Latest / The Innovation Show / Ep 666: Shuo Chen — Fractional Founders, Frontier Firms and Why Enterprise Data Is AI's Gold

Transcript

Ep 666: Shuo Chen — Fractional Founders, Frontier Firms and Why Enterprise Data Is AI's Gold · The Innovation Show

0:00
Oftentimes, Silicon Valley describes it as like, I just quit my job, I
0:03
just dropped out of school, and I go do this thing full time and keep
0:06
on working on it until it works out.
0:09
That's actually not what I've observed firsthand.
0:11
Oftentimes, what I see is who's working in a job has had, you know, 5, 10 plus
0:17
years of work experience where they've really deeply gotten to know that field,
0:21
gotten to understand the problems, and oftentimes experience that as a
0:26
personal pain point, or have served customers who are underserved and they
0:30
wish they could build a better solution.
0:32
And then instead of quitting their jobs cold turkey, they took the time to really
0:35
understand what the pain points are and how to fix it, and then they kind of
0:39
side hustled to building a solution.
0:41
Oftentimes, many of these intrapreneurs try to pitch this idea to their boss
0:46
and, you know, all the way up the command chain and try to do this in-house.
0:50
But oftentimes, as we were talking about earlier, Aidan, you know, the organism
0:54
rejects the organ and they can't do it in-house, and so they have no choice
0:59
but to go and do it as a startup.
1:01
And so I oftentimes see some of the smartest entrepreneurs being
1:05
reluctant because they wish they could have just done it in-house.
1:08
I've done a lot of work with EY, and one of the conversations I've
1:12
consistently heard is, "Hey, if now we don't necessarily need entry-level
1:17
employees to do very basic things like update spreadsheets and do more
1:20
tedious work, then we can actually bring them along to client re- meetings
1:24
and build relationships earlier on, have them do more at a younger age."
1:28
So that's one side of the coin.
1:30
As most folks know by now, Spirit Airlines had declared bankruptcy.
1:34
And what became somewhat surprising to people was that their most
1:37
valuable asset wasn't just, like, what terminals they had access to and what
1:42
airports, but rather their literal, literally decades of operational data.
1:46
This included things like employees' emails, anonymized, like, customer
1:50
data, operational data, literally how they were running processes.
1:54
So if you think about the world, um
1:59
95% plus of the world's data is not available on the
2:03
internet because it's private.
2:04
It's stored on within enterprise databases.
2:07
And that is gold for these AI companies because if you think about how AI works
2:13
Before we start, I wanna thank
2:41
our sponsor, Kyndryl, who runs and reimagines the technology
2:45
systems that drive advantage for the world's leading businesses.
2:48
With a unique blend of AI-powered consulting built on unmatched managed
2:52
service capability, Kyndryl helps leaders harness the power of technology for
2:56
smarter decisions, faster innovation, and a lasting competitive edge.
3:00
You can find out more about Kyndryl at K-Y-N-D-R-Y-L, kyndryl.com.
3:05
the top 1%, of startups spend 8.3 times more on AI per employee than the average.
3:11
It's a startling fact, and it leads to extraordinary outcomes.
3:16
It comes from a newsletter from a lady who I worked with last week
3:20
when we were in Vancouver, and I had to have her on the show.
3:23
She has a fantastic newsletter where she explores three-two-one, three
3:27
trends in startups, tech and venture, two theses on what comes next in the
3:32
world, and one tool that she loves.
3:35
She is an investor with IOVC.io Ventures, where she works in the Bay Area,
3:42
and she's a professor with Stanford and indeed Singularity University.
3:46
She works with corporates all over the world.
3:48
She keynotes, she MCs, she interviews.
3:52
She is Shuo Chen.
3:53
Welcome to the show.
3:55
Thank
3:55
you so, so much for having me, Aidan.
3:57
I probably didn't do you justice there.
3:58
So maybe you'll give, an explanation of the type of work you do, especially
4:03
because of the work you do with VCs.
4:05
It's really interesting, and you have a unique view of how you do that work
4:09
Really appreciate that.
4:10
So to give everyone a brief context to what I do, I guess entrepreneurship
4:15
has been the connecting dot across all of these different experiences.
4:19
I was a former entrepreneur after selling the company came back to
4:22
invest in early-stage entrepreneurs, and now of course, I also teach
4:25
entrepreneurship at Stanford and Berkeley.
4:27
And so you can see that, I, I very much caught the Silicon Valley
4:31
startup bug to say the least.
4:33
And one of the things that I found so interesting was when we s- when we
4:37
worked together, you talked about this concept of a fractional founder, and it
4:41
reminded me so much of so many listeners to the Innovation Show, and that's
4:45
what really caught my attention was it's very similar to somebody who works
4:50
inside an organization, an intrapreneur who's coming up against change, coming
4:54
up the corporate resistance, coming up against the politic of an organization.
5:00
And you talk about this concept of a fractional founder, and it really
5:04
has so many common traits to that intrapreneur inside an organization.
5:07
I'd love you to share who this person is and what the role is evolving to become
5:12
Yeah.
5:12
Thank for that because first of all, if there is another name that
5:17
listeners to the show identify by, please reach out and tell me, 'cause
5:20
I'm looking for a better name.
5:22
But the closest I've gotten to describing this phenomenon I've just observed out
5:26
in the wild, having done early-stage venture investing for the past 10-plus
5:30
years, is fractional founders, which is, there's multiple ways to think about it.
5:35
But the way I think about it is I think fractional founders are
5:38
essentially entrepreneurs in the earliest stages of starting something
5:42
where they're still side hustling.
5:43
So, The best entrepreneurs we've invested in for the past decade-plus,
5:48
I've just observed this trend.
5:51
And for context, I've done a hund- over 150 early-stage
5:55
investments in the past 10 years.
5:58
We've had about a dozen successful exits in the form of IPOs and
6:02
acquisitions by Fortune 50 companies.
6:04
We were one of the earliest investors into Instacart.
