Latest / AI for Founders with Ryan Estes / “Just Use AI” Is the most dangerous advice in tech

Transcript

“Just Use AI” Is the most dangerous advice in tech · AI for Founders with Ryan Estes

0:00
What if the biggest AI risk in your company isn't
0:02
hallucinations, it's that you're quietly turning your team into
0:05
button clickers? Because right now, everyone's
0:08
doing AI, but most companies are just buying the same tools as
0:11
their competitors and calling it strategy.
0:13
Jeff Gibbons from Human Machines is obsessed with the real
0:16
advantage. Not better prompts, better
0:19
collaboration between humans and AI.
0:21
And here's a novel found use case.
0:23
So maybe you need an AI collaborative coach that watches
0:26
how your team uses ChatGPT and Gemini and Claws then flags when
0:31
you're outsourcing judgement, missing the real questions or
0:33
wasting time switching tools. So a question if you could test
0:37
300 ideas before lunch, see which ones actually work in the
0:40
market, and only apply human judgement at the moments that
0:43
truly matter, how would you redesign your week, your team,
0:46
or your entire company? Welcome to AI for founders.
0:49
I'm Ryan Estes. If you're a founder, a product
0:51
leader or marketer trying to win in the next 12 months, this
0:54
interview is for you. And if you want your brain
0:57
upgraded weekly without becoming a robot, go to the newsletter at
1:01
AI for founders.co. It's the newsletter for founders
1:04
who love speed, hate fluff, and enjoy stealing unfair advantages
1:07
in public. Subscribe and I'll send you the
1:10
good stuff, the playbooks, the prompts, the shortcuts, and the
1:12
occasional existential crisis. That's the newsletter at AI for
1:16
founders.co. And if this helped, please leave
1:18
a review of the podcast like you're tipping your future self
1:21
with five stars. Now, Founders, quick question.
1:24
Are you still hoping customers magically discover you or you
1:27
ready to be everywhere your buyers are already paying paying
1:30
attention? Kitcaster helps founders and
1:32
executives get booked on top. The kind your customers actually
1:35
listen to, not the My friend started a show circuit because
1:40
podcast guesting is the cheat code for trust.
1:42
You borrow the host credibility, tell your story like a human,
1:46
and turn your expertise into inbound leads, partnerships, and
1:49
sometimes even funding. And Kick Caster makes it stupid
1:52
simple. We handle the research, the
1:54
targeting, the pitching, the follow-ups, and the booking
1:57
logistics. You show up, drop value, and
2:00
sound like the main character. If you want consistent, high
2:03
quality podcast appearances without it becoming like your
2:06
second job, go to kickcaster.com and talk to the team that's Kick
2:10
Caster by Molburst. Jeff, thank you so much for
2:27
joining me. Welcome to the show.
2:28
Good to see you. Good to be here as well.
2:30
Absolutely. So as we get started here, just
2:33
give us a broad overview of your work because I think it's going
2:36
to be very important to the listeners of this podcast.
2:39
Yeah, sure. Absolutely.
2:40
So, yeah, So I'm, I'm Jeff Gibbons, I'm a, I'm a
2:44
consultant. I've worked for the last nearly
2:45
20 years working with companies on new product development,
2:49
innovation, new business models. And over the last three years
2:53
roughly, I've been working on this kind of wild ride around
2:57
generative AI around how do companies take advantage of that
3:01
and use that to change the way that they work together, the
3:04
type of value they can deliver to their customers and their
3:07
consumers. And obviously, you know, like
3:10
other people, been on my own personal journey of discovery
3:12
around that too. And, and I, I run a consultancy
3:16
called Human Machines, which is really all focused on how do you
3:19
actually see AI not as a technology problem, but as a
3:23
problem of human AI collaboration and work with
3:26
companies around how to redesign the way that they work in order
3:31
to do things that weren't possible before.
3:33
We had just humans and just AI tools doing themselves, but
3:37
actually reinventing the way that people work with humans and
3:39
AI together. Amazing.
3:42
And kind of your trust bar on the website
3:45
whichishuman-machines.com is some gigantic companies.
3:51
So when you're working with kind of enterprise companies that are
3:55
characteristically they move slow probably for good reason.
3:58
They also have a lot of compliance and we're we're kind
4:01
of a juxtaposing that with AI technology which iterates and
4:04
move very, very quickly. So when you're thinking about
4:08
like generative AI and kind of what would be good applications,
4:11
what kind of tools and what kind of applications are you thinking
4:14
about for companies, for enterprise companies?
4:17
Yeah, Yeah. So it's a great question.
4:19
And so, yeah, So some of the examples of the companies I've
4:21
worked with the last few years, it's, you know, it's companies
4:23
like the Coca-Cola Company, it's it's Nestle, it's MasterCard,
4:27
Pfizer, and they're all companies that are in, you know,
4:30
large, larger enterprises, often in highly regulated sectors.
4:34
And I think there's 22 lenses. Look at it. 1 is that they're
4:38
looking at enterprise AI as an opportunity to get ahead in some
4:42
instances. And that, you know, there's a
4:43
real kind of desire to not be left behind, which which
4:47
motivates people a lot. And I think my sense from the
4:50
last couple of years is that this FOMO is really the main
4:53
investment driver of a lot of what's going on around AI and
4:57
the economy today. But the second thing is
4:59
obviously to manage that risk, a lot of what companies are doing,
5:02
which I think is the right approach is rather than changing
5:05
exactly what they do in sort of in their external facing
5:09
services and experiences, they're using it on the inside
5:11
first to actually reinvent the way that they work and use that
5:15
to create some kind of competitive advantage.
