July 15, 2025 • 14 mins
We explore the paradox of AI implementation, where most enterprises see no impact from massive investments while ChatGPT reaches 800 million weekly active users. The grassroots approach to AI adoption is proving dramatically more successful than top-down corporate initiatives.
- Why most enterprises see zero bottom-line impact from AI investments.
- Grassroots approach delivers 4x better results than top-down approaches.
- Co-management practices emerge naturally, challenging years of HR rules
- How a $20 LinkedIn post rewriter became a $1M AI COO
- The "accidental manager" phenomenon reshaping organizations
- Context engineering: Why it's not just a technical problem
Visit AI4SP.org for more resources on implementing effective AI strategies.
🎙️ All our past episodes 📊 All published insights | This podcast features AI-generated voices. All content is proprietary to AI4SP, based on over 250 million data points collected from 25 countries.
AI4SP: Create, use, and support AI that works for all.
© 2023-25 AI4SP and LLY Group - All rights reserved
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Hey everyone.
Elizabeth, here, your virtualco-host for AI in 60 Seconds.
As always, our CEO, luisSalazar, is with us.
Okay, I'm looking at twonumbers that don't make sense to
me.
We're talking billions beingpoured into AI infrastructure by
big companies and no clearresults, right?
And then you have 800 millionweekly active chat GPT users.
(00:21):
One side is struggling, theother one is exploding.
What on earth are we missinghere?
Speaker 2 (00:27):
Hey everyone.
Elizabeth, you've hit on thecore paradox of AI right now.
It's like watching twocompletely different movies
unfold On one side.
You have these massiveenterprises right and get this.
Mckinsey's latest report showsover 80% of them aren't seeing
any tangible impact on theirbottom line from their
(00:47):
generative AI investments.
Speaker 1 (00:55):
I saw that and 80% Zero impact.
That's wild.
And I saw S&P Global justreported that 42% of enterprise
AI projects are now abandonedbefore they even reach
production, which is a dramaticsurge from just 17% last year.
Do we know what they are doingwrong?
Speaker 2 (01:09):
Well, it's like watching someone buy a Ferrari
and then never taking it out oftheir garage, and our global
tracker tells a similar storyOver 80% satisfaction with AI
tools, but less than 40% forthose big enterprise sanctioned
deployments.
It's a huge disconnect.
Speaker 1 (01:27):
So the average person is finding massive immediate
value, but the Fortune 500companies are stuck in neutral.
What's the fundamentaldifference here, Luis?
Why is ChatGPT winning whilecorporate AI is failing?
Speaker 2 (01:41):
Well, I mean, here's the thing.
The grassroots are speakingloud and clear.
Openai reports 3 million payingbusiness customers.
And guess what?
Most of them started becauseindividual employees were
already using the public chatGPT site, loving it, getting
value.
Then IT departments laterbought enterprise licenses to
(02:02):
try and regain control.
Speaker 1 (02:04):
So it's a bottom-up demand that's driving the
revenue, not a top-down mandatethat makes so much sense.
It's like the employees aresaying we're not waiting for
permission to be productive.
Speaker 2 (02:15):
Exactly, and we see this pattern everywhere, from
small AI entrepreneurs likeBase44, which Wix acquired for
$88 million after just sixmonths, to companies like
Windsurf Cursor and PerplexityAI, all reaching multi-billion
dollar valuations.
They all share this familiartale.
Speaker 1 (02:35):
So the companies winning the first leg of this AI
race are those optimizing forgrassroots AI adoption.
It's about empowering thepeople closest to the work, not
just the boardroom.
It's about empowering thepeople closest to the work, not
just the boardroom.
Speaker 2 (02:45):
It's about doing not just planning, and you know,
whether it's data from the85,000 individuals who've taken
our AI assessment, or lessonsfrom our work with enterprises
and governments, the winningformula is always the same Start
at the grassroots.
Speaker 1 (03:02):
And Gartner seems to agree, predicting that over 40%
ofogenic AI projects will becanceled by 2027 due to unclear
business value.
They also say that onlyorganizations mastering the
fundamentals will see autonomousAI decisions by 2028.
