Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Andreas Welsch (00:00):
Today we'll talk
about how you can use Generative
(00:02):
AI to improve your userexperience in your applications.
And who better to talk to aboutit than someone who's been
teaching hundreds of teams on AIand analytics over the years?
Scott Clendaniel.
Hey, Scott.
Thank you
T. Scott Clendaniel (00:13):
So much for
joining.
Yeah, I am so excited to behere.
I can hardly contain myself.
Thank you for inviting me.
Andreas Welsch (00:20):
Awesome.
Hey, why don't you tell us alittle bit about yourself, who
you are and what you do.
T. Scott Clendaniel (00:24):
Sure.
So first of all, I am older thandirt.
I actually started in the fieldin March of 1986, so I've been
doing this for quite a while.
I absolutely love helping peoplelearn all about artificial
intelligence, machine learning,data science, all those fields.
Done a heck of a lot ofconsulting.
(00:45):
And I've had clients you mayhave heard of Audi, Biogen,
Mercedes, Los Angeles Times,folks like that.
And so now I'm trying totransfer that knowledge to
others and bring more peopleinto the field.
Andreas Welsch (00:58):
Scott, should we
play a little game to kick
things off?
T. Scott Clendaniel (01:00):
Okay.
Andreas Welsch (01:01):
Alright,
wonderful.
T. Scott Clendaniel (01:02):
Do I win a
free car?
Andreas Welsch (01:04):
Probably not.
T. Scott Clendaniel (01:05):
Okay.
Alright we'll do it anyway.
Andreas Welsch (01:07):
But how about
reputation and reputational
points?
Excellent.
All right.
Look, so this, game is called InYour Own Words, and when I hit
the buzzer, the wheels willstarts spinning.
When they stop, you'll see asentence.
I'd like you to answer with thefirst thing that comes to mind
and why, in your own words.
And to make it a little moreinteresting, you'll only have 60
seconds for the answer now.
(01:27):
Are you ready for What's theBUZZ?
T. Scott Clendaniel (01:29):
As ready as
I'm going to get I am.
Andreas Welsch (01:31):
Okay, perfect.
Here we go.
If AI were a song, what would itbe?
60 seconds on the clock, go.
T. Scott Clendaniel (01:40):
If AI were
a song, what would it be?
I'll tell you what it would be:
Help! By the Beatles because (01:45):
undefined
that's where we're going withthis, help.
I need some info help, not justany info.
How would that be?
Andreas Welsch (01:57):
That would be
awesome, right?
We all can use a little morehelp.
Fantastic.
And well within time, I've seena lot of creative answers.
But, help is, perfect.
Why don't we talk about theactual topic of the show around
user experience.
Look, I think we've allobviously seen Generative AI in
(02:19):
the news last year, still in thenews this year.
Good thing.
I think this year it's moreabout where the rubber meets the
road.
Implement AI, get your pilotsout of your lab into production.
But I also see that it obviouslydemocratizes access to
AI-powered applications andthose insights.
So whether you're a seniorexecutive or you're a first-year
(02:39):
student, English, it is the newprogramming language, which
makes it so universallyapplicable and accessible.
But I wonder, from yourperspective, what's so
remarkable about Generative AIwhen it comes to user
experience?
T. Scott Clendaniel (02:53):
All right,
so let me throw a question out
to you and the audience atlarge: when you go to send an
email.
Andreas, we're gonna start withyou.
Yeah.
What programming language do youuse?
Andreas Welsch (03:05):
Good question.
I don't know what's under thehood, but
T. Scott Clendaniel (03:09):
So when you
go to the web, which programming
language do you use?
You probably just use a browser,right?
Yeah.
When you go to create apresentation, you use some type
of app.
So the whole process of creatingsoftware and the point of
creating software was to putthings into apps so that people
didn't have to reinvent thewheel over and over again.
(03:30):
But somehow that memo seems tohave been skipped to our friends
in the artificial intelligencecommunity, whereas, no, you have
to be a ninja.
In Python to do anything?
No, you have to be able to dothis.