6:07
We, uh, have sold companies to Goldman Sachs, to Ford, to Caterpillar.
6:12
You know, you name your Fortune 50 company.
6:14
And the trend that I see over and over again across these entrepreneurs is
6:19
the smartest and most thoughtful ones don't just quit their job cold turkey
6:23
and go explore full time, contrary to what you see in pop culture and
6:28
how they describe Silicon Valley.
6:30
Oftentimes, Silicon Valley describes it as like, I just quit my job, I
6:33
just dropped out of school, and I go do this thing full time and keep
6:36
on working on it until it works out.
6:39
That's actually not what I've observed firsthand.
6:41
Oftentimes, what I see is who's working in a job has had, you know, 5, 10 plus
6:47
years of work experience where they've really deeply gotten to know that field,
6:51
gotten to understand the problems, and oftentimes experience that as a
6:56
personal pain point, or have served customers who are underserved and they
7:00
wish they could build a better solution.
7:02
And then instead of quitting their jobs cold turkey, they took the time to really
7:06
understand what the pain points are and how to fix it, and then they kind of
7:09
side hustled to building a solution.
7:11
Oftentimes, many of these intrapreneurs try to pitch this idea to their boss
7:16
and, you know, all the way up the command chain and try to do this in-house.
7:20
But oftentimes, as we were talking about earlier, Aidan, you know, the organism
7:24
rejects the organ and they can't do it in-house, and so they have no choice
7:29
but to go and do it as a startup.
7:31
And so I oftentimes see some of the smartest entrepreneurs being
7:35
reluctant because they wish they could have just done it in-house.
7:38
Most classic example of this is Marc Benioff, who was working at Oracle
7:43
before he had come up with the idea of doing something like a Salesforce
7:46
because he wished that he could have used that as solution himself.
7:50
He could have presented that to underserved customers that he
7:53
was dealing with at the time.
7:55
He didn't have the right infrastructure to do so, but when he left, he
7:58
actually got a check from the founder of Oracle to go do Salesforce.
8:02
So many other stories like this.
8:03
Jensen Huang, another classic example.
8:05
He and his two other co-founders were working full-time in the semiconductor
8:09
industry, understood all of the pain points, and then they would meet up
8:12
after work to just brainstorm, "How might we do another chip company
8:16
better, faster, cheaper, you know, in every single way?" And they had done
8:20
substantial amounts of the work before they decided to quit their full-time
8:23
jobs and make NVIDIA their full-time job.
8:25
And so we just see this over and over and over again.
8:28
And observing this from a startup perspective, I decided that I wanted to
8:32
do more systematic research to figure out, you know, i- is this true in the
8:36
broader world beyond the anecdotal case studies that I was seeing as an investor?
8:42
We ended up doing this research, and it turns out, in the world, there
8:45
are only 12 companies to ever be worth more than a trillion dollars.
8:50
Two of them are non-American, 10 of them are American.
8:52
The 10 American trillion-dollar companies, eight out of the 10 of
8:56
them were all started fractionally.
8:57
So only two where the founders were like, "Let's explore full time," and
9:02
they luckily landed on something.
9:04
But eight out of the 10 were all where it tied directly to what they
9:08
were doing before, whether they were in school or they were working full
9:11
time, and they had built substantial portions of things before they
9:14
left and, and did it full time.
9:16
And then we expanded this research to looking at Fortune 500 companies, and
9:20
it turns out that a quarter of all of them, so one in four Fortune 500 company,
9:24
they were all started fractionally.
9:26
And so I've, over time, developed a taste for fractional founders, and now
9:31
I love working with folks early on to just be a thought partner and figure
9:35
out, is this the thing to work on?
9:38
What is the most efficient way to test it, this hypothesis,
9:41
and run these experiments?
9:43
And, like, is this the right thing that you should go full in on and all
9:46
in on or, you know, should we test out other things before you decide to
9:50
quit your job and drop out of school?
9:53
you're gonna have , an influx of messages from people who listen to this
9:56
show . For, for people who do want to get in touch, where, is it LinkedIn
10:01
the best place to connect with you?
10:03
Um, I actually put all of my emails on LinkedIn as well, so
10:06
I
10:07
make it easy.
10:08
I'll link to that in the show notes as well, just to put that
10:10
straight up, 'cause you're gonna be exciting a lot of people over here.
10:13
But I wanted to come back to that.
10:15
T- so that idea of, that was really caught my attention when you were
10:19
giving a talk i- w- for MasterCard with me, and I really dived into that,
10:25
and I went back into, through the, all the research, and you saw so many
10:29
people, Steve Wozniak came out of HP.
10:32
They rejected his ideas he wanted to build for them.
10:36
And, you know, one of the things I see, and I'm sure you see this, is it's like, I
10:40
always think of that scene from Shawshank Redemption where he tunnels out the wall.
10:44
So the character who, who escapes, he tunnels out the wall, Andy Dufresne,
10:49
and then he puts the poster behind him.
10:52
And I always think that when, when somebody leaves like
10:55
that from an organization, the organization bad-mouths them.
10:59
So they're, like, going, "Oh, well, you know what? Aidan was,
11:01
he was crazy anyway, and he wasn't good for the company," et cetera.
11:05
And, and I often think it's because they don't want anybody else to
11:08
find your little hole and tunnel out after you, because they don't want
11:13
everybody leaving the organization.
11:15
But th- therefore, they should actually harness that energy, that
11:19
thinking inside the organization.
11:21
I, I don't know what if, what thoughts you might have on that
11:24
because, like, it's just such a shame.
11:26
I'm sure companies, when they see that, go, "Oh my God, we had that guy working
11:30
for us. We could have done it." You know, Eric Yuan in Zoom working inside Cisco.
11:36
There's so many of these people
11:38
I, I feel like is becoming increasingly a scarce resource these days, where
11:45
I think what you're describing is so spot on that in the past people would,
11:48
you know, put that poster back up.
11:50
Let, let's block the hole, prevent this leakage.
