Building AI That People Use: How entrepreneurs win over tech giants in the first leg of the AI Enterprise adoption

Share your thoughts with us

What if the tools designed to revolutionize your workflow are the very ones holding you back? This episode features Luis Salazar, founder of AI4SP, who sheds light on the staggering gap between AI adoption rates and user satisfaction, particularly in behemoths like Microsoft and Google. While corporations flounder in trying to retrofit AI onto legacy interfaces, AI-native companies such as Canva, CustomGPT, DanteAI, Khan Academy, Synthesia, and others are breaking new ground, offering a glimpse into an AI future where users become adept in days, not months. Join us as we dissect the discrepancy between seamless AI integration marketing promises and the gritty reality of skill gaps and change management challenges.

🎙️ 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

Episode Transcript

Elizabeth (00:00):
Hey everyone, I'm Elizabeth, and today we're
discussing how Google,salesforce, microsoft and other
large tech companies strugglewith low adoption rates from
their AI add-ons.
At the same time, a newgeneration of entrepreneurs
shows us how to build AI thatpeople actually use.
Luis Salazar, founder of AI4SP,has been tracking this pattern

(00:21):
across thousands oforganizations.
Hello everyone.

Luis (00:25):
As Elizabeth said, today we will talk about a fundamental
flaw in innovation.
Microsoft's approach withCopilot, google's with Gemini
and similar enterprise solutionsall make the same basic mistake
they're trying to patch AI ontointerfaces designed around
menus and search boxes a50-year-old paradigm.
Yes, 50 years of graphical userinterfaces 50 years of habits

(00:49):
that will not change in a blink.
Across 1,000 organizations,only 12% have succeeded in
deploying ChatGPT, enterprise orCopilot, while the success rate
of single-purpose AI tools isat the other extreme.

Elizabeth (01:02):
And Gartner's findings from large enterprises
are even more sobering, aren'tthey?

Luis (01:07):
Well, here's what's fascinating about these numbers.
While our broader study showsslightly better success rates
across organizations of allsizes, including smaller
companies and nonprofits, we'restill seeing the same
fundamental pattern.
Whether it's Microsoft Co-Pilot, chatgpt Enterprise or Claude
from Anthropic, theseone-side-fits-all approaches to

(01:27):
AI are struggling to deliver ontheir promises.

Elizabeth (01:30):
Those adoption rates tell a clear story, but what's
even more revealing is the usersatisfaction data you're
tracking.
What's particularly interestingis how this contrasts with our
recent coverage of specializedAI tools.

Luis (01:42):
Absolutely.
When we look at usersatisfaction rates, the
difference is dramatic.
These chat interfaces from techgiants see satisfaction rates
of around 45%, but purpose-builtAI tools they're hitting 84%
satisfaction rates.

Elizabeth (01:57):
That's a stark difference and, if I recall from
our October findings aboutfrontline workers, it's not just
about the technology, it'sabout the entire user experience
, right.

Luis (02:09):
It is about the user experience and about creating
the right expectations.
The marketing from all majorvendors Microsoft, google,
openai promises a seamlessintegration and no learning
curve, but in reality there is along learning path ahead.

Elizabeth (02:23):
And our research shows a striking disconnect
there.
90% of CTOs and CEOs reporthigh employee enthusiasm for
these tools.
But then what happens?

Luis (02:33):
That's where reality hits.
80% of employees struggle tointegrate these tools into their
daily workflows, and here's thehidden cost nobody talks about.
87% of users need ongoingsupport and training,
particularly in promptengineering, where we're seeing
massive skill gaps.

Elizabeth (02:50):
The change management costs must be staggering.
How are organizations handlingthis disconnect between
expectations and reality?

Luis (02:58):
Most aren't prepared at all.
80% of organizations severelyunderestimate the change
management needed.
They budget for licenses butnot for the fundamental
workforce transformationrequired.
And, what's really telling?
Traditional training approachesaren't working.

Elizabeth (03:13):
Right.
Watching videos about promptengineering isn't the same as
learning a new way to work.
What do you see in terms of AIadoption expectations versus
reality?
There is a big gap.

Luis (03:23):
While vendors promise immediate productivity gains,
we're seeing it takes three, sixmonths for users to become
proficient.
That's a far cry from theplug-and-play promise in their
marketing materials.