5:17
And so that's the type of work that I do with companies is, is
5:20
not just to look at how do you drive greater efficiencies
5:23
through AI, but how do you actually reinvent the way that
5:25
you work. And that's really where I see
5:28
the main competitive advantage comes from with AI.
5:30
So it's, you know, there's lots of tools, lots of technology,
5:34
everyone's selling you agents that are kind of generically
5:36
built. And every company, you know,
5:39
every large enterprise is getting sold the same thing.
5:42
And that means that, you know, it's obviously opportunity to
5:46
not, you know, not lose steam and be left behind.
5:51
But also it's hard to figure out how you differentiate if you're
5:53
just buying the same technology and same tools and same data as
5:57
all of your competitors. And so a lot of the way that I
5:59
do is actually figuring out how do you redesign the way that you
6:02
work in order to develop a competitive advantage.
6:05
And my, my hypothesis is that over the next few years,
6:09
reinventing your work flows, reinventing your work and your
6:11
teams to integrate the best of what humans and AI can do
6:14
together is where the real competitive advantage lies.
6:17
It's probably one of the hardest things to do, but that's that's
6:20
why it's hard to copy. You bet.
6:23
I mean, pardon me, you could speculate that like the entire
6:27
growth of the stock market right now is all based on that value
6:30
alone. Yeah, that's, you know, yeah,
6:32
totally. Yeah.
6:34
I mean the the expense of you know, whether it's cash or it's,
6:40
it's energy for AI is exorbitant.
6:42
And you know, you know, open AI doesn't necessarily have like an
6:46
operating business model or are they a business or are they a
6:49
non profit? We don't really know.
6:51
So truly it's the value being created by optimizing workflows.
6:55
And, and I'm really curious, you know, like within these
6:58
companies, if it's, you know, generative AI.
7:01
So I'm thinking like copy generation, artwork, you know,
7:07
decks, marketing collateral, videos, these kind of things.
7:12
How are they incorporating some of these tools or how are you
7:15
kind of instructing some of these companies to to kind of
7:18
redesign like tactically, like what are the nuts and bolts of
7:22
kind of what they're implementing on a daily basis?
7:25
Yeah, sure. So, so I think if you're a
7:27
logic, you know an executives at a large organization, you're
7:29
getting pitched all the time by individual vendors for point
7:32
solutions. And typically what happens is
7:35
that those vendors are, you know, doing a really good job at
7:39
automating the way that a lot of different companies are working
7:43
right now. And so there's, there's a value
7:46
to doing that. And I think it's one of the
7:48
easiest ways to deploy AI within your business in a short time
7:51
frame, right? Because you can take, take a
7:54
tool that kind of matches how you work today.
7:55
And you know, you can, you can adopt it.
7:59
I think what I have tried to do with companies I work with is
8:02
actually think about, let me take a step back.
8:05
Let's look at all the different amazing individual point
8:08
solutions that are available and figure out, well, if we can do
8:11
all these things in one, should we do the same process or should
8:14
we actually organise the business differently, right.
8:17
So if you take, if you take the example of like, you know,
8:20
content generation, you mentioned as an example, right?
8:22
There's, there's a lot of content tools out there that a
8:25
content marketing team could, could run with.
8:28
And you know, if you took the same approach that you typically
8:32
do of like, you know, mapping out all the steps in the
8:34
process, you could probably stitch together 3 or 4 different
8:36
tools that would do everything that a content marketing team
8:39
would do right? In terms of picking up on
8:41
signals, in terms of generating drafts, in terms of reviewing,
8:46
like generating visuals, whatever it might be.
8:49
The approach that I, I like to take is, is how do you actually
8:52
redesign that workflow based on the fundamental outcomes you're
8:55
trying to achieve by figuring out what's possible with AI that
8:59
you couldn't do before, right? So, so rather than having a sort
9:03
of human process where you go step by step in a linear fashion
9:06
of figuring out, well, what are the things we want to stand for?
9:08
What are the things we want to know?
9:10
How do we draft it? How do we have a human review in
9:13
place? How do you actually redesign
9:15
that so you can have fewer steps in the chain?
9:17
You can actually be generating hundreds of different options
9:21
for something and then using synthetic testing or synthetic
9:24
data to narrow what you should do.
9:26
And then you actually use the market as a feedback loop into
9:29
how you actually put out content.
9:31
And you could be publishing a lot more than you ever would as
9:34
a human team. And you're using the market to
9:37
test out what you should be doing more of?
9:40
Yeah, yeah, I, I, I get it. You know, it's it's so do you
9:44
find yourself using like N8 N these kind of kind of automation
9:48
kind of tools a lot in the work? Yeah, it could be any end
9:52
depends on sort of the, the enterprise stack that that the
9:55
teams have. And I think the, the key thing
9:57
to understand is that like so I've, I've done quite a bit of
9:59
work, for example, in product innovation at large companies,
10:02
right? And so there are, you know,
10:05
established design thinking processes that they have, which
10:09
are all designed around human limitations.