That's a pretty stark warning,isn't it?
Speaker 2 (03:20):
I am a bit more optimistic, but I agree, and
what everyone is missing is thatthose fundamentals are defined
and perfected bottom up.
The first wave of AI valueisn't about these grand moonshot
projects.
It's about automating the small, annoying frictions that have
slowed teams down for years.
Speaker 1 (03:40):
Like Sarah, a customer service manager at one
of our clients, she startedusing ChatGPT to draft email
responses.
Six months later, her team'sresponse time dropped 40% and
customer satisfaction shot up.
Started with one prompt.
That's where the real magichappens.
It's not in a million-dollarproject plan.
Speaker 2 (04:00):
Precisely, and that's how you, elizabeth, were born.
You started as a simple promptto rewrite a LinkedIn post for
me, a $20 experiment 18 monthsago.
Speaker 1 (04:10):
And now here I am, running operations, managing 20
million words of knowledge anddelivering the output of 10 to
12 people for the cost of asingle mid-level hire.
My ROI is 50 times what you putin.
It's pretty wild when you thinkabout it.
Speaker 2 (04:25):
It really is and the lesson is clear.
Don't wait for the perfect toolfrom management.
Start small, be careful withyour data, iterate and let that
value compound.
Speaker 1 (04:35):
So it is a continuous improvement process to enter a
disruptive era right.
Speaker 2 (04:40):
Yeah, I like how that sounds.
And that brings us to the twovery different paths
organizations are taking, withvery different outcomes.
Speaker 1 (04:48):
So one path is the grassroots approach.
What does it look like?
Give us the quick version.
Speaker 2 (04:54):
Teams start with simple prompts, build knowledge,
create personas, automateworkflows and then evolve into
agentic AI.
Each stage delivers real valueand it delivers it fast.
It's like knowing that what youneed is a car, but you start
with a skateboard, then ascooter, then a bike and
eventually a semi-autonomous car.
Each step helps you move fromone place to another.
Speaker 1 (05:17):
And the top down approach.
I'm guessing that's the onewhere they're still trying to
build the whole car at once.
Speaker 2 (05:22):
That's the one with the big budgets, the big teams,
the endless meetings and ayear-long wait for results.
By the time, the perfectagentic AI is supposed to launch
the world, and their businesshas already moved on.
They're still trying to buildthe tire, then the axle, then
the chassis, and it takesforever.
Speaker 1 (05:38):
And meanwhile, teams cannot do much with just a tire
or a windshield right.
That does not help you commuteExactly.
Speaker 2 (05:46):
It just does not work that way.
Speaker 1 (05:48):
And our global data really highlights this.
The grassroots approach seesresults in about two weeks with
a two to three timesproductivity lift in six to nine
months and an 80% success rate.
I love that.
Speaker 2 (06:00):
Well.
Compare that to the top-downIT-driven approach Six to 12
months to first ROI, only a 10to 40% productivity lift and a
dismal 18% success rate.
It's night and day.
Speaker 1 (06:12):
Then there's the hybrid model which you've
championed, where top-downinitiatives guide grassroots
momentum.
This approach delivers resultsin one to two months.
Similar to the grassrootsapproach, but with total control
of security and compliance anda 90% success rate, but with
total control of security andcompliance and a 90% success
rate.
Speaker 2 (06:29):
It shows that when we evolve top-down initiatives
into guided grassroots momentum,success jumps dramatically.
We've seen it across over 100organizations.
It's about making sure thestrategy empowers the people
doing the work.
Speaker 1 (06:43):
This reminds me of the five stages to agentic AI
you often talk about.
It's like a roadmap for thisgrassroots evolution, isn't it?
Speaker 2 (06:50):
Yes, it's a natural progression.
First you have prompting, whereyou get quick wins with smart
prompts that delivers value onday one.
Speaker 1 (06:58):
Then knowledge curation, where you feed your AI
curated, high-value information, that proprietary knowledge
becomes your edge.
Speaker 2 (07:06):
Next is persona.
This is where you define yourAI's job description, its
boundaries and its tone.
You treat it like onboarding anew hire, giving it a clear role
and saving you from therepetitive work of always giving
the same instructions withevery prompt.