11:52
Whereas now I almost feel like I'm starting to see culture
11:57
evolve around this topic.
11:59
So as an example, we talked about Marc Benioff earlier.
12:02
I think that's a great example where obviously Larry Ellison, who founded
12:06
Oracle, took Marc Benioff under his wing and said, "Look, I think this is a really
12:10
talented individual. If, you know, w- we're bound to, to lose him, let's make
12:15
sure we have an equity upside in what he's about to build next." And I think the
12:20
reverse is also true for intrapreneurs.
12:23
The way I would think about it is, like, you have all this
12:25
talent within your organization.
12:27
If you don't give them a way to build inside, they'll leave, so you
12:31
should find a way to stay in touch.
12:33
But if you empower them and inspire them within your own organization,
12:37
then they get to help you build incredible things that perhaps you
12:40
don't necessarily have the internal forces, motivation currently to do so.
12:47
And so I, I don't see it as like there's a hole and, you know,
12:50
all of the talent leaks out.
12:52
Maybe I keep on thinking about it in terms of like this actually attracts some
12:57
of the best talent to your organization when they can see that, hey, if I wanna
13:01
do something that's ambitious, that there is a way and a platform for me to do that
13:05
There is a hole there was like don't tell a lot of people about this, but we
13:09
have a place for you in here." It's like this sandbox where you can play with
13:13
your ideas, "But don't tell everybody.
13:15
We don't, we can't let everybody work there." So I'm, I'm not gonna get through
13:19
your work, your newsletter, so I'm gonna move on to the opening line that I talked
13:24
about, this idea that, and this was data from OpenAI, that frontier firms,
13:28
which you might give us a definition of, they have so much, almost 10 times
13:34
the use of tokens as a typical firm.
13:38
And we saw this thing of token maxing, where organizations had encouraged
13:42
their own people to use the tokens.
13:45
They ended up using them to tick the box.
13:47
Then they were using them to do menial tasks that they shouldn't have been
13:50
using, and using up all their tokens.
13:52
But maybe you'll give us a view of this part of your newsletter
13:56
Actually ties quite well with our earlier conversation
13:58
because let me set some context.
14:00
So you're right in there are companies famously like Uber that literally
14:04
burned up an entire year of budget within, like, less than a quarter.
14:09
And so there was this phase in Silicon Valley where companies were
14:13
encouraging their employees to tokenax.
14:15
Just, just use the tokens and show me that you can use it.
14:17
But as is with life, most things you know, you evolve.
14:21
As you become more sophisticated with usage of AI, you become smarter
14:25
about how you actually use it.
14:27
And so what we're starting to see at these frontier firms, and really
14:32
you-- there are different ways that we can define frontier firms.
14:35
But usually when we talk about frontier labs, we're really just talking about
14:38
the OpenAIs and Anthropics of the world.
14:40
When we're talking about frontier firms, at least the research that I
14:42
was citing, is really firms that are in kind of the top ten percent of usage and
14:47
performance of being able to leverage AI.
14:49
These frontier firms, we used to think of, "Is this company good at using
14:54
AI?" And we kind of look at their usage, like, in terms of, "Do they
14:58
have access to using these tools?" For many firms, and it's still true that
15:02
around the world, most of these firms don't even have access to, like, the
15:06
OpenAIs and Anthropics of the world.
15:08
And so we used to evaluate, like, is this firm AI native?
15:12
Is this firm AI empowered by looking at do they give their employees access?
15:17
And what this latest research is telling us is actually access is no longer enough.
15:21
It's not just about giving employees access to the tools.
15:24
It's actually about looking at usage.
15:26
How deeply are they actually using these tools?
15:29
So it's great if 100% of your employees have access to the latest
15:33
tools from OpenAI and Anthropic and all the other AI companies, great.
15:37
But if they're not using it, then, you know, just giving
15:41
them access really isn't enough.
15:43
And what we're starting to see in this data is that these leading
15:46
firms do have employees that actually just already have learned how to use
15:51
AI efficiently and natively that.
15:53
And I think it has a lot to do with the earlier conversation we just had around
15:57
scarcity of talent because All of these firms are now looking at internal data
16:04
points and saying, "Hey, we actually think revenue per employee is a really
16:08
important metric to be tracking." And they're looking at it to say, "Well,
16:12
how much revenue can each employee I hire into my organization help bring?"
16:18
so the more AI and AI agents that they can use, the highest, higher leverage
16:23
each employee has, the more revenue they can generate for us as a firm.
16:27
And if my revenue per employee is high, that means my capital efficiency is high.
16:31
It means literally for every single dollar I spend on hiring a human
16:35
being, I'm getting max useful output.
16:38
And so perhaps a broader context of this is we've moved on from just
16:42
giving employees access to these tools.
16:45
We're now looking at depth and quality of usage.
16:48
So how much are they using these tools, and are they using
16:50
it in a very efficient way?
16:52
In fact, something that I'll talk more deeply about in the next newsletter
16:56
is this new metric that OpenAI very recently has started talking about.
17:00
And I have a number of good friends who are there, leading the teams, and one of
17:04
the things they start to look at is in terms of human-to-agent leverage ratio.
17:10
So for every human that is working, how many AI agents are you running
17:14
in the background to actually make this human more efficient?
17:17
And of course, um, I think what is coming next, this is a prediction.
17:22
The trend is we're seeing us evaluate companies going from access to usage,
17:27
and the prediction is I think we're gonna go from usage to actually
17:30
looking at max output per dollar.
17:33
So how do I maximize intelligence and useful output per dollar spent,
17:38
whether it's on humans or on agents?
17:41
But oftentimes the highest leverage is getting humans that
17:44
also manage additional agents.
17:46
You were talking in the newsletter as well about an index where you also see
17:49
this, the best performing startups tend to be the biggest and heaviest users of AI
17:54
100%, because the only ch- the only way that a startup has a chance at
17:58
going up against an incumbent is if they're somehow more agile, more
18:04
adaptable, and also more efficient.