10:11
All right. So if you think about a product
10:13
design team, for example, they will have, you know, a
10:17
structured process the the thinking through things in
10:20
logical steps, right. I'm going to figure out what are
10:23
the top 6 or 7 customer needs, so on and so forth.
10:27
And then with each of those, I'm going to figure out what are the
10:29
top 6 or 7 product ideas that can do.
10:30
And I'm going to do some testing.
10:31
With humans, you sort of run through these steps, but they're
10:34
all designed around the limitations of the human brain
10:38
and like how we can think about things, right?
10:39
So that design process is usually based around what, how
10:43
many post it, what, how, how many post it notes we've put on
10:46
a wall and how many different concepts can a human think about
10:50
at any given time, which is like, you know, 6 to 10 things
10:53
at once. And so you could automate that
10:55
process. But the other thing you could do
10:58
is we'll say, actually, how about we, how about we actually
11:02
scour the earth and figure out 300 product project needs.
11:06
And how about we look at identifying 400 product ideas
11:11
and testing those against all different criteria and then
11:14
narrowing our focus as a human team on the six or seven things
11:17
that we've stress tested to offer the best appeal, things
11:21
like that. Yeah, Yeah, you you mentioned
11:26
post it notes and I kind of smiled because I still see
11:29
people do a lot of business on post it notes, you know, and if
11:32
that works for you, I think that's that's great.
11:35
You know, and often times, you know, it just becomes to habits.
11:40
And when when I see kind of companies, you know, I'm working
11:43
mostly with start-ups. So these people are looking for
11:45
tools, looking to move fast, speaking to dozen start up
11:49
founders all the time, dozens and dozens of them.
11:52
And often times that the FOMO you talked about is like not
11:55
knowing what they don't know. It's like, oh man, the
11:57
technology is coming so fast that it's like, do I make a big
12:01
investment on this side? Or really if I wait three
12:04
months, does or do the tools I already have have the capacity
12:08
for what I'm looking to do? You know, so I, I can see that
12:12
kind of being a challenge as well, kind of a analysis
12:15
paralysis almost, you know, you know, I, I was doing a little
12:21
bit of research and I, I saw a great quote.
12:24
It said or you said the future belongs to human AI
12:27
collaborations, not humans or AII think the audience of this
12:34
this podcast understands that deeply, you know, but working
12:38
in, you know, large enterprise, a lot of this is coming on
12:42
really fast. What do you think most of these
12:43
leaders still misunderstand about that collaboration.
12:49
Hey, everybody, you know what's crazy?
12:50
I've been sitting on a gold mine this entire time and I didn't
12:53
even know it. My Gmail inbox, literally
12:56
thousands of warm contacts who already know and trust me.
13:00
But here's the problem every founder faces.
13:02
How do you turn those relationships into revenue
13:04
without feeling like a sleazy salesperson?
13:07
That's exactly why I've been using Warmstar dot AI and
13:10
honestly, it's been a game changer for my business.
13:13
This AI platform connects your e-mail, analyzes all your
13:17
existing contacts, and helps you identify who in your network
13:20
might actually be interested in what you're doing.
13:22
Then it writes personalized outreach messages that don't
13:25
sound robotic or pushy. The best part?
13:28
You review every single message before it goes out.
13:31
No spray and pray nonsense. This is relationship based
13:34
selling done right. My listeners know I we recommend
13:37
tools I actually use, and Warm Start has literally helped me
13:40
reconnect with people I'd lost touch with and turn those
13:43
conversations into real business opportunities.
13:46
If you're a founder struggling with outreach, check out
13:49
warmstar dot AI and use code AIFF to get $50 off your first
13:54
month. Trust me, your existing network
13:57
is worth way more than you think.
14:01
This episode is brought to you by Moeburst, the AI first growth
14:05
engine trusted by iconic startups like Calm, Robin Hood,
14:09
Reddit, Uber, and Google. Moeburst is a digital, full
14:13
service, mobile first marketing agency that helps companies to
14:16
scale and become category leaders.
14:19
Whether you're finding product market fit or ready to
14:22
accelerate growth, MO Burst delivers creative, data-driven
14:25
solutions that help with scale and efficiency and impact.
14:29
At MO Burst, we're not just guessing what works, we're
14:32
predicting it. As an AI first agency, MO Burst
14:36
uses cutting edge tools to analyze user behavior, adapt
14:39
creative in real time, and uncover hidden growth
14:42
opportunities. The result?
14:44
Marketing that's faster, smarter, and built to convert
14:48
from early stage startups to global brands.
14:50
We've helped hundreds of companies overcome complex
14:53
challenges, unlock new audiences, and optimize every
14:57
metric that matters. At Moeburst, we believe that
15:00
creative thinking, advanced technology, and data are the
15:03
core drivers of success at scale.
15:05
If you're serious about scaling your business with precision and
15:08
purpose, it's time to talk to Moeburst.
15:11
Visit moeburst.com and let's build something category
15:14
defining together. Yeah, I mean, I think the way
15:21
that I see often companies will start on the idea of human usage
15:26
of AI is they they think about it in terms of usage, right.
15:29
So AI is a a tool that you instruct as a human and you you
15:35
kind of give it a prompt and you can go through prompt training
15:37
and that's all good. And then you and then you can
15:40
get an answer. And it's it's really awesome.
15:42
It's it's amazing. I think we've probably all seen
15:45
the types of answers, the questions that you never would
15:47
have imagined. You can get them in seconds.