You have to be able to code, youneed to know IDEs.
So not only am I gonna teach youall the statistics, but I'm also
gonna teach you computer scienceand all these obscure languages
and all this kind of stuff.
(03:52):
Really, we don't require on yourdriving test to be able to
understand compression ratios ona gas pass powered engine,
right?
Or at least mine didn't.
So the whole process ofapplications to simplify things
and make things easier.
It's also about user experience,but programmers, a small group
(04:15):
of folks, not programmers, butbro-Grammers.
Thank you to Nicole for givingme that phrase is the fact that
no.
You have to be able to doeverything I can do before you
can touch your information andthen consider the screen that
you see when you go to ChatGPT,as an example.
It's largely a blank screen witha place to type.
(04:38):
You don't need to code anything.
You don't need to know how to doit.
Even GUI interfaces, it'sbasically you type your answer
and hit enter.
And so the beauty is people whohave been scared off from our
field for a long time now haveaccess to things that they never
would've had before.
And I think it's remarkable.
A lot of the stuff that AI doesnow, it did before.
(05:00):
It did search before you couldgenerate text before.
So what's the difference?
Why is it so popular now?
And I'm going to be reallycontroversial here and I'm gonna
say it is not so much thatartificial intelligence has
improved, it's that we'vesimplified the user interface.
Andreas Welsch (05:16):
I love that.
I remember working withcustomers a couple years ago on
AI first pilot projects.
We didn't call it AI then, wecalled it machine learning,
because we wanted to be veryprecise and and accurate.
And accuracy, for example, wasone of the variables or, one of
the piece of information that wewanted to display to business
(05:37):
users.
And they saw 80% accuracy and Ithought, that's awesome.
And they tried it out and theresults were garbage.
I said actually, if it's 95 or97%, that's when it gets really
good and 99%, that's where youwanna be.
And they couldn't believe that80% was garbage.
To your point, if you don't,absolutely, you need to expose
(05:58):
that if you, make it simpler tobegin with to interact with
these systems.
That's a huge opportunity.
T. Scott Clendaniel (06:03):
That's well
and it's funny'cause you even
talk about the labels.
We now call it artificialintelligence.
Back when I started, it was datamining and no one uses that
phrase anymore.
And then it was KDD.
Knowledge, discovery anddatabases, and then it was
predictive analytics.
No, that's not good enough.
No.
It's gonna be machine learning.
No, it's gonna be data science.
Lots of the stuff has been thesame the whole way through, but
(06:24):
we keep changing the name justto make things harder for
people.
And that's a great example.
If you type in any of thosethings, you're gonna get similar
results back.
Because we have removed thatcomplexity.
Andreas Welsch (06:36):
No I, really
love that.
And, again for, business users,I think it makes it so much
easier to finally get theresults and get useful results
without the complexity of havingto understand the technical
details in, the background.
So by the way, for you in theaudience, if you have any
questions for Scott, feel freeto pop them in the chat and
(06:56):
we'll take a look in a coupleminutes and take some of those
as well.
So speaking of business users,what do you think there is the
potential for them, maybe what'san unexpected aspect of large
language models that you've seenfor improving the user
experience there?
T. Scott Clendaniel (07:13):
I think a
lot of it is it opens doors.
And what I mean by that is I cantell you that I, myself and
probably the worst graphicdesigner in the history of the
universe.
I may have the heart of aDaVinci, but I have the skill
set of a one-armed troll with areally bad hangover.
(07:34):
So I can have ideas, oh, I'dlike to have this image to be
able to convey this message, anactual business use case.
I wanna convey this message toothers, and it's very
complicated, and I'm trying tothink of how to do it.
There is no way in hell youcould give me all the tools in
the world and that was not gonnabe helpful.
But now I can go into one of theGenerative AI tools and say,
okay, I'm looking for somethingthat sort of looks like this.
(07:56):
I can prototype all kinds ofdifferent things, and I'm not
even saying that's gonna be thefinal result, but the amount of
time savings, I wouldn't haveeven ventured into that realm
before.
Now play around.