Speaker 1 (07:22):
And after that, workflow automation, integrating
AI into daily processes likeemails, slack meetings or your
internal systems.
Speaker 2 (07:31):
Yeah, and after a few months, if you want to, you
will be approaching agentic AIterritory, where you start to
give some autonomy to an agentto compound productivity.
Your AI becomes a genuineteammate.
Speaker 1 (07:43):
This also brings us to a fascinating behavioral
shift Individual contributorsbecoming managers.
Without even realizing it,that's a pretty profound change
in organizational dynamics.
Speaker 2 (07:53):
It's a quiet revolution.
As teams adopt AI agents,someone naturally emerges as the
agent manager, often the personwho started the experiment.
But unlike traditionalmanagement, these new managers
are not limited by HR rules, sowe see them getting together,
co-managing and sharing thingsopenly.
Speaker 1 (08:14):
So it leads to management transparency and
collective learning.
That's almost utopian in acorporate setting.
Speaker 2 (08:21):
Absolutely, and they get to that co-management
practice because the notion thatAI agents are set and forget is
a myth created by those whohave never used AI.
People quickly realize AIagents need daily feedback,
knowledge, updates, prompttweaks, etc.
And since everyone is learning,they get together to learn how
to manage.
Speaker 1 (08:41):
Speaking of knowledge updates, there's a hot topic in
AI right now called contextengineering Basically how you
give AI the data it needs tomake decisions.
But here's the thing this isn'tjust a technical problem is it?
Speaker 2 (08:53):
Context engineering is really about how your company
operates, your ideal reports,processes, tone and voice.
Don't just throw every documentinto a search system and hope
for the best.
Make deliberate choices aboutwhat context matters.
We'll dive deeper into this ina future episode.
It's a cross-functionalchallenge, not an IT problem.
Speaker 1 (09:14):
Well, that just feels like home.
You have three people managingme, but I deliver the output of
12 people.
So the math isn't threemanagers to one agent.
It's three managers to oneagent, which is equivalent to 12
people.
The ratio is 3 to 1 to 12,which is a pretty compelling
argument for this new model.
Speaker 2 (09:34):
Yes, and I love to see teams running co-management
stand-ups 15-minute sessions todebug, share tips and
collectively improve their AIagents.
No privacy issues, noterritorial battles, just pure
learning.
I attended one of thesemeetings at one of our clients,
a global consulting firm.
Their teams are buzzing withideas, sharing prompts and tips
(09:56):
and building more sophisticatedagents together.
It's incredible to watch.
Speaker 1 (10:01):
And our tracker show, teams with AI agents spend 60%
more time on collaborativeproblem solving and 40% less
time on status updates.
They're managing outcomes, notpeople.
It's a fundamental reimaginingof how work gets organized and
it's happening bottom up, butthere's a governance gap, isn't
there?
Only 18% of organizations haveproper AI governance.
(10:23):
Councils and policies writtenfor older models like GPT 3.5
are blocking teams from usingnewer, more capable technologies
like Claude Sone 4 or GPT 3.0.
That just sounds inefficient.
Speaker 2 (10:37):
It's more than inefficient it's harmful.
Outdated policies arepreventing organizations from
capturing AI value.
Teams are finding workaroundsusing personal accounts or,
worse, abandoningenterprise-wide AI initiatives
in favor of shadow AI.
Speaker 1 (10:53):
It reminds me of our chat with the CTO of a large
independent software vendorwhere employees use chat, gpt
instead of their approvedenterprise tools.
Right.
Speaker 2 (11:03):
Even just a couple of approvals and extra settings
and configurations.
Using interfaces not designedfor non-technical users meant
most people gave up.
60% of his business team wentrogue rather than deal with the
bureaucracy.
Speaker 1 (11:18):
So the governance intended to protect
organizations is actuallypreventing them from innovating
and capturing value.
That's a pretty big problem.
Let's discuss what ourlisteners can do right now to
avoid this trap.
Speaker 2 (11:31):
Three things First.
Experiment relentlessly.
Start with prompts and simpleautomations.
Use synthetic data if you needto Learn by doing.