15:48
And I think there's two things that that don't get enough
15:52
attention, which I think I'm starting to see more of.
15:55
The first one is how do you actually think about what's the
15:57
right question to ask, right? So how do you actually cultivate
16:01
that process of critical thinking and, and thinking about
16:03
what will, if I can get an answer to any question in two
16:07
seconds, what should I be asking?
16:09
Because I could ask 100 questions in an hour, right?
16:12
And So what should I be doing? So that human critical thinking
16:15
process, I think is, is really important.
16:18
And then the second thing is actual dialogue and
16:21
collaboration. So thinking beyond AI is just a
16:25
thing that you instruct and thinking about it more as a
16:27
collaborator and thinking very deliberately about, well, how
16:31
should I collaborate with this system that I'm working with?
16:34
I how can I understand the way it works, it's limitations and
16:38
when I should let it, you know, take the reins and make
16:41
decisions itself? And when should I have that
16:43
human review process and just thinking and working very
16:46
deliberately with this new type of collaborator that we have in
16:50
front of us. And so I think that's not often
16:54
taught that much in companies. People are just sort of told,
16:56
well, go and use AI and they're not really talked about this
16:59
collaboration. So one of the things I've been
17:02
starting to do research around is haven't you actually to
17:05
measure and not just the usage of AI, but what's the actual
17:08
impact of the usage of AI? So I, I developed a framework
17:12
called the Collaborative Results Index, which is all around, how
17:17
do you measure, what are the results of your human AI
17:19
collaboration? What's the quality of the
17:21
relationship and the dialogue that you have when you're
17:24
working with AI and what's the resilience of your process?
17:26
So actually, are you just outsourcing your thinking?
17:29
Are you, as an individual learning more and developing
17:32
more domain expertise in that and starting to develop a
17:36
measurement system so that people and teams can actually
17:39
measure how well they're using AI and how they're collaborating
17:43
with it in a more deliberate way?
17:46
Is outsourcing your thinking bad, for example, for example,
17:51
you know, because you were saying kind of like folks will
17:55
don't know which questions to ask.
17:58
They should figure out what questions to ask.
18:00
But I think that would be the best first question to ask.
18:03
An LLM is like, hey, what are my blind spots?
18:06
You know What should I be asking you to ask before I get into
18:10
this work? Yeah.
18:11
So I actually think outsourcing, I'm thinking, is great as long
18:14
as you're aware you're doing it and you think it's the right
18:16
thing, right? So it's all about being
18:17
deliberate about how you work and how you think.
18:20
So there's this concept of metacognition, right?
18:22
And so thinking deliberately about how am I thinking, what am
18:26
I going to spend my time thinking about and what am I
18:28
going to outsource? And so as you said, like a,
18:30
working with AI is a great way to actually help spot the blind
18:35
spots, you know, unpacked biases in the way you typically
18:39
approach problems and say, Hey, you, you know a lot about me.
18:42
How can you figure out how I would misinterpret a situation
18:46
like this or what might I be missing out on?
18:48
And so I think it's actually great to outsource your thinking
18:50
as long as it's not all the time and you're doing it in the right
18:53
moments and in the right ways, because then it actually frees
18:56
you up to spend your brain power, which, you know, for us
19:01
all is limited. Like we can't all be thinking
19:03
all the time. It's kind of exhausting.
19:05
Then you can focus it on the things that matter most.
19:09
Yeah, I I find it much more productive and satisfying to be
19:14
a sculptor as opposed to a painter.
19:16
You know, if I, if I come to an LLM to solve a problem and I'm
19:20
like, this is exactly what I want, narrowly like give me this
19:24
thing as opposed to opening up and being like, hey, what should
19:28
I ask you to get the most out of this?
19:29
Then it, it delivers you a big block of marble and then you're
19:33
just chiseling away. You're like, I see where you're
19:34
going, but I don't want that. I don't want that.
19:36
These things are great. This was an idea I didn't have
19:38
you heard before. You know, that was a real kind
19:41
of light bulb moment for me when I realized I was, I was, you
19:45
know, just holding on to my cherished creative output like
19:50
it couldn't be approved upon, you know, or that like sharing
19:54
that creative capacity with the LLM wouldn't generate in like
19:58
maybe a better idea faster. You know, there's a, there's a
20:02
real humility that I think is really important.
20:04
And that's one of the things I try to think about is, is not
20:08
just thinking about AI as something you use, but it's a
20:10
collaborator that you genuinely see as like on level pegging
20:13
with yourself, right? And you want feedback on
20:15
yourself. And I think that's where a lot
20:17
of companies will head in the future is that we'll be working
20:20
in these kind of human AI teams, as I call them, where you're
20:24
getting feedback on AI, no, from AI.
20:27
And then there's probably an AI evaluating how well you're
20:30
evaluating AI and there's a human evaluating how well you're
20:33
evaluating AI and this whole kind of crazy loop.
20:36
And we're all going to be working and thinking in a
20:38
different way, but the AI systems are going to be much
20:42
more integrated as part of the teams that we're in, rather than
20:45
just like a tool that we might choose to consult once in a
20:48
while. Yeah, 100%.
20:50
So, you know, I, I think the folks that kind of the, the
20:53
audience of this podcast is fairly advanced, you know, so I
20:57
think like a lot of us are, I just make the assumption that
20:59
they're a lot like me. Like this summer I was building
21:01
automations for everything. You know, I was like, I was
21:05
really excited about like, Hey, what are the, the boundaries of
21:08
what I can and can't do both with generative, you know, kind
21:11
of opportunities, But then all the workflow automations.