I get a little moreself-confidence.
I'll try refining my promptsover time.
I never would've done that.
I love adobe Photoshop and allthose products, they're lovely,
(08:19):
but, yeah, that was not gonnahelp.
This removes that fear andremoving that fear allows
business users to do all typesof tasks that they would've been
afraid to approach before.
Andreas Welsch (08:31):
Yeah.
I like how you, also broaden thedefinition, if you will or the
variety of Generative AI if you,in business, a lot of the use
cases that we see today arestill focused on text and text
generation, summarization.
What are the key points out ofthis meeting, out of the meeting
minutes or even things like Bardor now Gemini.
(08:54):
Summarize this YouTube video forme.
If you can access it.
T. Scott Clendaniel (09:00):
Absolutely.
Let's just take a simpleexample.
It is very hard for most peopleto sit in a series of meetings
back-to-back.
There's been all kinds ofresearch that even a
fifteen-minute separationbetween meetings does wonder,
the human brain to be able torecover just a bit when you're
even have something as simple asan AI assistant in a meeting.
To take the notes, I have toworry less.
(09:23):
That I'm typing and writing andsummarizing the point and
listening and coming up with myresponse all at the same time.
It seems simple'cause we have todo it all the time, but it does
get complex.
Instead, I can really listen towhat's happening and form my
response and not be justobsessed with note-taking.
It's frees up that cognitiveload for me to do the things I
(09:43):
hope I'm better at.
Andreas Welsch (09:45):
I'm really
excited to see where this is
going.
Maybe even get some coaching.
What are the questions that Ishould be asking, following the
conversation?
And having context about me.
Yeah.
T. Scott Clendaniel (09:56):
Can I share
a magic trick
Andreas Welsch (09:57):
with you?
Oh, only by exception, yes.
T. Scott Clendaniel (10:00):
Okay,
perfect.
Actually, one of the things thatif you are new to Generative AI,
one of the beautiful things isif you get stuck, you can ask
your question about GenerativeAI.
In ChatGPT or whatever, orGemini or whatever your weapon
of choice is.
If you are stuck or you get aninstruction that you don't fully
understand, you can clarify andyou can say what do you mean by
(10:22):
that?
And a lot of times studentswon't do that in a full
classroom setting because theydon't wanna feel embarrassed.
They don't wanna feel like, oh,my boss is gonna think I'm an
idiot if I didn't catch thatlast thing.
But if you're in something likeGemini, you can actually Gemini
for.
Use Gemini for some tips on howto create better Gemini
products.
(10:43):
That is one of the things that Ithink just opens up all kinds of
possibilities.
Andreas Welsch (10:48):
I love just the
power, right?
That, now is, at our fingertips.
It's not the first time that wealso hear it's all about
language.
A couple years ago we've talkedabout chatbots as the new thing,
as the new interface for voicecommands in the enterprise and
what have you.
And I think it hasn't quitehappened yet whether it's in an
(11:08):
office where you have an openspace and everybody's talking to
their assistant.
It gets a little loud andannoying maybe or let alone
field service.
T. Scott Clendaniel (11:17):
I've been
in those environments.
Yes, I understand.
Andreas Welsch (11:20):
All of a sudden
you trigger your your desk
neighbor's assistant to dosomething or whatever that maybe
we've seen that with GoogleAssistant, with Alexa in our
homes when they suddenly go off.
Or now even let alone fieldservice or on an oil rig where
it's just impractical thatsomebody speaks to that system.
Given all the environmentalconditions and noise and
(11:42):
everything.
So what do you think isdifferent this time then when it
comes to language, and voice,large language models?
What's different this timecompared to chatbots from a
couple years ago?
T. Scott Clendaniel (11:56):
I'm so glad
you asked that question because
fortunately I was prepared witha response because that is this
guy, your cell phone.
Why?
Because, most people in the past10 years have given up to a
large extent on calls back andforth within the office.
They've just surrendered.
But what do we use all the time?
(12:17):
Text messages.
Text messages is the primarymean of communication,
especially for folks youngerthan I am, which is most
everybody.