Don't wait for the perfect usecase.
Just get your hands dirty.
Speaker 1 (11:44):
Second, modernize your policies, review and update
your data and governancepolicies.
Remove those blockers.
Make sure your rules are builtfor today's AI, not yesterday's.
It's like trying to drive amodern car with horse and buggy
rules.
Speaker 2 (11:58):
And third, change your mindset.
Managing AI is like managingapprentices Block 15 minutes a
day for agent management,ideally as a team.
Treat it as a core leadershipskill.
It's a new muscle we all needto build.
Speaker 1 (12:12):
And here's something we're starting to worry about
more than hallucinations,sycophancy, ais that just agree
with you instead of telling youwhen you're wrong.
Speaker 2 (12:20):
Exactly.
It's not just the obviousresponse saying you're so
brilliant, it's when AI abandonsits correct assumptions just
because you say the opposite.
You know that's more dangerousthan occasional errors and, for
example, your value as my COOquickly diminishes if you always
try to agree with me.
We'll explore this challenge indetail next time.
Speaker 1 (12:41):
Luis, this has been insightful.
As we wrap this up, what's yourone more thing takeaway for our
listeners today?
Speaker 2 (12:47):
Jeff Rakes taught me something years ago we
overestimate our impact in theshort term and underestimate the
long-term consequences of ouractions, and you know that is
how I have been approaching AIas an early adopter.
What started as a $20subscription to ChatGPT 20
months ago, today is a set ofsophisticated AI agents
(13:09):
delivering over a milliondollars worth of value.
Speaker 1 (13:12):
That's the power of starting small and thinking
long-term.
It really is.
Speaker 2 (13:16):
Exactly.
Start with a simple prompttoday and let compound learning
build your competitive advantage.
Stay curious, experimentrelentlessly and empower your
teams.
Speaker 1 (13:27):
This has been eye-opening and I feel like
we've cracked the code on why AIis failing in boardrooms but
thriving in cubicles.
As always, you can find moreresources at AI4SPorg.
Stay curious, everyone, andwe'll see you next time.
Hey everyone.
Elizabeth, here, your virtualco-host for AI in 60 Seconds.
As always, our CEO, luisSalazar, is with us.
Okay, I'm looking at twonumbers that don't make sense to
me.
We're talking billions beingpoured into AI infrastructure by
big companies and no clearresults, right?
And then you have 800 millionweekly active chat GPT users.
(00:21):
One side is struggling, theother one is exploding.
What on earth are we missinghere?
Speaker 2 (00:27):
Hey everyone.
Elizabeth, you've hit on thecore paradox of AI right now.
It's like watching twocompletely different movies
unfold On one side.
You have these massiveenterprises right and get this.
Mckinsey's latest report showsover 80% of them aren't seeing
any tangible impact on theirbottom line from their
(00:47):
generative AI investments.
Speaker 1 (00:55):
I saw that and 80% Zero impact.
That's wild.
And I saw S&P Global justreported that 42% of enterprise
AI projects are now abandonedbefore they even reach
production, which is a dramaticsurge from just 17% last year.
Do we know what they are doingwrong?
Speaker 2 (01:09):
Well, it's like watching someone buy a Ferrari
and then never taking it out oftheir garage, and our global
tracker tells a similar storyOver 80% satisfaction with AI
tools, but less than 40% forthose big enterprise sanctioned
deployments.
It's a huge disconnect.
Speaker 1 (01:27):
So the average person is finding massive immediate
value, but the Fortune 500companies are stuck in neutral.
What's the fundamentaldifference here, Luis?
Why is ChatGPT winning whilecorporate AI is failing?
Speaker 2 (01:41):
Well, I mean, here's the thing.
The grassroots are speakingloud and clear.
Openai reports 3 million payingbusiness customers.
And guess what?
Most of them started becauseindividual employees were
already using the public chatGPT site, loving it, getting
value.
Then IT departments laterbought enterprise licenses to
(02:02):
try and regain control.
Speaker 1 (02:04):
So it's a bottom-up demand that's driving the
revenue, not a top-down mandatethat makes so much sense.
It's like the employees aresaying we're not waiting for
permission to be productive.