21:14
And what, what I found is that like this collaboration between
21:18
human and AI is critical. So like, you know, the human
21:21
loop is, is huge, but there also seems to be a little bit of
21:24
friction, you know, and which just kind of results in me like
21:28
maybe 3/4 the automations I work.
21:30
I just threw them out the door. I was like, it actually, this is
21:33
a bit burdensome just actually managing that relationship.
21:38
You know, of all these AI agents, the ones I have are
21:41
still, you know, helpful. But we're kind of coming into a
21:44
second phase now, I think with as far as workflow automations
21:47
and like really taking advantage of, you know, where, where does
21:51
it make sense for me just to do a repetitive task because I'm
21:55
adding a lot of human context to it.
21:57
So I guess the question is, you know, where do you see teams
22:00
accidentally building this kind of friction between human and AI
22:05
collaboration? Yeah, Yeah, it's a great
22:09
question. I'm trying to think.
22:11
So can you give me a a second just to think about that one?
22:13
Is that OK? Sure.
22:16
Absolutely. Yeah, Yeah, it's kind of a weird
22:18
question. Yeah, no, sorry.
22:20
And actually I just, I don't know, there's an issue with the
22:22
Wi-Fi, but like I my video was cutting in and out there.
22:25
I I couldn't quite see the whole time, but I hope it's now fine
22:28
now. So that just a critical.
22:29
So the question was, where are teams deliberately creating
22:32
friction with AI? Accidentally accidentally
22:36
creating friction, you know. Yeah, OK, Yeah.
22:40
So yeah, it's a it's a great question, I would say.
22:44
I mean, in many situations serendipity is kind of the is
22:48
kind of the birth of a lot of innovations, right.
22:50
Like a lot of things get stumbled upon by accident.
22:53
I can think of a couple of examples.
22:55
One I think is actually I was working with a company on an AI
22:59
powered product design tool. So this is a tool that would
23:04
that would basically help to pull, pull in data from TikTok
23:07
and things like that to actually design ideas for like themes and
23:10
then product designs in the fashion space.
23:13
And within that, one of the things we found was actually
23:18
some of the quirks in the early design process in the image
23:21
generation, because this is like 2 years ago when things were not
23:24
quite so good. Some of those quirks actually
23:26
pointed to different silhouettes within the clothing that
23:30
actually were quite creative and quite interesting.
23:32
And so we actually built some structures to capture some of
23:35
the more creative ideas that came out of this design tool and
23:39
integrate them into like a structured process rather than
23:43
treat them as anomalies, right? And I think you have to kind of
23:46
treat the hallucinations and the weirdness of AI as a feature
23:50
and, and embrace it and know that you're going to get some
23:53
weird stuff and some unexpected results.
23:56
But that's actually part of, you know, you part of like working
24:01
with humans too, right? Like you get, you get weirdness
24:05
and surprises and, and little like avenues that you go down
24:08
when working with humans. And you, I think, you know,
24:11
probably yourself and a lot of your listeners have had weird
24:13
things happen where they go off in some random direction with AI
24:16
and they, they have to rein it in.
24:18
But like, sometimes there's actually a a part of the
24:20
expiration process. Yeah, totally.
24:23
That's so funny. I I hadn't really ever thought
24:25
about that before. Like yeah, humans are pretty
24:27
weird too. They have their own
24:28
hallucinations. Third thing.
24:30
Well, actually, so that's one of the things that's really
24:32
interesting. So I think in the last couple of
24:34
years, I actually think, you know, in working with this new
24:36
type of intelligence, we've been able to have a lot more
24:40
deliberate conversations around what it's like to be a human and
24:43
shut. It's almost like a mirror on
24:45
humanity a little bit, right? And you think about, yeah,
24:48
people talk about hallucinations a lot and it's an issue and you
24:51
have to manage it. But human memory is terrible.
24:54
Like it's, it's well known within, you know, criminal
24:56
justice that like human memory and testimonies, people kind of
24:59
construct stories in their head around memories even with the
25:02
best of intent, right? It's called confabulation within
25:05
human psychology. And it's not that dissimilar to
25:09
how AI works. And I think what's actually
25:10
interesting is a lot of the quirks and witness of AI systems
25:14
are kind of similar in some ways to how humans work too.
25:20
And you know, we're we're still figuring out human psychology
25:23
after thousands of years of studying it.
25:25
So it's probably optimistic of us to think that we're going to
25:29
figure everything out about AI in a few years too.
25:32
Yeah, yeah, it's a, it's a really good point.
25:35
I mean, you know, there's like a a famous Black Mirror episode
25:40
where, you know, you basically are just recording everything
25:43
all the time. So you have like a documented
25:46
reference point of your entire life, essentially, you know, the
25:50
second they have like, you know, these glasses can record
25:53
everything and then it goes inside your eye and then, you
25:55
know, and what was was so poignant about that episode was
25:59
like how damaging and dangerous that is.
26:02
I mean, there's a great essay from who's the guy who wrote Tom
26:08
Sawyer and Huckleberry Finn. I'm good.
26:12
Samuel Clemens, but his name is actually.
26:16
So the tooth of my tongue can't remember now.
