February 4, 2025 • 26 mins
In this solo episode of the podcast, I address some recent questions I've gotten specifically about A.I. in CS. A few tangents are included as per usual:
Chapters:
00:00 - Intro
02:42 - When is your program ready for A.I.?
04:10 - Data readiness for installing A.I.
08:14 - Using AI for content generation
11:05 - Staying current or getting up to speed on A.I.
13:25 - Ticket deflection with A.I.
16:00 - Utilizing A.I. in establishing integrations and configurations
17:03 - A.I. Chatbots
18:03 - Google’s NotebookLM use cases
20:35 - What to watch out for in adopting A.I.
23:10 - Start with the Simple Things!
Enjoy! I know I sure did...
Special shoutouts in this episode go out to Ariglad, Clueso, HeyGen, QueryPal and Vitally!
+++++++++++++++++
Like/Subscribe/Review:
If you are getting value from the show, please follow/subscribe so that you don't miss an episode and consider leaving us a review.
Website:
For more information about the show or to get in touch, visit DigitalCustomerSuccess.com.
Buy Alex a Cup of Coffee:
This show runs exclusively on caffeine - and lots of it. If you like what we're, consider supporting our habit by buying us a cup of coffee: https://bmc.link/dcsp
Thank you for all of your support!
The Digital Customer Success Podcast is hosted by Alex Turkovic
🎬 This content was edited by Lifetime Value Media.
Learn more at: https://www.lifetimevaluemedia.com
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:03):
Today we got another Q&A episode for you, heavily
focused on AI.
Stay tuned Once again.
Welcome to the Digital CustomerExperience Podcast with me,
alex Turkovich, so glad youcould join us here today and
every week as we explore howdigital can help enhance the
customer and employee experience.
My goal is to share what myguests and I have learned over
(00:25):
the years so that you can getthe insights that you need to
evolve your own digital programs.
If you'd like more info, needto get in touch or sign up for
the weekly companion newsletterthat has additional articles and
resources in it.
Go to digitalcustomersuccesscom.
For now, let's get started.
Greetings and welcome back tothe Digital CX Podcast, the show
(00:46):
where we talk about all thingsdigital and CX.
My name is Alex Tergovich.
You're listening to episode 90of the show and another solo
episode for you.
We've had an interesting coupleof weeks here just because we
had the holidays and I did acouple of additional solo shows
over the holidays where wetalked about like Notebook LM
(01:06):
and having that create podcastsfor you, which is super cool,
very fun, and I'd love to hearfrom you if you've played around
with it at all and if you'veused it.
You know one of the things thatI realized is that we didn't
really talk about you know,specific use cases of that
during that episode, so we maydig into that a little bit today
(01:30):
.
But first and foremost, I dowant to extend a really warm
welcome to our two show sponsorsfor the next little while here.
First off, I'd like to welcomeback Vitaly, appreciate the
sponsorship there and keepingthis content going.
Vitaly has been a sponsorbefore they're back for another
round, and so you'll be hearingfrom them throughout these
(01:54):
episodes.
Big difference between lasttime and this time is that I've
actually become a heavy Vitalyuser these days, and so I like
to talk about stuff that I canstand behind vitally user these
days, and so I'd like to talkabout stuff that I can stand
behind.
Also want to extend a bigwelcome to QueryPal as a second
show sponsor, who do amazingthings in terms of support,
(02:15):
ticket deflection and artificialintelligence.
We're actually testing themright now as well.
Welcome to our new showsponsors.
So today I have a couple ofthings that came to my inbox
after last week's episode that Ifelt were worth kind of
bringing up in this episode.
A couple of questions that camein that are AI related got a
couple of non-AI related thingsas well, but one of the
(02:38):
fundamental things that did comethrough was this question of
when is your program ready forartificial intelligence?
We talk a lot about thetechnology, we talk a lot about
the benefits that it brings toyou.
We talk a lot about the usecases and all that kind of fun
stuff, but what we don't reallyspend a lot of time on, and what
(03:04):
vendors also don't spend a tonof time on, is what you need in
place in order to be ready forartificial intelligence.