But that, that, that sort ofinterface we're really
comfortable with and we use itall the time.
And one of the other things thathelped us was the fact that with
the rise of search engines overthe past 20 years is the fact
(12:40):
people are used to typing inquestions.
And so I don't have to foolabout the timber of my voice.
I can take as long as I want.
It's basically texting to areally smart.
And for the moment, we're gonnakeep the hallucinations and the
accuracy issues off to the sidefor right now.
But in general, we have thiswhole means of access using an
(13:02):
interface that we're so familiarwith, which is you're basically
a giant text message system whenyou're using one of these LLMs.
Andreas Welsch (13:10):
So that's your
answer?
T. Scott Clendaniel (13:12):
That's my
answer.
Andreas Welsch (13:13):
Oh, okay.
Good.
T. Scott Clendaniel (13:14):
Even
remember the question'cause that
happens to me.
Andreas Welsch (13:17):
Yeah.
All, all good.
So I see an Anil here is askingin the chat if you have any tips
how to deal or handle copyrightissues when using UX generated
from GenAI tools.
T. Scott Clendaniel (13:30):
Yes, I do.
Think of all of the GenerativeAI tools in one of two realms.
One would be brainstorming.
Help to come up with somedifferent ideas, developing
prototypes, that type A.
Number two is being able to comeup with different ideas that you
(13:51):
couldn't have thought of on yourown.
So you're brainstorming.
The system's brainstorming backto you.
That's like a good first draftin terms of the copyright
issues.
Do not use anything that comesout of Generative AI.
It's already been pretty wellestablished that you can't
copyright stuff that comes outof Generative AI.
And I don't know if anyone elsenoticed, but in the past month
(14:11):
the patent office has now saidthat anything, any patent that's
written or comes out of AI willnot be recognized.
And that patent is, so this is.
Your first step field, this isyour brainstorming field.
This is the area where I'm gonnacome up with new ideas and test
some things out.
So in terms of copyright, avoidit like the plague.
Think of it as coming up withthe different ideas.
(14:34):
If you're gonna have an image,take the image that you created
as a good example, and then sendit to whoever your internal
artist is, or do it on your own.
Wonderful.
Thank you.
And yeah, to be really careful.
Yeah.
Andreas Welsch (14:46):
Yeah.
When it comes to UX, one of thetools that I was really
surprised by was one that TobiasZwingmann mentioned a couple
months ago when he was on theshow and there was Vercell AI.
And you can give it a prompt andyou say, Hey, create a layout
for our website to do leadcapture, for example, or as a
landing page or a sign up for anewsletter or something else, or
(15:07):
even more complex things.
And using Generative AI, it doescome up with the design and it
shows that to you and you canmake modifications and you can
transfer that over to a actualprogramming language.
So that's where I got excitedbecause it's not just text
generation, not just imagegeneration or synthetic voice or
video, but there are otherreally powerful and, useful
(15:29):
applications where you can use
T. Scott Clendaniel (15:32):
Let me tell
you an opportunity that's missed
a lot by organizations that Ithink that Generative AI can
also help with.
For decades, when I worked inmarketing, I was desperate to
understand what exactlycustomers were thinking.
I either had to sit down withthe customer service department,
listen to phone calls, come inor look at messages from email
(15:53):
requests or whatever else.
One of the things you wanna lookinto is the ability to actually
capture what the first promptsthat someone used for an
internal.
Tool were for several reasons.
One, it helps you improve the UXof the product overall, even
when it's not a genre of AIinterface.
But number two, what was theirstarting point?
(16:15):
We, in business, have a nastytendency to talk to ourselves.
We come in and we can't erasewhat we think we know, so we
just assume that everybody elseknows that we assume that they
want this new benefit or thisnew feature or whatever else.
You have a non-stop collectionof user-generated research
(16:36):
information.
Take some time to read some ofit for heaven's sake.
I'm not saying read all of it,but you can also ask your
Generative AI tool to reviewwhat different questions or
comments that have come through.
What are the most commonaspects?