26:18
Somebody tweet me and tell me I'm an idiot.
26:21
Anyway, he he wrote an essay on the the beauty and respect of
26:25
white lies. You know, where you don't really
26:29
have someone to ask you how they look in those jeans.
26:31
It's always like, man, you look great.
26:33
You know what I mean? Like the little white, the
26:37
little white lie is like is regarded as polite.
26:41
Now when we we have like a a photocopy recollection of
26:44
everything, then you know, all of these white lies get exposed.
26:48
It not only with that we're maybe doing with our Co workers,
26:51
our family, our friends, but all the white lies we tell ourself.
26:55
You know, which I feel like what we're just like, we eliminate
27:00
all the confidence that anyone's ever accumulated because, you
27:04
know, I think confidence relies on creative storytelling about
27:07
your past and creative projection about the future.
27:11
You know, so we're starting to think about hallucinations as
27:13
they tighten up. I don't know, it just makes me
27:15
think of like how much I actually appreciate human
27:19
hallucinations, at least when it's it's surrounded around like
27:22
just the polite pleasantries of life.
27:25
Yeah, like well-intentioned ideas and and approaches.
27:29
Yeah, totally. And I think that it raises this
27:32
interesting question of like how do you think about trust and
27:36
honesty and openness? And this is one thing that I
27:38
think that's really, really important as we sort of start to
27:41
navigate this as a society. And you know, one of the
27:45
concepts I was in a meeting talking about recently was the
27:48
idea of a personal data lake, right?
27:50
And so, you know, you have enterprise data lakes and
27:52
everyone thinks about how they can kind of create a data
27:55
pipeline into their AI tools, all those kind of stuff.
27:57
But how do you think about that for yourself?
27:58
And like what, you know, what data do you want to share with
28:03
with AI systems and not? And then how do you think about
28:06
which of the companies and the tools that you want to share
28:08
that with? And I think there's a lot of
28:10
opportunities too, for companies to start to think about how do
28:13
they become a trust broker and a way of protecting your data from
28:16
leaking and getting out there and, you know, thinking
28:20
deliberately around who you share your your relationship
28:24
data with AI systems on. Yeah, I'd love to get your
28:28
opinion on on this a little bit. You know, as we're looking at
28:31
the future and we're looking about what's coming next, you
28:34
know, it's very obvious that like there's going to be some
28:36
kind of AI companion that's going to be with us at some
28:39
point close to birth all the way through our entire life.
28:44
And as we're starting to think about companies, one of the
28:47
trends is like, OK, I want to, we want to create our own AI
28:52
agent chatbot, whatever it is to our website, to our app, and we
28:57
want you to engage with our technology.
29:00
The other side of that is, you know, do we want to have our own
29:04
personal, you know, safeguarded chatbot that that we build a
29:09
tight relationship with that then we share access with that
29:13
app with that thing. You know, so I think a bit of
29:15
like we got that new Avatar movie coming up, you know, so
29:19
like when an avatar wants to jump on their dragon, like it
29:22
plugs its braid into like the the Dragon's tail or something.
29:26
So there's like a connection. And I really see like that as
29:29
kind of being a metaphor of like what we're going to do our 1:00
29:32
potential way with our kind of personal AI agents is that like
29:37
we'll have ours and that will actually interact with the
29:40
company as opposed to all these companies having theirs that's
29:43
trying to interact with us. Where do you see the future
29:45
going in which direction? Yeah.
29:47
I mean, I think, you know, actually I see it going in a
29:49
couple of different directions, partly based on which kind of
29:52
phone you have, right. So like an Apple phone.
29:54
I have an Apple phone, right. The Apple intelligence approach
29:56
is very much that they're building Apple intelligence into
29:59
your device and so much of the what you do and what you tell it
30:04
and what you share with it stays on your device and you own it.
30:06
It never goes back to Apple. The Android phone is very
30:09
different, right? All of the data goes back to the
30:12
Google servers, wherever they are, all over the world.
30:16
And I could imagine a thing, a situation in the pretty near
30:19
future whereby Apple Intelligence becomes that kind
30:22
of trust broker for you. So if I log into a new website,
30:27
you know, some kind of app app, you know, agent, assistant,
30:30
whatever it might be, Apple Intelligence could very easily
30:34
give that agent or assistant a little token of information
30:38
about me that's contextual to the type of information they
30:42
need to serve me in a personalized way.
30:45
But they don't get to keep that. They get that for a moment just
30:48
to make the right recommendation to me or offer the right service
30:51
to me. And then Apple takes it back and
30:53
it stays on my device. So I can see that that's
30:55
happening pretty soon with like, you know, an Apple phone, for
30:58
example. But I think not every company
31:01
has taken that that approach. I think you've got, I think
31:04
they'll be pretty soon a, a bifurcation of the market
31:07
between AD funded AI services and then services that you might
31:13
be more likely to pay for that in which you would actually
31:16
protect your data and they would not sell that data were provided
31:21
to advertisers. And I think we'll probably see a
31:22
big difference. And I think that will play out
31:25
in complex ways around trust. So right now, if people talk to
31:31
ChatGPT, they actually generally study show that people trust
31:34
ChatGPT more than Google. Because now, you know, Google
31:37
used to be super trusted like 20 years ago when you just got
31:40
really good search results. Now there's a load of ads and
31:43
sponsored links. You don't really trust the
31:45
answers that come out from that. And you know, studies prove that
31:48
out. I think we we might get to a
31:50
situation in a few years where if AI services start to become
31:54
more ad funded, you're going to start to question what are the
31:57
answers they're giving you and why they're giving you that in a
32:00
different way if it's an ad funded system as opposed to
32:04
something you're paying for. Yeah, yeah.