There are several use casesthat I have in mind as I'm
talking about this.
One would be implementing achatbot, a customer-facing
chatbot, for instance.
(03:24):
Implementing a chatbot, acustomer-facing chatbot, for
instance, you know.
Another might be implementing aco-pilot for your internal
teams, which are two kind ofinterrelated things.
Another one might be a chatbotto help you generate content,
and so the fundamental kind ofanswer here is it depends a
little bit on the type ofartificial intelligence tool
(03:44):
that you're looking for andshopping for, but fundamentally
speaking, one of the things thatyou are going to likely need in
order to be ready is the properdata.
I know you've heard us talkabout data over and over and
over again.
We talk about it a lot becauseit's super important.
(04:06):
In this case, though, I want toraise two specific things in
terms of your data, and let'suse the use case of a chatbot.
The first is you know you don'treally need super, super clean
customer data, super, superclean customer data per se.
(04:30):
If you're implementing thatbase level of chatbot that
answers some product questions,right, you can literally throw
it out in a link or however thatplatform works, to your
customer so that they can answersome basic questions, and it
won't necessarily need to betrained on the customer-specific
data for that foundationallevel.
However, if you want to getmore sophisticated with the
(04:54):
artificial intelligence that iscustomer-facing, for instance,
you might want it to use somesemblance of personalization.
Use some semblance ofpersonalization.
You may want it to respond in away that's appropriate for the
kind of user the person is.
Are they an admin, are they anexecutive?
Are they an end user?
Whatever that may be, you maywant to personalize the
(05:19):
responses based on their role.
You may want to include theirfirst name in responses.
You may want the chatbot toknow their email address so it
can email things out.
Those kinds ofcustomer-specific
personalization elements needcustomer data.
But I think more fundamentally,if you're going to install a
(05:40):
chatbot, you got to have thecontent on the backend to train
the artificial intelligence toanswer the questions that are
coming in appropriately.
If you're expecting to installa chatbot and have it be able to
answer your product questionswithout doing much training,
(06:02):
then I think you've been ledastray by someone, because
ultimately, the more data andthe more quality data that you
can feed the chatbot, the moreaccurate and the more thorough
your responses are going to bewhen somebody goes to ask it a
(06:23):
question.
Now, what does that look like?
That can be any number ofthings and these are all kind of
platform specific.
Some do it, you know they alldo it a little bit differently.
Most chatbots will allow you totrain it using your knowledge
base, right?
(06:43):
So if you have written a bunchof articles, that is prime
territory for you know havingtraining a chatbot.
Another is your support tickets.
You can elect to have a chatbottrained on your support tickets
and what responses that you'veprovided to customers previously
that may not be reflected inyour knowledge base.
(07:06):
Some chatbots allow you totrain it on Slack channels and I
think the important call outhere is that you do have to be a
little bit careful about, um,what information you train your
chatbot and what you don't.
And you know, for instance, youdon't you don't want to release
(07:26):
the hounds, so to speak, on yourentire slack instance, because
then you'll effectively betraining that chatbot to respond
to things that you may not wantyour customers to know about.
So you do have to be prettyguarded in in what you're
training that chatbot.
So those are really the twofundamental things Like,
(07:47):
fundamentally speaking, whatcontent are you feeding it?
What content are you trainingit on?
The more the merrier, but notat the sacrifice of quality.
And so when people ask me aboutdata readiness or readiness to
start using AI especially ifit's content generation on a
(08:09):
product and those kinds ofthings my advice is always start
writing.
Start getting your team writingarticles.
If you're sitting on aknowledge base with 50 articles
in it, it's probably not goingto cut the mustard.
In fact, 100 may not be enough,200 may not be enough.
It just depends a little bit onyour product complexity as well
(08:29):
and how many different topicsthat are potentially you know
going to be asked by yourcustomers.
Now there are some pretty cooltools that can help you with
that sort of content generation.
A couple of things, a couple ofones that kind of stand out.
(08:51):
There are a few solutions outthere which help you
specifically with knowledge base.
One in particular is calledAriglad A-R-I-G-L-A-D.
We've been testing them out alittle bit as well.
