, listen, try to understand how it
31:00
can affect your industry and really just kind of be a sponge
31:03
to everything that's happening and observe.
31:05
Like I said, we're all figuring this out together.
31:07
We're figuring it out. So just listen and learn and see
31:10
where it could be an opportunity to you in your day-to-day life
31:15
or in your job. Yeah, and remember, you're in
31:17
the driver's seat. It's a tool.
31:19
It's a tool. Thanks everyone.
31:22
And that's a wrap. Thanks for joining us on
31:24
Unspoken Career Truths. Don't forget to follow us on
31:27
Spotify and Instagram. And if you like this episode,
31:30
share it with a friend.