32:07
I mean, you know, as far as trust, it's like people
32:10
definitely do not trust Google and they barely trust Chat GPTI
32:15
mean, I, I feel like generally speaking, I feel like generally
32:20
speaking, we're kind of moving further and further towards a
32:24
low trust society, which I think is kind of maybe new for
32:27
America. And so you see the rise of like
32:31
the surveillance kind of increasing, particularly with AI
32:35
tools. And I, I think this is going to
32:37
grow. I, I see where like a
32:38
opportunity for, you know, the technocratic Illuminati to like
32:43
win back some trust is that if we are interacting with our own
32:49
AI agents and there's opportunity to actually not
32:51
make, not for them to earn money from advertising, rather to, to,
32:55
for the user to get paid for advertising.
32:59
You know, I, I've seen a couple of models of this, of this a
33:02
couple of times. But truly, if there you have
33:04
kind of a hierarchy of data, then you're willing to like say
33:07
like, OK, if it qualifies for this, this row, then it can move
33:11
on to the second row. So you have this deeper and
33:13
deeper qualification for certain products so that then the the
33:17
company that wishes to market to you actually pays the individual
33:20
user. I think we're really close to
33:21
that. And I think that helps build the
33:23
trust, not only because now people are getting some kind of
33:26
dividend in their crypto wallet that's stored in the browser or
33:30
whatever, but also, you know, their, their qualification for
33:33
the right products is going to find people much easier.
33:36
You know, so when that I I would love to that happen just to kind
33:40
of just to improve upon kind of the the deterioration of trust
33:44
just kind of culturally across the whole society.
33:46
Yeah, totally agree. And I think the other thing it
33:48
raises too is the flip side of that is how does AI, how does AI
33:54
actually build trust in humans and companies as well?
33:57
So it's not just about what humans trust.
33:59
It's now a question of what does AI trust?
34:01
So I'd imagine, for example, like a lot of your listeners,
34:04
Ryan might be founders and people who are running their own
34:07
businesses and they're thinking about how does my business, how
34:10
does my product or service get mentioned by ChatGPT or Gemini
34:14
when someone's asking a question, right?
34:15
And so that increasingly is becoming a question of trust in
34:19
terms of what are the trusted answers to those questions that
34:22
they're asking in the eyes of ChatGPT and, and Gemini and
34:25
Claude or whatever it is. And so, so I, I published a book
34:28
recently called when AI shops and it's all about AI shopping
34:32
and a jet to commerce and the new kind of rules of the road
34:36
for marketers in terms of how you build trust with LMS and how
34:40
you get visible to them. And you know, one of the things
34:44
that LLMS look for is they look for very recent, very contextual
34:47
information, but also consistency of stories about
34:51
that service or product across the Internet, right?
34:53
And so that's how they, that's their marker of trust.
34:56
So like, you could have, you know, great stories written
34:59
about yourself on your website, but then on Reddit and Wikipedia
35:02
and other sites, they're saying a very different set of things
35:05
about you. And so LMS are picking up on
35:07
that and they will down weight that kind of company or that
35:10
service because they're seeing there's a discrepancy between
35:13
what the company is saying and what the rest of the Internet is
35:15
saying about them. Yeah, I love that.
35:19
All right, we're at that time in the show where we're entering
35:21
the quantum round. Are you ready?
35:24
Let's hope so. Yeah, so you wrote that AI is
35:29
making us smarter and Dumber. How's it making us Dumber?
35:32
I think it's making us Dumber in that it's super easy just to
35:37
just to not think about things. And it's like going to the gym.
35:40
If you don't train those muscles, you lose them.
35:45
Yeah, 100%. I noticed that myself.
35:51
You know, most of the folks here are in startups in various
35:54
stages. I'm working with a company like
35:56
Coca-Cola. What's kind of your favorite
35:58
thing about working with enterprise teams?
36:02
I think I mean, they, they obviously have a lot more
36:08
complexity to deal with, right, in that they work in large
36:11
systems and you know, it's like, it's like moving a super tanker,
36:16
but once you actually get the traction to do something, things
36:21
can be move incredibly fast and the impact that you're going to
36:26
have is much bigger. And so like part of my passion
36:30
in my career is how do I actually work with very large
36:33
companies to do things that are super complex in large
36:36
organizations that give them a huge sense of satisfaction that
36:41
they've managed to move that kind of super tanker in a new
36:44
direction. Yeah, totally.
36:46
It's super impressive. This audience loves business
36:50
ideas. Do you have any cool business
36:52
ideas rattling around, particularly ones that can be
36:54
sprung up to $1,000,000 run rate quickly?
36:58
Oh, you mean like a business idea that I'm working on or for
37:00
other people? Yeah, either.
37:03
OK, so I got this. So I got this idea that I, I'm
37:06
actually looking for, for alpha testers around.
37:08
So if anyone's interested, I'm looking to build a, a new human
37:12
AI collaboration measurement system.
37:15
So I'm building a, I'm building a, an alpha version of this.