So, basically, you train it onyour existing content, you train
(09:12):
it on your support tickets, youtrain it on specific Slack
channels, and what it will do isit will suggest updates to
existing articles as well assuggest new articles and write
them for you.
Very cool tech and does areally good job of uncovering
additional articles you may wantto write some things about.
(09:35):
And there are other solutionsthat do that.
In fact, some chatbot vendorsdo that intrinsically so, like
Maven.
Agi, for instance, has thiskind of feature where it will
suggest articles and suggestupdates to articles based on.
You know some of theconversations that have been
happening with customers usingthe chatbot.
(09:55):
So there's some cool things outthere that can help you
essentially maintain yourknowledge base.
Some others that are worthmentioning something like Clueso
is really cool.
We've been checking out Cluesorecently as well.
Clueso I could say Clueso onemore time.
Clueso is a platform thatallows you to either record
(10:16):
video using the platform orupload video of like product
demos and screen shares andthose kinds of things, and it'll
turn it into basically AIregenerated, rewritten video
snippets with professionalsounding voiceovers, as well as
articles.
So by uploading one kind ofkarate demo, you can spit out
(10:38):
videos and articles withouthaving to write them all from
scratch.
You can even do things like youknow create GIFs out of the
images and things like that thatit creates.
So there's all kinds of cooltools.
I'm just scratching the surface.
There is a tech stack page onthe website that you can go
check out if you want to go seewhat other platforms I've
(11:00):
recommended in the past or youknow are out there to help you
with some of that contentcreation.
Another thing that came out oflast week's episode is Mike and
I had a somewhat lengthy backand forth about being behind on
AI and if you are behind on AI,what you should kind of do about
(11:21):
it.
We didn't really get into thespecifics there, so I wanted to
provide a couple of specificsoff the back of that
conversation, because I did geta couple of questions about okay
, well, you know, what should Ido to really increase my, I
guess language in artificialintelligence.
The first thing I would justrecommend blanket you know
(11:43):
blanket statement is start usingchat, chat GBT, start using
perplexity and I've talked aboutthis, the difference between
the two on the, on the channelor on the podcast before.
But essentially chat GBT it'schanging a little bit just in
terms of its access to theinternet, but ChatGPT tends to
(12:04):
be kind of like a companion,more so than Perplexity, which
is basically a research engine.
It's kind of like you know, alot of times I'll use Perplexity
over Google just because itdoes, you know, really good
internet research and summarizesthings for you really nicely,
kind of like Google searchresults do today, but just
better.
Kind of like Google searchresults do today, but just
(12:28):
better.
So my first piece of advicewould be just start using that
stuff on a regular basis.
Anything you would Googlenormally, just pulling up Google
real quick, use ChatGPT orPerplexity to do that and start
getting in the habit of using itand seeing what it comes back
with, because one of the otherbenefits there is that you can
follow up with it andessentially have a conversation
(12:50):
with artificial intelligencethat way, so that it learns a
little bit about how you want tolearn and how you think, and
then you know you in turn getthe details that you're looking
for and get the accurate answersthat you're looking for,
because we all know a Googlesearch typically doesn't return
what you want it to the firsttime around.
(13:11):
These next few are a couple ofsuggestions just for leadership
in general and some of thetooling that you might want to
look at implementing if you arelooking to essentially get your
organization into AI when it'sany number of like ticket
deflection, ticket co-pilotsystems the sponsor of the show,
for instance, querypal, is agood example of that but
(13:34):
essentially systems that willsuggest answers to questions
that come in in real time sothat when your agents or when
your CSMs go to respond, they'llhave a baseline or maybe even a
draft of an answer ready to gothat they can just edit and fire
off.
That can save a ton of time andreally increase the amount of
(14:00):
throughput that your teams canhandle, just because some of
that grunt work is being donefor them and research is being
done for them.
So, basically, like a co-pilotsituation.
Another is just in the realm ofcontent creation.
There are so many cool toolslike HeyGen, like Cluso that are
(14:23):
really focused on contentcreation and especially video
content creation.
