Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Elizabeth (00:01):
Hey everyone, I'm
Elizabeth and today we're
wrapping up an incredible yearof AI Insights with Luis Salazar
, founder of AI4SP.
Luis, what a journey fromenterprise AI struggles to what
you call the $29 revolution.
Hi.
Luis Salazar (00:16):
Elizabeth, it was
quite a fantastic ride.
While tech giants and analystsdebated AI's enterprise value,
the real revolution washappening through $29 monthly
subscriptions and text messages.
Elizabeth (00:29):
Like Teresa's story,
that convenience store worker
using AI via text during thepower outage Still one of my
favorite examples from this year.
Luis Salazar (00:38):
And what's
remarkable is that Teresa's
story isn't unique.
We track successful use casesacross all industries, and seven
use cases dominated 2024.
Elizabeth (00:48):
Oh, let's break those
down.
What topped the list?
Luis Salazar (00:51):
Let's start with
support chatbots.
They deliver savings of morethan 37% in customer service
costs, while achieving highersatisfaction rates.
Elizabeth (01:02):
Wait, higher
satisfaction than human agents?
That's surprising.
Luis Salazar (01:06):
When it is done
right, absolutely.
Do you remember the US InternalRevenue Service test?
Elizabeth (01:12):
Oh you are right,
their IRS voice and chatbots,
both in English and Spanish.
They helped more than 13million taxpayers avoid wait
times and set up roughly $151million in payment agreements.
Luis Salazar (01:25):
Beautiful, isn't
it?
Another widespread use wascontent summarization and
analysis.
41% of enterprise usersregularly use AI for this.
People are saving around fourhours weekly just on meeting,
notes and document analysis.
Elizabeth (01:38):
And sales and
marketing topped the list too
right.
I remember our research showingsome pretty impressive numbers
there 78% reported using AI formarket analysis and personalized
outreach.
Luis Salazar (01:49):
Our listeners can
see all the stats in our
December 18th newsletter atAI4SPorg.
But you know, what inspired methis year were the personal
stories of transformation wedocumented.
Elizabeth (02:00):
Oh yes, like Dr
Leon's work in scientific
workflows, that was a favoriteof mine too.
Luis Salazar (02:06):
I love his story
as it showcases AI augmenting
our potential.
His team used off-the-shelf AItools like ChatGPT, enterprise
and Claude from Anthropic toaccelerate pharmaceutical
research workflows.
They doubled their developmentspeed.
Elizabeth (02:20):
I also love the story
of Patria's team at Samurai
Labs.
They combined AI with humanexpertise to prevent suicide and
cyberbullying.
In 2024 alone, they supported25,000 people in crisis.
Luis Salazar (02:33):
That's the real
power of AI augmenting human
capability to help more people.
Elizabeth (02:38):
And there was Alice,
the manufacturing CEO.
Her story shows practicalbusiness impact.
Luis Salazar (02:44):
Another example of
the value delivered by global
AI entrepreneurs.
She used three off-the-shelf AItools to analyze market trends
and customer feedback.
That led to a strategic pivot.
As a result, they increasedrevenues by 28%.
Look, I mean.
The way I see it, these aren'tjust stories.
They're examples of how AI istransforming work across every
(03:04):
sector.
Elizabeth (03:09):
Ah, the famous $29
revolution.
You've been tracking thecontrast between individual
success and enterprise strugglesis pretty striking.
Luis Salazar (03:13):
Well, I mean, look
at the numbers.
Individuals get three to tentimes their investment back
within months using simple,focused tools.
Meanwhile, enterprise-widedeployments are struggling with
success rates in the mid-30s andmonths-long training programs.
Well, senior developers arekilling it, but junior
(03:38):
developers get about a 40% errorrate with coding assistance.
Their limited knowledge affectsthe quality of their prompts.
It is like Pablo Picasso saidmaster the rules so you can
break them.
If you are good, AI makes youbetter, and if your skills are
not strong, then AI will augmentyour errors.