Because you wanna learn aboutyour customer, you want to hit
the customer needs, and a greatway to understand why people
(16:57):
don't like your particular UXapproach.
You would really want to askthem the question if you've only
got buttons on a website, youcan't do that.
But if you can capture whatpeople are entering into your
customer service team throughthe actual text, you will have
amazing insights that you didn'thave before.
Andreas Welsch (17:16):
So in a way,
even use Generative AI to
summarize the, information, theraw data that, that you're
getting and, draw some,conclusions Absolutely from it.
T. Scott Clendaniel (17:25):
And that's
where RAG comes into play, and I
think that's gonna become moreand more important.
There was some research that wasreleased by Wharton University
of Pennsylvania, I guess twoweeks ago, three weeks ago, or
that's where I found it.
I don't know that they generatedthe research, but that's where I
found the articles from aWharton professor was the fact
that if you have a solelyisolated, resource to create
(17:48):
your own genre of AI interfacedoes not tend to do well at all.
Because it loses all thelanguage clues.
So we've got the specializedinformation.
So I fixed that problem, but nowI've still got the interface
problem.
So that's why I think RAG'sgonna help because it's going to
allow the best of both worlds.
Andreas Welsch (18:07):
Yeah.
And make it tangible andpreserve that context.
Even now, now I'm, wondering if,you're an AI leader, if you want
to become an AI leader in abusiness, what do you need to be
aware of then when you want tointroduce large language models?
Be it for improving the userexperience or be it for other
reasons.
T. Scott Clendaniel (18:28):
Yeah.
One of the things that I'm alittle worried about is the fact
that we now wanna calleverything artificial
intelligence, like everything inthe world.
And this is a funny but truestory because I was like, gosh,
and this was five years ago, AIon this, AI on that, oh, our
product has AI, AI everywhere,right?
(18:48):
And I was like, wouldn't it befunny if someone said that they
used artificial intelligence ina toaster?
I will be darned if in 2017 Ididn't find an old ad.
For a toaster with artificialintelligence.
Alright, stop using that phrasefor everything.
Just because it's an interfacewith a computer does not make it
(19:10):
artificial intelligence.
So I think as a leader, thereare a lot of folks out there,
bless their hearts executiveswho come back from the golf
course and our lady CEO EO isplaying with other lady CEOs and
our guy CEO is playing with hisguy CEOs, and they're all
playing golf and they come backand say, okay, we gotta do AI.
(19:33):
Why?
Because everybody else on thegolf course is doing AI.
And I can't go out on the golfcourse and say, we're not doing
AI.
I don't wanna be embarrassed.
What?
That's terrible.
So when you're talking aboutthis field, don't do AI because
you want to be able to say youare doing ai.
You need to find an appropriateproblem where this is gonna be a
(19:54):
good fit.
So throwing around the terms allover the place is not going to
help you.
Also, everyone is now using thatphrase.
So whatever type of marketadvantage you thought you had 18
months ago by using the phrase,artificial intelligence is gone
now because everyone else issaying it's, you're not
differentiating yourself.
(20:15):
So that's the first sort ofproblem area.
The second problem area is plainold statistics and machine
learning.
Our old friend logisticregression.
Decision, trees, random parts,anybody remember those?
Algorithms that we use forproblem solving.
You're lot more likely to get ahigher ROI from solving problems
(20:35):
like fraud detection responseanalysis, email, open rates, all
those simple, binary outcomeswith a simple algorithm than you
are to go full out and buyingall these platforms for
artificial intelligence.
Crawl, walk, run.
You don't have to be full speedat the moment to be able to make
(20:57):
this stuff work.
Take it easy.
Andreas Welsch (21:00):
That really
brings it back to the basics,
right?
And also a bit of thatmisconception that Generative AI
now renders everything elseobsolete in all of the logistic
regression and statisticalmethods and so on and it does
not, right?
ChatGPT does not generate yourdemand forecast.
It probably does, but not veryaccurately.
(21:22):
There is that.
Now one other question here inthe chat I think is really
interesting.