18:08
Otherwise, there's just no chance that they have competing against someone
18:12
who is larger, who has had more time, who has more capital and resources.
18:15
And so I see startups really not token maxing.
18:20
Token maxing was just about using AI, but like really maximizing
18:24
intelligence per dollar they put in
18:25
the sponsor of the show is Kyndryl.
18:27
I mentioned that to you beforehand, and the leader of that, the guy who
18:30
actually looked at the innovation show for us was a brilliant guy,
18:33
an author called Ismail Amla.
18:35
And Ismail said to me before, he said, " we're seeing this shift where having
18:40
a big company and having lots and lots of people was the advantage before, and
18:45
the startup didn't have that advantage.
18:46
Now it's becoming a disadvantage because you have to almost up-skill all those
18:51
people or bring all those people with you on the transformation journey, while the
18:55
startup, if they're using AI, they're way more nimble and they can actually be so
19:00
much more efficient." If you even think about a meeting , like they can have
19:03
a meeting with a small team, get stuff done so much faster now than ever before.
19:09
What's, what are you seeing out there in that field?
19:11
I actually think that is so spot on because what I'm starting to see in the
19:16
world of startups is a couple of things.
19:18
For one, they definitely pay attention to that revenue employee per employee
19:23
number really upfront, and every single time before they hire, the question they
19:26
ask is, "Is this new employee I'm about to hire going to drag up my average or
19:31
drag down my average?" So they're super thoughtful about who they bring on board.
19:35
So if I think this person is gonna really substantially grow and scale
19:39
the business, then it makes sense.
19:41
But why would I hire a human who's less efficient than all the other
19:44
people that I already have on my team?
19:46
The second thing I'm starting to see that is kind of, uh, a borderline between a,
19:52
a trend I'm observing versus, like, a prediction I'm making is, I'm starting
19:57
to see some early startups do this.
19:59
I think it will become more of a trend going forward.
20:02
But what I'm starting to see is some startups are saying, "Look, instead
20:05
of a job opening and then looking for candidates to fill that hole, let's do
20:12
it the other way around, which is let's just find the smartest people out there
20:15
that have connections that I want access to, that can do things that we know our
20:19
organization will benefit from. Let's just hire that person into a general title."
20:24
Like, we are now starting to see this happen not only at OpenAI and Anthropic,
20:28
but for example, Microsoft is starting a job at this where they just call the
20:31
member of the technical staff, right?
20:34
So you just join the member as a member of the technical staff, we
20:37
plug you in, and then we figure out how we're gonna make most use
20:40
of all the talent and resource and relationship and network that you bring.
20:45
And so instead of plugging in, like, a missing hole, we actually bring something
20:49
that's more additive to the organization.
20:51
So I'm starting to see that conversation happen.
20:54
And because talent, at the end of the day, is increasingly m- more scarce,
20:58
especially people who can deploy AI agents effectively, startups are really
21:02
starting to think about that question.
21:04
thing I will say here is that I think when we look at any sort of situation
21:10
for startups where they need to attract talent y- it's not like a large
21:15
organization where you're attracting them based off of, you know, perhaps the
21:19
compensation package that you can offer.
21:21
It's not just the, the salary, it's the future equity, it's the
21:25
growth, the upside, the ambition that you now get to be a part of.
21:29
And so there is somewhat of a self-selection effect there.
21:32
But, I think the way that I hear startup founders talking about this is they know
21:37
very, very well that we are the last generation of leaders to be managing only
21:42
people Increasingly, they're seeing more and more employees manage people and AI
21:47
agents, and they can just be faster at adapting to that than an organization that
21:53
already has tens of thousands of people.
21:56
the same time, what I will say, though, is there's a huge advantage to an
21:59
organization that has already attracted tens and thousands of talent, 'cause those
22:04
are the people that startups are trying to, to tempt away from your organization.
22:08
So, if you find a way to inspire and empower those people that you already
22:12
have within your, your organization, that's already a massive win, 'cause
22:15
you're sitting on top literally the most valuable resource of people that
22:20
startups are trying to just build out
22:22
There's one thing that comes to mind.
22:23
I was telling you I was doing a two-day, four-session program
22:27
the last two days with the Enterprise Ireland here in Ireland.
22:30
It's called Peer Circles, and it's groups of companies with s- similar challenges,
22:35
similar opportunities so for example, we had med tech and life sciences together.
22:39
We had a CFO group.
22:41
We had R&D and innovation, and we had a family business circles.
22:45
We call them peer circles.
22:46
But one of the, some of the questions that came up were around, for example,
22:51
if I have my kids, it's like, what, I'm fearful for my kids' future.
22:57
And what, why I think it's really interesting to throw this to you
23:00
is 'cause you lecture to okay, I'm aging myself here, but to kids, so
23:04
undergraduates as well as postgraduates.
23:07
So you're not trying to tr- tr- like, if you have a company that's like, "Okay,
23:12
we, we have a revenue per employee metric," but they're gonna have, they're
23:17
gonna, they're gonna look for different skills from graduates, from, from new
23:22
employees, and I'd love you to share that.
23:25
For those people who are thinking that way, what are you encouraging
23:29
your own students to be doing?
23:31
100%.
23:32
I'm glad you asked the question because I'm not only an investor
23:35
and educator, I'm also a parent.
23:37
And so when I think about what ki- what world my kid will grow
23:43
up in, I'm often brewing over it.
23:45
Well, how does that then impact my role also as an investor?
23:49
What am I investing in so that that future is something that I wanna be a part of?
23:53
And what am I making an impact on as an educator so that, again, that's
23:58
shaping the world in a way that I think will be better for everyone.
24:01
And so when I think about, you know, what changes are currently happening,
24:07
kind of two sides of the coin.
24:09
One side of the coin is you know, that AI has been an incredible tool
24:15
in terms of automating a lot of the repetitive and boring aspects of
24:19
work so that kids nowadays graduating from, you know, college and graduate
24:25
school, they are expected to do more at a relatively more junior role.