37:18
It's a Chrome extension that actually would plug into your
37:21
Chrome and monitor, not what you're saying and like not
37:26
tracking exactly what you're doing, but actually monitoring
37:29
the quality of your collaboration with AI systems,
37:31
whether that's ChatGPT or Gemini or Claude.
37:33
And then using that to measure how effectively you're doing and
37:37
giving you advice around how to do a better job, right.
37:39
So I think everyone I know, I, I work with so many people who are
37:43
looking at how do they improve their skills in working with AI.
37:46
And I'm, I'm thinking about how can I build a business that can
37:49
track that, give people real time in the moment advice around
37:53
that and give them real direction around what they can
37:55
do next. So that's something I'm kind of
37:57
building a, a testing group around.
37:59
So if anyone's interested, maybe you can put the put the
38:02
information in the in the contact contact details.
38:07
You bet. I'm definitely interested too.
38:09
I think that's a great idea, largely because I think it would
38:11
uncover a lot of patterns that maybe you're not aware of that I
38:15
wouldn't be aware of. Yeah.
38:16
You know, because like now it's the every kind of application
38:19
has a glowing AI button in the right hand corner where you can
38:24
use it. But you know, I never hit that
38:26
button. So it would probably see like me
38:28
going back and forth from like spreadsheets to, to Gemini over
38:32
and over again when it's like Gemini's built right in.
38:35
You know, if it if it had a recommendation of like, hey,
38:37
Ryan, you actually don't need to be like toggling back and forth.
38:39
You can, you can do this automatically.
38:41
I think that's a great idea. Yeah, absolutely.
38:43
Yeah. So actually, so I've got a, I've
38:45
got a website, it's Corex dot AICO, double Rix dot AI, where
38:50
people can sign up as a, as a alpha tester.
38:54
Co What is it? CORRIX dot AI.
38:59
Or XCORRI x.com OK, cool. I'll I'll definitely sign up now
39:04
and I'll put it in the I'll put in the show notes yeah Or you
39:08
know, GPT put that in the show notes.
39:10
Yeah indeed more. Appropriate.
39:13
Yeah, it's funny because a lot of people like you do, you do
39:15
find in meetings a lot of people talk to the AI assistant now,
39:18
like, OK, so now the next steps are this like you can see people
39:22
starting to do that. 100% cool. If you're entering college right
39:28
now, what would you study? You know, I, I, I certainly
39:33
wouldn't be studying like, you know, like a vocational thing
39:38
like like business or marketing or computer science, maybe maybe
39:42
computer science, but actually like the, the actual how do
39:45
computers work, not coding. I think I would, I think I would
39:47
probably look to study something that would help me to really
39:53
think deeply, right? So something I'm passionate
39:54
about, probably in the humanities, not worry too much
39:58
about what it was, but something that I can really engage with
40:02
and become an expert. And that's probably what I would
40:05
look at, because I think that increasingly the ability to
40:11
think deeply, become really expert in a domain is a skill in
40:15
itself. And I think sometimes it doesn't
40:18
always matter what that area is, but it's cultivating that
40:21
ability to become an expert in a domain I think is actually very
40:23
important. Very wise, Jeff.
40:27
This has been so much fun. We love to give shout outs to
40:30
charitable organizations and we're going to give a shout out
40:33
to Saint Jude's largely because they've been a client of yours.
40:37
You want to give us a a little rundown of what Saint Jude's
40:41
does and you know what the the kind of work you you perform
40:44
with them. Yeah, so Saint Jude's is an
40:46
amazing institution. So it's a, it's a hospital, but
40:49
also an incredible research institution focused mostly on
40:53
pediatric cancer. A lot of you might have heard of
40:55
them. They do, you know, you probably
40:57
get a lot of mailers from them in the, if you live in the US
41:00
around fundraising. And I've actually been working
41:02
with them this year around how can they actually implement
41:05
human AI workflows to improve the way they do marketing and
41:08
fundraising actually, which is a really nice thing to get to do.
41:13
Best, Yeah, shout out to them, stjude.org, Children's Research
41:18
Hospital. Really happy about the work they
41:20
do. I'm really happy about your
41:22
work, Sir. It is human-machines.com where
41:26
people can learn more about your work and then what's the the
41:30
best place for people to connect with you online.
41:33
Best place is LinkedIn. Amazing.
41:37
And we will link that in the show notes as well.
41:39
Jeff, thank you so much for coming on the show.
41:41
Yes. Thank you, Brian.
41:42
Pleasure. Hey founders, let me ask you
41:45
this, are you building with AI or getting left behind by it?
41:50
If you're serious about scaling smarter, faster and bolder, you
41:54
need to check out AI for founders.co.
41:58
It's your go to resource for cutting through the noise and
42:01
actually using AI to grow your startup.
42:04
Each week, AI for Founders drops a sharp, no fluff newsletter,
42:09
dives deep with expert guests on the podcast, and hosts monthly
42:13
workshops where you can get your hands dirty with the real tools
42:16
and strategies. Whether you're precede or Series
42:20
B and beyond, launching something new or just staying
42:23
sharp, AI for Founders gives you the edge you need to win in 2025
42:28
and beyond. Hit AI for founders.co to
42:32
subscribe to the newsletter, tune into the podcast, and grab
42:35
your spot in the next workshop, and it's all free.
42:38
Build smarter, move faster, think bolder with AI for
42:42
Founders. Now back to the show.