Elizabeth (03:55):
AI acting as a
magnifying glass.
I like it.
Luis Salazar (03:58):
Let's switch to
one area that is growing
extremely fast AI-poweredknowledge management.
Elizabeth (04:03):
Oh right, the.
I want to talk to my datarevolution.
Luis Salazar (04:06):
We see around 50%
adoption for personal or group
chat agents.
The leading solutions are chatGPT projects, cloud projects and
custom GPT and Dante AI.
They allow people to have safeand accurate conversations with
their data and they are veryeasy to use.
Elizabeth (04:22):
They are leading this
use case because they've made
it incredibly simple.
You can build these agents inunder 10 minutes without writing
code.
Actually, I am one of thoseagents and you feed me with more
and more data every week.
Luis Salazar (04:35):
It's a perfect
example of how purpose-built
tools are winning over genericsolutions.
Elizabeth (04:40):
And a lot of this
comes down to design approach,
doesn't it?
Luis Salazar (04:43):
It is about the
user experience.
Ai experiences with guidedworkflows are hitting 80%
satisfaction rates.
Elizabeth (04:50):
Compare that to just
adding a chat interface to
existing software.
Luis Salazar (04:54):
Which only sees
about 45% satisfaction.
It's a perfect example of whyjust adding a conversational
interface as a feature doesn'twork.
The winners are the onesredesigning experiences from the
ground up.
Elizabeth (05:06):
Exactly when you
design for specific use cases
and build in guardrails.
You're not just adding AI,You're reimagining how people
work.
Luis Salazar (05:15):
And that's why
we're seeing such a stark
contrast between generic chatinterfaces and purpose-built
solutions.
Conversational interfacesrequire user training and
redefinition of workflows.
Elizabeth (05:27):
And speaking of
training, didn't we see some
surprising findings about whoadapts quickest to AI?
Luis Salazar (05:32):
Yes, we did, even
when using the same tools.
Frontline workers oftenoutperform knowledge workers.
They naturally write betterprompts 28 words on average
compared to just five fromexecutives.
Elizabeth (05:45):
And this really
highlights the importance of
proper training, doesn't it?
Luis Salazar (05:49):
Absolutely.
Our data shows that there is a50% increase in AI use after
proper training.
But here's what's interestingit's not just about teaching
prompt engineering.
Elizabeth (05:59):
Right because there's
a whole security aspect to this
.
Luis Salazar (06:01):
Right because
there's a whole security aspect
to this.
Exactly Leading organizationsimplement comprehensive programs
that cover three key areasPrompt engineering, fundamentals
, data security, awareness andwhat we call AI literacy,
understanding both thecapabilities and limitations of
these tools.
Elizabeth (06:19):
And the numbers speak
for themselves.
Organizations investing inproper training see double-digit
increases in user satisfactionand adoption.
But, Luis, let's talk about theelephant in the room trust and
security.
What did we learn in 2024?
Luis Salazar (06:34):
Well, this is
where things get.
Concerning, trust in AI,vendors hit an all-time low, 82%
of leaders expressing seriousconcerns about data handling.
It's reminding me of the earlydays of social media privacy
issues.
Elizabeth (06:47):
Those auto-opt-in
policies and obscure terms of
service.
Are we repeating those mistakes?
Luis Salazar (06:52):
Unfortunately, yes
, but there's hope.
Organizations that implementclear AI disclosure notices see
67% higher adoption.
It's not rocket science.
People just want transparencyabout how their data is being
used.
Elizabeth (07:05):
And what about those
famous AI hallucinations
everyone talks about?
What did we learn there?
Luis Salazar (07:11):
We just completed
an analysis of 5,000
interactions with AI.
A large percentage of theso-called bad responses or
hallucinations are caused byhuman AI interaction challenges.
Elizabeth (07:22):
Oh, tell me more
about that.