That has come a long way in avery, very short amount of time
and there really isn't an excuseanymore for your knowledge base
articles to not have videos inthem and for you not to have
really thorough productwalkthroughs and demos and
(14:44):
things like that that you canthen go and install either in
product or have, you know, sendout via guides or there's any
number of things that you can doonce you have that content
created, but a lot of times, asyou know, content creation tends
to be, you know, the root of ofa lot of problems, just in
terms of getting off the groundwith self-serve motions.
(15:07):
So look for tools, look forartificial intelligence tools
that can help you create thatcontent.
And I'm not saying that thecontent should be solely created
by artificial intelligence,because while it's good, it's
not there yet.
I can tell you right now Iwould not want to sit down and
(15:28):
listen to 30 minutes ofAI-generated video, because the
voiceover still you can tell,you can just tell it is
AI-generated.
Do I mind listening to that fora minute or two or three, for a
quick walkthrough of something?
Absolutely not.
I think that is a prime usecase for AI video generation,
(15:52):
those really short walkthroughsthat you can just pump out one
to the next Great use case forAI.
Another place to look atimplementing artificial
intelligence is via yourintegrations.
I will use Zapier as an examplefor right now.
Makecom also has somethingsimilar.
(16:15):
A lot of tools are nowlaunching artificial
intelligence configuration help.
In other words, you prompt whatyou want to accomplish and the
tool will then spit out aconfiguration suggestion based
on what it is you wanted to do.
Useful, especially when you'rebuilding very complex
(16:43):
integrations and when you'rewanting to do a lot in terms of
connecting different tools anddata sets and those kinds of
things.
It can really take the gruntwork out of building
integrations between systems.
So definitely look into youknow how AI can help you with
those kinds of things.
And then you know.
Another thing that we mentionedin the last question was just
around chatbots in general.
(17:04):
There are tons of vendors outthere.
Now A couple that come to mindMaven, agi, intercom is
interesting because Intercom hasa standalone AI chatbot that
you whether you use intercom forsupport tickets or Zendesk for
support tickets or whatever youuse for support tickets um,
(17:27):
intercom and their AI chatbotcan actually um, you know it can
.
It can plug into any any systemthere.
So that is definitely somethingthat is relatively easy to
stand up, but again, you need tohave the content to train the
chatbot.
It's kind of a vicious cycle and, as I mentioned at the
(17:50):
beginning of the episode,another thing I wanted to make
sure that we touched on todaywas in response to the two
holiday episodes that Ipublished, all about using
Google's Notebook LM to promptand then to have it pump out
podcast episodes for you.
Now, there were a couple ofresponses that I got that were
(18:15):
like well, why would you wannado that?
And it's a fair question,because it seems, at first
glance it seems a little bitrandom that you would want to
produce a podcast episode fromtwo personas that seemingly had
nothing to do with you.
You'd never like.
These are just, you know, afemale and a male persona.
(18:36):
They're chatting about whatever.
Whatever it is you want them tochat about via your prompt
persona.
They're chatting about whateverit is you want them to chat
about via your prompt.
But a couple of things did cometo mind as I was going through
that exercise.
The first is if you are drivinga specific type of let's call
it thought leadership.
(18:57):
One example would be if you arein vertical SaaS and you serve
a very specific industry, it isquite likely that you have a
blog out there with articlesrelated to that industry.
But maybe the industry isunderserved in podcast land and
so you might want to produce apodcast that is specific to that
(19:19):
industry.
It might get a little bit oldif you're producing this podcast
using solely Notebook LM, butif you want to do you know you
(19:39):
were writing that could be areally good use case of using
Notebook LM to produce thatprofessional sounding podcast
for your industry.
And very similarly, if you'renot serving a particular
industry you're serving multipleindustries, but you a software
platform that is focused on Idon't know e-commerce or
(20:02):
accounting or whatever acrossyou know multiple industries.
That could be another instancewhere you do want to produce
some thought leadership aboutthe services and the space that
you are in.
So considering using NotebookLM for that sort of thing and
you know you might even toy withhaving it answer questions,
(20:23):
answer customer questions viaaudio that could be an
interesting use case, though youknow you may want to vet that
it is actually answering thequestion correctly.