24:30
For example as we were talking about earlier, Aidan I've done a lot of work
24:33
with EY, and one of the conversations I've consistently heard is, "Hey, if now
24:39
we don't necessarily need entry-level employees to do very basic things
24:43
like update spreadsheets and do more tedious work, then we can actually
24:46
bring them along to client re- meetings and build relationships earlier on,
24:50
have them do more at a younger age." So that's one side of the coin.
24:54
The other side of the coin is that it's easier than ever before
24:58
to graduate from school, but harder than ever to find a job.
25:01
Because the expectation for what you're able to bring to
25:05
a job is that much higher.
25:07
so how I'm responding to that is that I try to give my students as many real-world
25:13
exposure and experiences and project experience as early on as possible.
25:18
So that's partly why I think the fractional founder concept is so
25:22
empowering because you find what you wanna do in life by working on side projects
25:27
that you were genuinely interested in, where you weren't even compensated
25:30
at the beginning, but it was just a problem that you obsessed over, or this
25:33
thing that you really enjoyed building, and then that turns into, you know,
25:38
what sparks the rest of your career.
25:40
I, I oftentimes think that these projects are the best way to get you started.
25:44
And so one of the top things I encourage my students to do is build side projects.
25:49
It could be something that's not clearly what you're gonna do for the rest of your
25:53
life, but that's the best way to explore.
25:55
So there are negative forces, you know, that make it harder for entry-level
25:59
employees to get a job, um, that, where the bar is higher for them to
26:04
even be able to perform at that level.
26:06
But I think there are now also more tools that they can play around earlier on and
26:10
build side projects with that just make it that much more interesting for them.
26:14
And so I always encourage my students to, to try out as many side projects as much
26:19
as they can earlier on in their journey.
26:22
This is a computer science concept that I think is very true for careers as well.
26:26
I think there are two phases to careers.
26:28
There's the exploration phase, and then the exploitation phase where,
26:32
you know, you wanna make sure you explore the landscape, otherwise
26:34
you're stuck on a local maximum.
26:36
You explore so you can find what your global maximum looks like.
26:39
So like these are the things that I like, these are the things that
26:41
I'm good at, and like t- let's find where I'm the best at it.
26:45
And then once you've done that exploration, then you can get
26:47
really deep on that front and become genuinely one of the best at that.
26:51
And most students now I think are under a lot of pressure to almost skip past
26:56
the exploration phase, which is why there are people who go work in a job
26:59
for five years, 10 years, and then realize, "I, I don't wanna do this at all.
27:02
I want a complete career change or a complete career pivot," in their 40s.
27:07
I'm seeing this very regularly happen for folks in their 30s, 40s, 50s even,
27:11
that come back to executive programs at Stanford and try to figure out,
27:15
like, "How am I gonna do this career pivot?" So I think there's a lot
27:18
of merit to doing this exploration early on through side projects.
27:22
Even, Whether you're in school or you're already in a full-time job,
27:25
try to build these side projects.
27:26
Figure out what you like, what you don't like, what you're good
27:28
at, what you're not good at.
27:30
And I think that will be way more informative than, you
27:33
know, anything else you can do.
27:35
And so I tend to be an optimistic person by nature, as you can tell.
27:40
Otherwise, I wouldn't be an investor because I would just think all startups
27:43
would die if you look at the data.
27:46
so I try to take the more optimistic side, when I look at things,
27:51
which is what is something I can do that makes a positive impact?
27:54
So as an investor, I'm trying to invest in companies that
28:00
are built upon relationships.
28:03
Not just that it's a fundamentally better tool, but that there's a really important
28:06
human element where it's rooted in humans and our connections to each other.
28:11
And as an educator, I try to encourage my students to go build
28:16
that human relationship, to try to build that larger network earlier on.
28:20
I think oftentimes networking can come across the wrong way.
28:23
Like sometimes people hear that and they just think, "Oh, that like
28:25
feels kinda ick, like that's not for me." Maybe the way I like to think
28:29
about it is like you're growing up with people around you, right?
28:32
Like no matter what age you are, oftentimes I talk to people in
28:36
their 60s and they're like, "I'm still trying to figure out who I
28:38
wanna be when I grow up," right?
28:39
So I feel like it's about surrounding yourself with people that inspire you,
28:45
that like pique your interest, that you have interesting conversations with,
28:48
and you build deeper relationships with, get to know better, and then
28:52
eventually you grow up together and, um, and then you expose yourself to
28:55
a whole world of really cool things.
28:57
I love it.
28:58
I love it.
28:58
And I just wanted to w- say one more thing on the, the education aspect,
29:02
which is, I'm sure you see this, and I know Stanford came out with a, a policy
29:07
recently that many other organizations are looking at, which is students using
29:11
AI to write their papers, et cetera.
29:13
And it's something that I see as a real threat to human cognition in
29:19
general, as in, I would say to many organizations, they're using AI to write
29:24
policies, then they're getting AI to summarize those policies and give them
29:29
questions to ask about those policies.
29:31
The board then read , use AI to summarize the paper that came out of
29:36
those policies, and they ratify it, and no human has been in the loop at all.
29:41
But beyond the human being in a loop, nobody has gone through the
29:44
struggle that is absolutely required to learn or to write or to understand.
29:50
And for me, I say this to my kids, like, for example, I was like, to my older
29:54
son who's, who was a bit of a rebel when he came to school, I was like,
29:57
"Listen, it's not about going to school.
30:01
It's about learning to be able to struggle through something, some type of content,
30:06
and then deliver an answer i- on time to somebody who wants that answer." And it's
30:13
that, that's a me- that's a mechanic.
30:15
That's a journey.
30:16
That is life.
30:17
And sometimes that means doing stuff that you don't wanna do, because everything,
30:20
every job has aspects that it doesn't want, that you don't want to do.
30:24
And I just wondered what your thoughts were on, what your advice were to your
30:27
students, 'cause I have this myself, and you can spot it a mile away.