That's how soon will we seecommerce where users can
purchase goods or serviceswithin the chat interface versus
using a specific app?
What do you think is it, amulti-year journey?
Is it just connecting anotherAPI, like a plugin like, like
(21:44):
Kayak or something else that youcan already connect in ChatGPT,
what do you think?
T. Scott Clendaniel (21:50):
I think you
could probably jury-rig it
today.
I wouldn't necessarily recommendit, but yeah with enough blood,
sweat and tears.
You can actually use a genre ofAI interface today as to be able
to purchase goods.
I don't recommend it becauseit's still very buggy, but yeah,
(22:11):
I absolutely think that in 2024there're gonna be a bunch of
products that offer that.
Remember that the purpose ofAmazon creating the product,
which shall not be mentioned.
Hint, hint, it rhymes withMarexa.
The purpose was not for peopleto be able to play music.
The purpose was they thoughtthat this amazing device is
(22:32):
gonna be in every room in thehouse and people are gonna be
making purchases with it all thetime.
At one point, they had 10,000folks working on hardware,
software, all that.
That's a lot of folks, and theycompletely missed the boat, in
my humble opinion, on theinterface of the power of things
like.
Chat GPT, but they thought thate-commerce was gonna be solved
(22:52):
by this.
I'm not sure that peopleparticularly want to go into
that type interface or trustedenough.
So I don't think the issue isthat we can't do the technology
when it's gonna happen.
It's when our customers reallygonna prefer that.
Andreas Welsch (23:05):
What I think is
really interesting about that
point and, obviously I'm usingsome of those devices myself and
have been using them for awhile, is it at least personally
I want to see the item.
I want to see what does it looklike?
Is it large?
Is it small?
Is it white?
Is it black?
Is it a shade of white or ashade of blue so in, absence of
(23:26):
the visual cues and information,just relying on language and
then also trusting that thedevice accurately understands
what I mean or gives me theoptions.
And, then if it's reciting theoptions what was option number
one again, out of those three orout of those five?
And are these even the threethat I want to see?
T. Scott Clendaniel (23:45):
If you went
into a store today with a live
salesperson, 70% of allcommunication is nonverbal.
Yeah, it's tone of voice, it'sbody language.
You lose all that when you havea text interface.
You're not gonna have any ofthose components.
And we still have a hard timegetting sales professionals to
understand what we want, andthis is what they specialize in
(24:07):
life to be able to do.
So if we're expecting to get.
That type of clarioncommunications, let alone your
excellent point about the fact,but this doesn't help me see it.
This doesn't help me hold it canI hold it up to my couch and
make sure it matches orwhatever.
Andreas Welsch (24:24):
Yeah.
I love the VR by the way.
See what this looks like in yourroom.
I.
Yes.
For, some things.
That's really helpful.
Yes.
For a bunch of batteries or yourdish soap.
Not so much, but definitely.
All hey we're, getting close tothe end of the show and I was
wondering if you could summarizethe key three takeaways for our
audience today when it comes toGenerative AI and improving user
(24:46):
experience with it.
T. Scott Clendaniel (24:48):
Sure.
Number one, encourage employeesto use it as a tested ground for
brainstorming, not for finalresults.
Two, trying to ban it is aterrible idea'cause they're
gonna use it anyway.
And you would rather have sometype of level of control on
understanding what they'redoing.
Three, I would say use it as away of researching what people
(25:10):
are really interested in andunderstanding customer needs.
Just don't think of it as abandaid.
Think of it as a fantastic wayto improve the overall customer
experience.
Those would be my top three.
Andreas Welsch (25:21):
Wonderful.
Thank you so much.
Scott, it was a pleasure.
Thank you.
I love being here.
Thank you for sharing yourexpertise with us and for those
of you in the audience forlearning with us.
I think it was a very insightfulsession.
Really appreciate you giving usthat, that broad overview and
bring it back to, business andwhat really matters.
T. Scott Clendaniel (25:36):
You bet.
I hope I can come back one day.
Take care.
Andreas Welsch (25:38):
Yeah.
Awesome.
Thank you so much.