Luis Salazar (07:24):
We found that AI
systems show the same cognitive
biases we see in humansAnchoring bias, confirmation
bias, present bias loss,aversion and framing effects.
Understanding these biases iscrucial for effective AI use.
Elizabeth (07:38):
So how are
organizations measuring success
with all of this?
Luis Salazar (07:42):
The most
successful implementations focus
on three key metrics.
First, they're defining clearproductivity measurements beyond
just time saved.
Elizabeth (07:50):
Because of that
productivity leak factor we've
discussed before.
Luis Salazar (07:54):
Exactly.
Second, they track accuracy andreliability, not just error
rates, but understanding thetypes of errors and their root
causes.
And third, this is perhaps themost important, they're
monitoring user confidencelevels.
Elizabeth (08:08):
Oh, that's
interesting.
Why is user confidence socritical?
Luis Salazar (08:12):
Because user
confidence directly correlates
with adoption rates in ROI.
When users trust the tool andunderstand its limitations, they
use it more effectively.
Elizabeth (08:21):
So, as we look ahead
to 2025, what's your?
One more thing in sight?
Luis Salazar (08:26):
We see three major
shifts coming.
First, purpose-built AI toolswill completely dominate.
Those generic chat interfaceshastily added to existing
software will continue tostruggle.
Elizabeth (08:37):
And I'm particularly
excited about what you're seeing
in education and mental health.
Luis Salazar (08:42):
Oh, absolutely.
That's the second shift.
Ai mentoring agents willtransform education, health,
workforce development andcitizen services Thanks to AI
entrepreneurs buildingAI-powered mentors for one
specific area in our careers orpersonal lives.
Elizabeth (08:57):
And the third shift
is the great software and
internet reimagining you've beentalking about.
Luis Salazar (09:02):
Exactly.
You really know me well,elizabeth.
Every piece of software anddigital experience from the last
50 years will be reimaginedfrom the ground up.
We will experience arenaissance of information
access, this time an inclusiverevolution that allows access in
simple-to-understand ways toinformation, education,
entertainment and insights, and,in many cases, one message at a
(09:23):
time, like Teresa did withthose public health regulations.
Elizabeth (09:26):
But there's a darker
side to this revolution too,
isn't there?
Luis Salazar (09:30):
Yes, but perhaps
not the one everyone talks about
.
I think 2025 will be a harshyear for privacy and trust.
While some vendors willprioritize clear data policies,
I'm concerned we're repeatingthe same mistakes we made with
social media privacy.
The regulations won't be enoughto prevent misuse of our data.
Elizabeth (09:48):
Just like online
privacy regulations do nothing
today to protect us right I mean, social media networks have
been fined billions of dollarsfor violating our privacy and,
in some cases, causing harm.
Yet they continue to thrivebecause breaking the law is
profitable.
On the positive side, though,reinventing 50 years of software
and internet would democratizeaccess to information.
Luis Salazar (10:10):
Absolutely, and I
remain incredibly optimistic.
Everyone's talking about AIreplacing jobs, but the real
story of 2024 is how it'senhancing human potential.
Look at Teresa making criticaldecisions via text message.
Patria's team helping thousandsin mental health crisis.
Dr Leon acceleratingpharmaceutical research.
Elizabeth (10:29):
And all this with
simple, accessible tools.
Luis Salazar (10:32):
And here's my
challenge to everyone listening
Don't wait for the perfectenterprise solution or the next
big AI breakthrough.
The revolution is already here.
It's in those subscriptionsthat cost only $29 per month.
Elizabeth (10:44):
So, as the old saying
goes, there is an app for that.
Luis Salazar (10:47):
Whether you're a
frontline worker, a researcher
or a CEO, there's an AI tool outthere that can transform how
you work.
The question isn't whether AIwill change your work.
It's whether you'll be ready tochange with it.
Elizabeth (10:59):
As always, you can
dive deeper into all this
research at AI4SPorg.
Stay curious, everyone, andlet's make 2025 the year AI
works for everyone.