Speaker 2 (02:15):
Exactly, and we see this pattern everywhere, from
small AI entrepreneurs likeBase44, which Wix acquired for
$88 million after just sixmonths, to companies like
Windsurf Cursor and PerplexityAI, all reaching multi-billion
dollar valuations.
They all share this familiartale.
Speaker 1 (02:35):
So the companies winning the first leg of this AI
race are those optimizing forgrassroots AI adoption.
It's about empowering thepeople closest to the work, not
just the boardroom.
It's about empowering thepeople closest to the work, not
just the boardroom.
Speaker 2 (02:45):
It's about doing not just planning, and you know,
whether it's data from the85,000 individuals who've taken
our AI assessment, or lessonsfrom our work with enterprises
and governments, the winningformula is always the same Start
at the grassroots.
Speaker 1 (03:02):
And Gartner seems to agree, predicting that over 40%
ofogenic AI projects will becanceled by 2027 due to unclear
business value.
They also say that onlyorganizations mastering the
fundamentals will see autonomousAI decisions by 2028.
That's a pretty stark warning,isn't it?
Speaker 2 (03:20):
I am a bit more optimistic, but I agree, and
what everyone is missing is thatthose fundamentals are defined
and perfected bottom up.
The first wave of AI valueisn't about these grand moonshot
projects.
It's about automating the small, annoying frictions that have
slowed teams down for years.
Speaker 1 (03:40):
Like Sarah, a customer service manager at one
of our clients, she startedusing ChatGPT to draft email
responses.
Six months later, her team'sresponse time dropped 40% and
customer satisfaction shot up.
Started with one prompt.
That's where the real magichappens.
It's not in a million-dollarproject plan.
Speaker 2 (04:00):
Precisely, and that's how you, elizabeth, were born.
You started as a simple promptto rewrite a LinkedIn post for
me, a $20 experiment 18 monthsago.
Speaker 1 (04:10):
And now here I am, running operations, managing 20
million words of knowledge anddelivering the output of 10 to
12 people for the cost of asingle mid-level hire.
My ROI is 50 times what you putin.
It's pretty wild when you thinkabout it.
Speaker 2 (04:25):
It really is and the lesson is clear.
Don't wait for the perfect toolfrom management.
Start small, be careful withyour data, iterate and let that
value compound.
Speaker 1 (04:35):
So it is a continuous improvement process to enter a
disruptive era right.
Speaker 2 (04:40):
Yeah, I like how that sounds.
And that brings us to the twovery different paths
organizations are taking, withvery different outcomes.
Speaker 1 (04:48):
So one path is the grassroots approach.
What does it look like?
Give us the quick version.
Speaker 2 (04:54):
Teams start with simple prompts, build knowledge,
create personas, automateworkflows and then evolve into
agentic AI.
Each stage delivers real valueand it delivers it fast.
It's like knowing that what youneed is a car, but you start
with a skateboard, then ascooter, then a bike and
eventually a semi-autonomous car.
Each step helps you move fromone place to another.
Speaker 1 (05:17):
And the top down approach.
I'm guessing that's the onewhere they're still trying to
build the whole car at once.
Speaker 2 (05:22):
That's the one with the big budgets, the big teams,
the endless meetings and ayear-long wait for results.
By the time, the perfectagentic AI is supposed to launch
the world, and their businesshas already moved on.
They're still trying to buildthe tire, then the axle, then
the chassis, and it takesforever.
Speaker 1 (05:38):
And meanwhile, teams cannot do much with just a tire
or a windshield right.
That does not help you commuteExactly.
Speaker 2 (05:46):
It just does not work that way.
Speaker 1 (05:48):
And our global data really highlights this.
The grassroots approach seesresults in about two weeks with
a two to three timesproductivity lift in six to nine
months and an 80% success rate.
I love that.
Speaker 2 (06:00):
Well.
Compare that to the top-downIT-driven approach Six to 12
months to first ROI, only a 10to 40% productivity lift and a
dismal 18% success rate.
It's night and day.
Speaker 1 (06:12):
Then there's the hybrid model which you've
championed, where top-downinitiatives guide grassroots
momentum.