Elizabeth (03:34):
Those adoption timelines are revealing, but I'm
curious about companies thatare getting it right.
You mentioned in our previousepisodes how some companies are
taking a completely differentapproach to AI integration.
How are they achieving suchdifferent results?

Luis (03:48):
When AI is built into the core experience from the ground
up, users become proficient indays, not months.
Companies like Canva, a leadingproductivity software with
hundreds of millions of users,are showing how to do this right
.
They're not just adding AIfeatures, they're reimagining
the entire user experience.

Elizabeth (04:09):
And the market is responding to this approach.
You mentioned AI-nativecompanies are growing twice as
fast as traditional softwarecompanies.

Luis (04:17):
Yes, and here's why Look at how these specialized tools
approach real problems.
Instead of promising totransform everything, they focus
on specific challenges.
A non-profit developmentofficer using AI for donor
communications gets personalizedoutreach in minutes.
A sales team gets proposalsthat actually reflect their past
win rates.
Users of these niche AI toolsbecome proficient in hours or

(04:39):
days, not months.

Elizabeth (04:41):
And these aren't just faster, they're delivering
better results.
Right, I'm thinking about whatyou shared last week about
Teresa, the convenience storeworker using AI during that
power outage.

Luis (04:52):
Well, let me put it this way Teresa did not need to learn
anything new, because the AItool was designed to mimic the
way she works.
Teresa didn't need to learnprompt engineering she just
texted in Spanish about herrefrigeration problem and got
clear guidance based on FDAprotocols.
Another great example iscustomer service teams using

(05:13):
focused AI tools to craftresponses.
They're seeing 90% satisfactionrates because the tool
understands their specificindustry and common issues.

Elizabeth (05:24):
So, while tech giants promise to transform every
aspect of work, thesespecialized tools are quietly
revolutionizing specificworkflows.

Luis (05:32):
That is right and think about grant writers using AI to
analyze successful proposals andcraft stronger applications, or
construction teams gettingimmediate safety protocol
guidance on their phones.
While enterprise solutions seeless than 5% reaching
company-wide deployment, thesesingle-purpose solutions see
adoption rates in the mid-doubledigits because they solve real

(05:55):
problems in ways that make senseto use it.

Elizabeth (05:58):
It's fascinating how these solutions succeed by
thinking smaller but deliveringbigger results.

Luis (06:04):
Exactly.
They're not trying to boil theocean.
They're just making specifictasks dramatically better, one
workflow at a time.

Elizabeth (06:11):
So what's the takeaway for organizations
trying to navigate thislandscape?

Luis (06:15):
First, be very skeptical of one-size-fits-all solutions,
even from leading tech companies.
Look at specific workflows youwant to improve and consider
purpose-built tools and, mostimportantly, understand that
adding AI isn't like adding anew feature it's fundamentally
changing how people work.

Elizabeth (06:31):
And for software creators.
This seems to demand a wholenew approach to building
software.
What are you seeing on thatfront?

Luis (06:38):
Here's our one more thing.
The next wave of software won'tjust add AI features to old
interfaces.
It requires a fundamental shiftin how we think about human-AI
interaction.
At AI4SP, we learned thislesson early.
Our team includes philosophers,neuroscientists,
anthropologists and behavioraleconomists, alongside our tech
talent.

Elizabeth (06:58):
It's like what Jeff Rakes said about the evolution
of computer sciences and whatStanford University is doing
with its symbolic systemsprogram.

Luis (07:05):
Right, Exactly, Stanford recognized that bridging the
human-AI gap requires more thanjust technical expertise.
They're combining computerscience with linguistics,
psychology and philosophy.
It's about understanding howhumans think and communicate,
not just how machines work.

Elizabeth (07:21):
This reflects what we're seeing in successful AI
tools they're designed withhuman behavior in mind, not just
technical capabilities.

Luis (07:29):
Right.
When we studied why native AItools achieve 84% satisfaction
rates, we found it wasn't justbetter technology.
It was a better understandingof human behavior.
Software teams won't just needengineers.
They'll need linguists whounderstand how people naturally
express needs.
Engineers.
They'll need linguists whounderstand how people naturally

(07:50):
express needs, psychologists whograsp cognitive patterns and
anthropologists who can map howwe interact with technology.

Elizabeth (07:54):
And, as always, if you want to dive deeper into
this research, you can find allthe details at AI4SPorg.
Stay curious, everyone, andwe'll see you next time.