30:31
You can spot LinkedIn posts that are written by AI that clearly ha- is like,
30:35
it shouldn't deserve a, a LinkedIn post but is made expand into a formula.
30:41
And when people do that, they're not learning and
30:44
they're not retaining knowledge
30:46
It's so true.
30:47
Um, I, I think everything you described is true because when you have a lot of
30:55
poor quality content flood the internet, and now we have more AI-generated content
30:59
than human-generated content by, like, a wide multiple, that, like, it decreases
31:04
the quality of the human-written content.
31:06
For example, like, I've been an avid user of em dashes before AI existed,
31:12
and now that AI uses so many em dashes, like, my messages look like they're
31:15
written by AI even if I wrote them.
31:18
So it decreases the quality of even the human-written posts 'cause you
31:22
look at it and you're like, "Wait, maybe this was AI generated." And
31:25
then, like, even the human-written work doesn't get as much recognition or,
31:29
you know, uh, attention than before.
31:32
And so I think it's very easy to end up in this kind of negative spiral where,
31:36
like, there's no genuine human in the loop, and it's just my agent will talk
31:40
to your agent and, you know, that's it.
31:42
And I think this is partly why I see so much of the value of education of just
31:49
going back to that in real life component.
31:51
Like, in, like, the whole movement in Silicon Valley around I- IRL, in real
31:57
life, I think is a thing because people are realizing, like, anything online,
32:01
that can all be AI, but when we show up in the same room, have a dinner
32:05
discussion sharing, like, breaking bread together, sharing a meal, and,
32:09
like, that is not replaceable by AI.
32:12
And so I think a huge value to the educational experience is partly
32:19
what you described, like struggling through something, delivering an
32:21
answer, and, like, understanding how to work under pressure and all of that.
32:25
I do remember when bringing calculators was controversial right?
32:31
Like, that was a whole thing, and I think there are some ways that
32:35
the current era of AI that, where, like, hey, it, it can be a tool.
32:40
At the same time, I think AI can automate things to a whole new
32:43
level, unlike calculators, where, like, humans can genuinely get
32:46
away without much thinking at all.
32:49
And so that's why, on the other hand, I think it's so, so, so important that
32:52
college is not just about learning and the process of learning and
32:56
struggling, but also about, like, these are all the people that are around
32:59
you and building those relationships.
33:02
I think honestly that is the most important thing that is not
33:05
replaceable by any sort of technology.
33:07
And if you look at most of the real world, industries still
33:11
thrive based off of relationships.
33:13
So even though I invest in fractional founders who are building software
33:17
startups that sell B2B Very few of them are because like, "Hey,
33:23
look, like you are the best, the only, the, you know, whatever."
33:27
Like of course those numbers matter.
33:28
But oftentimes it's, "I trust you.
33:30
I am buying your software because I know when things go wrong, that
33:34
like I'm going to come to you and you're gonna help me out.
33:37
You're not gonna leave me stranded in my job and my role.
33:39
Like, you're gonna make sure that I'm okay." I think that's how a lot of
33:44
the world still functions, and I'd like to believe that even with AI that
33:47
that will still be very important.
33:50
I do think we'll need better AI governance for sure to prevent a
33:53
situation like you described where there's literally no human in the loop.
33:56
But when I look at students, I always tell them, like there are students
33:59
who prematurely drop out of school, and there are absolute valid examples.
34:04
Like, I've had students who, whose startup, was already making like
34:08
$13 million in annual recurring revenue, and I'm like, "Of course
34:11
you should do this full time.
34:12
Of course you should be dropping out of school." then there are also
34:15
other examples of students who wanna drop out of school, and I always
34:18
remind them like, "Hey, know, AI tell you answers to everything.
34:22
So you're not just in school for like the learning piece of it.
34:25
It is being in this environment where you get to learn and grow
34:28
with other people around you.
34:29
Let's not forget the value of perhaps an acquaintance today that could become
34:34
someone who's really important in your life later on." I am now working with a
34:39
venture partner, who I had met fleetingly back in undergrad when we were at the
34:43
bookstore just buying the same textbook, and we struck up a conversation,
34:47
became good friends, and stayed in touch for literally well over 15 years
34:51
before we ever did any work together.
34:53
And so I think it's just the depth of relationships that compound over time.
34:58
. I was mentioning about those groups that I worked with yesterday and, and we mined
35:02
them for their worries or their concerns when it came to usage of AI, and one
35:06
of them was some of the companies had genuine IP where they'd created really
35:12
globally unique content or u- u- unique formulas for their, for their output,
35:18
for their services, for the product.
35:20
And some of them were like, "Well, if I put it into an AI like ChatGPT
35:27
or Claude, where else is it going?
35:29
'Cause I, I know it says it's unique to me and it, it won't go anywhere
35:33
else, but do I really trust it?" And it's a nice segue to something that
35:37
you write about where you have the tech companies actually competing and bidding
35:42
for proprietary data, and it wasn't data that is openly available, it was
35:47
data that was from bankrupt companies.
35:49
And in this case, you talked about Spirit Airlines.
35:51
I'd love you to share tho- those two flip sides of the coin, the story of Spirit
35:55
Airlines and the, the bidding war, but then the concerns over enterprise data.
36:00
Absolutely.
36:01
So, As most folks know by now, Spirit Airlines had declared bankruptcy.
36:05
And what became somewhat surprising to people was that their most
36:09
valuable asset wasn't just, like, what terminals they had access to and what
36:14
airports, but rather their literal, literally decades of operational data.
36:18
This included things like employees' emails, anonymized, like, customer
36:22
data, operational data, literally how they were running processes.
36:26
So if you think about the world, um
36:31
95% plus of the world's data is not available on the
36:34
internet because it's private.
36:36
It's stored on within enterprise databases.
36:39
And that is gold for these AI companies because if you think about how AI works
36:45
and by the way, if this is too, too basic for you, like skip 10 seconds ahead.
36:49
But like AI basically is made of three things, right?
36:52
It's data, it is compute, and it's algorithms.