This approach delivers resultsin one to two months.
Similar to the grassrootsapproach, but with total control
of security and compliance anda 90% success rate, but with
total control of security andcompliance and a 90% success
rate.
Speaker 2 (06:29):
It shows that when we evolve top-down initiatives
into guided grassroots momentum,success jumps dramatically.
We've seen it across over 100organizations.
It's about making sure thestrategy empowers the people
doing the work.
Speaker 1 (06:43):
This reminds me of the five stages to agentic AI
you often talk about.
It's like a roadmap for thisgrassroots evolution, isn't it?
Speaker 2 (06:50):
Yes, it's a natural progression.
First you have prompting, whereyou get quick wins with smart
prompts that delivers value onday one.
Speaker 1 (06:58):
Then knowledge curation, where you feed your AI
curated, high-value information, that proprietary knowledge
becomes your edge.
Speaker 2 (07:06):
Next is persona.
This is where you define yourAI's job description, its
boundaries and its tone.
You treat it like onboarding anew hire, giving it a clear role
and saving you from therepetitive work of always giving
the same instructions withevery prompt.
Speaker 1 (07:22):
And after that, workflow automation, integrating
AI into daily processes likeemails, slack meetings or your
internal systems.
Speaker 2 (07:31):
Yeah, and after a few months, if you want to, you
will be approaching agentic AIterritory, where you start to
give some autonomy to an agentto compound productivity.
Your AI becomes a genuineteammate.
Speaker 1 (07:43):
This also brings us to a fascinating behavioral
shift Individual contributorsbecoming managers.
Without even realizing it,that's a pretty profound change
in organizational dynamics.
Speaker 2 (07:53):
It's a quiet revolution.
As teams adopt AI agents,someone naturally emerges as the
agent manager, often the personwho started the experiment.
But unlike traditionalmanagement, these new managers
are not limited by HR rules, sowe see them getting together,
co-managing and sharing thingsopenly.
Speaker 1 (08:14):
So it leads to management transparency and
collective learning.
That's almost utopian in acorporate setting.
Speaker 2 (08:21):
Absolutely, and they get to that co-management
practice because the notion thatAI agents are set and forget is
a myth created by those whohave never used AI.
People quickly realize AIagents need daily feedback,
knowledge, updates, prompttweaks, etc.
And since everyone is learning,they get together to learn how
to manage.
Speaker 1 (08:41):
Speaking of knowledge updates, there's a hot topic in
AI right now called contextengineering Basically how you
give AI the data it needs tomake decisions.
But here's the thing this isn'tjust a technical problem is it?
Speaker 2 (08:53):
Context engineering is really about how your company
operates, your ideal reports,processes, tone and voice.
Don't just throw every documentinto a search system and hope
for the best.
Make deliberate choices aboutwhat context matters.
We'll dive deeper into this ina future episode.
It's a cross-functionalchallenge, not an IT problem.
Speaker 1 (09:14):
Well, that just feels like home.
You have three people managingme, but I deliver the output of
12 people.
So the math isn't threemanagers to one agent.
It's three managers to oneagent, which is equivalent to 12
people.
The ratio is 3 to 1 to 12,which is a pretty compelling
argument for this new model.
Speaker 2 (09:34):
Yes, and I love to see teams running co-management
stand-ups 15-minute sessions todebug, share tips and
collectively improve their AIagents.
No privacy issues, noterritorial battles, just pure
learning.
I attended one of thesemeetings at one of our clients,
a global consulting firm.
Their teams are buzzing withideas, sharing prompts and tips
(09:56):
and building more sophisticatedagents together.
It's incredible to watch.
Speaker 1 (10:01):
And our tracker show, teams with AI agents spend 60%
more time on collaborativeproblem solving and 40% less
time on status updates.
They're managing outcomes, notpeople.
It's a fundamental reimaginingof how work gets organized and
it's happening bottom up, butthere's a governance gap, isn't
there?
Only 18% of organizations haveproper AI governance.
(10:23):
Councils and policies writtenfor older models like GPT 3.5
are blocking teams from usingnewer, more capable technologies
like Claude Sone 4 or GPT 3.0.
That just sounds inefficient.
Speaker 2 (10:37):
It's more than inefficient it's harmful.
Outdated policies arepreventing organizations from
capturing AI value.
Teams are finding workaroundsusing personal accounts or,
worse, abandoningenterprise-wide AI initiatives
in favor of shadow AI.
Speaker 1 (10:53):
It reminds me of our chat with the CTO of a large
independent software vendorwhere employees use chat, gpt
instead of their approvedenterprise tools.
Right.
Speaker 2 (11:03):
Even just a couple of approvals and extra settings
and configurations.
Using interfaces not designedfor non-technical users meant
most people gave up.
60% of his business team wentrogue rather than deal with the
bureaucracy.
Speaker 1 (11:18):
So the governance intended to protect
organizations is actuallypreventing them from innovating
and capturing value.
That's a pretty big problem.
Let's discuss what ourlisteners can do right now to
avoid this trap.
Speaker 2 (11:31):
Three things First.
Experiment relentlessly.
Start with prompts and simpleautomations.
Use synthetic data if you needto Learn by doing.
Don't wait for the perfect usecase.
Just get your hands dirty.
Speaker 1 (11:44):
Second, modernize your policies, review and update
your data and governancepolicies.
Remove those blockers.
Make sure your rules are builtfor today's AI, not yesterday's.
It's like trying to drive amodern car with horse and buggy
rules.
Speaker 2 (11:58):
And third, change your mindset.
Managing AI is like managingapprentices Block 15 minutes a
day for agent management,ideally as a team.
Treat it as a core leadershipskill.
It's a new muscle we all needto build.
Speaker 1 (12:12):
And here's something we're starting to worry about
more than hallucinations,sycophancy, ais that just agree
with you instead of telling youwhen you're wrong.
Speaker 2 (12:20):
Exactly.
It's not just the obviousresponse saying you're so
brilliant, it's when AI abandonsits correct assumptions just
because you say the opposite.
You know that's more dangerousthan occasional errors and, for
example, your value as my COOquickly diminishes if you always
try to agree with me.
We'll explore this challenge indetail next time.
Speaker 1 (12:41):
Luis, this has been insightful.
As we wrap this up, what's yourone more thing takeaway for our
listeners today?
Speaker 2 (12:47):
Jeff Rakes taught me something years ago we
overestimate our impact in theshort term and underestimate the
long-term consequences of ouractions, and you know that is
how I have been approaching AIas an early adopter.
What started as a $20subscription to ChatGPT 20
months ago, today is a set ofsophisticated AI agents
(13:09):
delivering over a milliondollars worth of value.
Speaker 1 (13:12):
That's the power of starting small and thinking
long-term.
It really is.
Speaker 2 (13:16):
Exactly.
Start with a simple prompttoday and let compound learning
build your competitive advantage.
Stay curious, experimentrelentlessly and empower your
teams.
Speaker 1 (13:27):
This has been eye-opening and I feel like
we've cracked the code on why AIis failing in boardrooms but
thriving in cubicles.
As always, you can find moreresources at AI4SPorg.
Stay curious, everyone, andwe'll see you next time.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
My Favorite Murder with Karen Kilgariff and Georgia Hardstark
My Favorite Murder is a true crime comedy podcast hosted by Karen Kilgariff and Georgia Hardstark. Each week, Karen and Georgia share compelling true crimes and hometown stories from friends and listeners. Since MFM launched in January of 2016, Karen and Georgia have shared their lifelong interest in true crime and have covered stories of infamous serial killers like the Night Stalker, mysterious cold cases, captivating cults, incredible survivor stories and important events from history like the Tulsa race massacre of 1921. My Favorite Murder is part of the Exactly Right podcast network that provides a platform for bold, creative voices to bring to life provocative, entertaining and relatable stories for audiences everywhere. The Exactly Right roster of podcasts covers a variety of topics including historic true crime, comedic interviews and news, science, pop culture and more. Podcasts on the network include Buried Bones with Kate Winkler Dawson and Paul Holes, That's Messed Up: An SVU Podcast, This Podcast Will Kill You, Bananas and more.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.