36:55
So very analogous to the human body.
36:57
Data is like all the knowledge that it learns.
36:59
Compute is like the physical brain, like the actual hardware compute
37:02
storage where things are stored on.
37:03
And then algorithms is like literally what we were talking
37:05
about, learning how to learn.
37:07
So some people are more visual learners, they're better at visuals,
37:09
and other people are like better with tac- like tactile, right?
37:13
So it's like the different models.
37:14
Different models out there.
37:16
There are some that are really good at images, some that are really good at
37:18
videos, some that are really good at math.
37:20
So exact same analogous situation to humans.
37:24
And so when you think about AI you can teach an be really good at, let's say,
37:30
looking at a particular kind of data.
37:31
So you're training its algorithm.
37:33
You're storing it on the most efficient compute on state-of-the-art chips.
37:37
But you can't do any of the training unless you have knowledge, right?
37:41
This is what we go to school for.
37:42
This is what we live for.
37:44
We absorb a lot of knowledge, and AI precisely needs that kind of knowledge.
37:48
So imagine trying to teach your kid to go to school with just by
37:53
watching stuff on the internet.
37:54
Like you're not gonna have a very thoughtful, smart child, even though
37:58
the internet is full of information.
38:00
If you wanna have-- like raise a kid or teach a kid, or in this case, teach
38:05
an AI how to be a really thoughtful adult that can work in the enterprise
38:09
environment, you need to have it learn in an enterprise setting.
38:12
It needs to be able to have access to like how do people talk to
38:14
each other in a corporate setting?
38:15
Like what data are they transferring?
38:17
How do operations and procedures and workflows actually
38:20
function inside a company?
38:22
And this is why this kind of data is so valuable to these AI companies.
38:26
So Spirit Airlines declare bankruptcy.
38:28
All these companies wanna bid for the data.
38:31
The latest that we're aware of is Google sent in the largest bid for that data
38:34
because they wanna be able to use that to train their AI so that the AI can
38:38
do better in an enterprise setting.
38:40
So that's the first side of the story.
38:42
The second side that Aidan, you were talking about is, well, like there are
38:46
all of these people that are sending their information off to AI, and it's
38:48
like, are they using it for their own training purposes, et cetera?
38:52
of all, if you're not paying for your ChatGPT, Claude, Gemini, whatever
38:57
it is, they're definitely using your data because when you don't
39:01
pay for it, you are the payment or your data is the payment, right?
39:05
Now if you pay, there are more stipulations that you can read through
39:09
it doesn't fully protect you, so you need to read through all kind
39:12
of the details to figure out what exactly are you trying to protect.
39:16
Typically, now, if you have an enterprise account, it's much more
39:19
clearly laid out, and you can negotiate upfront that, hey, like none of my
39:23
proprietary enterprise data will be used to train models that then can
39:26
benefit other companies, for example.
39:28
But one thing to keep in mind is that this is actually an emerging conversation,
39:33
maybe a topic that I'll cover in a future monthly note as well, where
39:39
companies are saying, "Given that these proprietary workflows are my most valuable
39:43
resource, why would I just give it away to someone else to train the data?
39:47
Why don't I figure out how to train the data and actually own all of it?
39:52
Because that's literally my most valuable thing, like how I do
39:55
business, how I manage people, how my people talk to each other.
39:58
So why would I just give it away to the equivalent of a frontier
40:01
lab and have them own that?
40:03
Why don't I try to do that in-house?" Now, for anybody who's spent time in corporate,
40:08
you'll know that sometimes it doesn't work for a lot of reasons due to sometimes
40:12
it's re- resource constraints, sometimes it's politics, sometimes it's lack of
40:15
will, like tons of different reasons. I definitely see an increasing move
40:20
from companies that are saying, "Look, let's figure out how we can do a healthy
40:24
balance where maybe we use one of the open-weight models for just basic tasks.
40:28
These are models that are cheaper and open-sourced there's no proprietary
40:33
data for basic workflows there.
40:35
But then for the more proprietary stuff, maybe we wanna retain that, 'cause that's
40:39
literally my most valuable resource.
40:41
Wanna keep all of that data within the company, and then be able to
40:44
train a super specialized model that is really good at what I do.
40:48
And it could be in a specialized sector like legal or like financial
40:51
services, and it could be even more niche, like just credit or just lending
40:55
within financial services." And so I'm definitely starting to see a shift there.
41:00
But that will require a lot of human capital, a lot of political
41:05
capital, a lot of will to, to make this all happen within a company.
41:08
We are not gonna get near the newsletter.
41:11
I thought we'd get through it in half an hour.
41:13
Not a chance.
41:14
Maybe we'll just share.
41:15
There's stuff in there.
41:16
There's your future-oriented trends, things that you see coming down the line.
41:20
But let's share your tool, so the tool that you love.
41:22
This is from the August newsletter, and hopefully we'll get together
41:26
and do the September one.
41:27
But let's share this one.
41:28
This, I thought this was really cool
41:31
in my monthly note, tried to share three trends, two thesis, which are kind of
41:36
predictions of the world, and one tool.
41:38
And, um, uh, I try to keep the newsletter really concise, so Aidan, thank you for
41:43
being so patient and letting me ramble on about each of these points because
41:46
a lot of these are things that I wish I could have included in the note,
41:49
but I just don't have the space to do so and had to keep it really concise.
41:54
Tool I usually feature something that either I've worked with a founder or
41:58
crossed paths with them somewhere and just found what they do to be really valuable.
42:02
And the tool that I just talked about is this startup called Woolf.
42:07
Joshua, the founder of Woolf, actually was a longtime professor, um, in the UK.
42:13
He was, at Oxford for the longest time, I believe as a philosophy professor,
42:17
and then also served on the board.
42:20
He has done some fascinating work, since he decided to leave and start the startup.
42:26
Also somewhat of a fractional founder story, where he had experienced this
42:29
pain point working in the administration of the university and then said,
42:33
"Look, I need to build a better tool." So what he started by building
42:36
initially was just, look the university administrator to student ratio is way off.
42:42
I think in the US there's some even worse numbers.
42:45
Don't quote me on this.
42:46
I can Google and find out the exact ratio, but like I think the Yale administrator
42:50
to student ratio is double digit.
42:52
It was like 30 administrators to a student or something like that.
42:57
Uh, They do a lot of work, administrators do.
42:59
Like there's all of this paperwork and reporting that you have to keep up to
43:03
date with the regulators so that you can stay accredited as a university.
43:07
And so Joshua's starting point with Woolf, named after Virginia Woolf, was
43:12
just let's use technology, including AI, to automate a lot of the work
43:16
to make it easy for a university to become an accredited institution.
43:20
So there are a lot of educational programs out there like coding boot
43:24
camps or like architectural programs or even startup programs like the
43:27
one that you and I do that very much could and maybe should be accredited.
43:32
For example, you and I both have done this, where we teach on campus and,
43:37
all the students that are in class of course receive credits for that class.
43:40
And then we literally s- like let's say take one block off campus, teach the
43:44
exact same content, and students who take that class get no credit whatsoever.
43:48
That doesn't seem to make sense at all.
43:50
So at first he was just building this tool to make it easy for educators and
43:55
institutions to make it easy to create content that is accredit- accreditable.
44:01
And then eventually he started building this, to make it available to the
44:05
student and corporate side as well, where for students, what if you had a
44:08
lot more flexibility and you could take a couple classes at Oxford, a couple
44:11
classes at Cambridge, a couple classes at Harvard, a couple classes at MIT,
44:15
a few more from Berkeley and Stanford.
44:16
And you DIY that entire thing into a degree.
44:22
Makes a lot of sense because you've just taken all these accredited courses.
44:25
Universities didn't historically offer you the flexibility to do that, but now
44:29
we have technology to track your learning across all these different campuses and
44:32
actually stitch it together into a degree.
44:35
Similarly for enterprises, when they need to do employee training, if I
44:39
need to do AI training, why shouldn't you get credits towards a degree?
44:43
You can absolutely earn a master's in AI, or you can learn-- you can start to earn
44:48
towards an MBA in business administration um, as long as you're taking all
44:54
the right courses along the way.
44:56
I think we're seeing this shift where in the past we transitioned from a role
45:01
of a student to a role of an employee, and, like, it was a one-way street.
45:06
That's not the case anymore because things are changing so quickly, where
45:09
oftentimes both of those roles coexist in our lives, where we're employees,
45:13
but we're constantly learning.
45:14
In fact, we're expecting to be learning on the job.
45:16
That's part of the fun of being on the job.
45:19
So if you're consistently learning, why shouldn't you
45:21
receive credits towards that?
45:22
So that's Woolf-- what Woolf does.
45:24
It started off just in terms of the administrator side, and now it
45:27
offers that flexibility to students who might wanna take courses across
45:31
multiple universities and still graduate with an accredited degree.
45:34
Offers that flexibility to companies that provide training to their employees to
45:38
say, "Look, you can actually train your employees and give them an MBA degree
45:42
in the process as well." one of the top things that they had just published
45:46
was a piece where they were writing about how more employees were spending
45:50
money on AI training, more so than ever before, but the completion rate for
45:54
these self-paced courses are so low.
45:56
They're sitting at, like, nearly ten percent, only about twelve percent or so.
46:00
And so they were doing more investigation to figure out what
46:02
will actually incentivize employees to finish these training degrees.
46:06
And it turns out that accreditation is one of the top incentives, because wanna
46:11
know that I'm working towards something.
46:13
Like, I'm not just doing this random course that I'm required by my employer,
46:18
but that, oh, like, this will add up towards, like, a graduate degree.
46:21
Like, I would love to have that, right?
46:23
So, um, the tool that I just shared with everyone.
46:27
Whether you're an employer or an-- you're employee, it's a great platform
46:30
to check out because now have partnered with Accenture to offer the first time
46:36
ever where you can do an MBA degree online and you pay on a monthly basis.
46:40
So you work towards the degree at your own pace.
46:44
You don't have to do this exorbitant tuition at the beginning and then
46:46
get into a lot of student debt.
46:48
You can actually work in your full-time job, do a part-time degree,
46:51
like MBA with specialization in AI, and then work towards that, so
46:55
you can do both at the same time.
46:57
Brilliant, brilliant.
46:57
Mic drop.
46:57
Brilliant work.
46:57
I,
46:59
I'm glad you cleared up w- it was Virginia Woolf, 'cause I thought it
47:02
was the Woolf from Pulp Fiction, the guy that cleans up all the dead bodies.
47:06
I thought it was the fixer.
47:07
It was the whole idea of being a fixer.
47:09
I watch too many movies, obviously, Shuo.
47:12
It's been an absolute pleasure, and I know you're under time pressure.
47:14
L- you're, you just had a board meeting.
47:16
I got you in between the board meeting and your next appointment.
47:19
Very, very grateful to have you on the show.
47:21
Where can people find that newsletter?
47:23
I will link to it as well, but maybe you'll tell us
47:26
Well, thank you so much for having me on the show.
47:28
The Monthly Note is at shuo.iovc.io.
47:33
Can remember my name as shushing someone and saying, "Whoa, she's awesome," Shuo.
47:38
So it is S-H-U-O.I-O-V-C.I-O
47:43
Fantastic.
47:44
It was an absolute pleasure having you on the show.
47:46
Shuo Chen, thank you for joining us.
47:48
you so much, Aidan
47:50
our sponsor, Kyndryl, who runs and reimagines the technology
47:54
systems that drive advantage for the world's leading businesses.
47:57
With a unique blend of AI-powered consulting built on unmatched managed
48:01
service capability, Kyndryl helps leaders harness the power of technology for
48:06
smarter decisions, faster innovation, and a Lasting competitive edge.
48:09
You can find out more about Kyndryl at K-Y-N-D-R-Y-L, kyndryl.com.