I think, ultimately, when itcomes to all of these artificial
intelligence tools, the thingthat you're going to want to
watch out for is just kind ofover-rotating on it, because you
(20:46):
know, as I mentioned in thatepisode with Mike, you know
there are people out there, Iknow, that are listening, where
you feel kind of behind thetimes and you kind of feel
guilty about that a little bit.
And I'm here to tell you thatit doesn't really take a whole
lot to get yourself fullycomfortable with what artificial
(21:07):
intelligence is today versuswhat it can also be, because it
is really only limited by ourimagination, honestly, and you
know we'll see where all of thatgoes.
But fundamentally, it can beequally, I guess, intimidating
as well and it can be confusingas well.
(21:27):
So I would, I would, I wouldstrongly encourage you to be
super crisp on what your usecases are going into it.
If you know that you want toleverage artificial intelligence
for helping your support agents, or for a phone tree or
something like that, or for achatbot or a content creation or
(21:51):
process automation, whateverthe use case is, my
recommendation is to go in witha really strong use case and
then work backwards from there.
A lot of people make the mistakeof purchasing the tool or
purchasing the solution, orgetting into a tool only to find
out that it kind of doesn'treally do what you imagined it
(22:14):
might do.
But if it's you just likeliterally wanting to explore
what the possibilities ofartificial intelligence are, my
recommendation is just to dig in.
Do some chat, gpt, do someperplexity.
Replace your Googling withthose kinds of tools.
There are also some reallygreat courses out there.
Coursive is one, but it's kindof like clickbaity.
(22:36):
I don't know.
I've been through some of itC-O-U-R-S-I-V.
You've probably seen it in yourTikTok feed or something like
that.
If you're on TikTok, linkedinLearning, there's a lot of good
stuff on there.
Coursera has some great stuffaround artificial intelligence.
Just educate yourself on thisstuff and I guarantee you if you
(22:57):
spend a concerted week or twoon this stuff, and I guarantee
you if you spend a concertedweek or two on this stuff,
you'll be in good shape.
So don't let your lack ofcurrent knowledge detract you
from adopting it tomorrow To dosimple things, because there are
very simple and powerful thingsthat you can do.
A couple of things that I useChatGPT for constantly is if I
(23:20):
have a long document and I don'thave time to read through it.
I'm going to upload it toChatGPT and I'm going to ask it
to summarize it for me.
Images, things like that.
I have it help me with my son'shomework.
My son is in algebra and it'sbeen 30 years since I've done 20
, 25 years I'm not that old 25years or so since I've done 25
(23:41):
years.
I'm not that old 25 years or sosince I've done algebra.
And so literally I'll pick upChatGPT, I'll take a picture of
the problem that he's working onand it'll solve it for me and
explain how it's done, so thatthen I can explain how it's done
.
Like you know, the use casesare just insane.
(24:02):
The other favorite, favorite,favorite thing that I love to do
and I'll leave you with thisone is I've been walking a ton.
I've been trying to do like 10,you know the 10,000 steps a day
, three miles kind of situation,some days more successfully
than others.
But sometimes, when I'm workingon a specific problem or
(24:25):
thinking through somethingspecific and I'm just a little
bit stuck, what I will do ispull up ChatGPT and go into
conversation mode, where you canliterally talk to it and just
have a conversation aboutwhatever it is I'm talking about
and it will help me organize mythoughts and it'll have a
conversation with me about thatthing until I have a clear
(24:48):
mental image or you know apicture of what that is and what
my next steps should be.
I've used it so many times tounblock myself about certain
things where either I justdidn't have that little spark of
creativity or that little bitof knowledge, and that's what
one of the things that ChatGPTis really really good at.
(25:10):
It's great at unlocking thatkind of thought process.
So I guess that's your top tipfor the episode go talk to chat
gpt literally.
Uh, you won't regret it, uh,though it gets a little creepy
sometimes.
Anyway, um, this episode hasbeen a little bit all over the
(25:31):
place, but I hope you've taken,uh, something away and I've hope
I I hope I've answered some ofthe questions that I've gotten
in from you from recent episodes.
Please keep the questions andthe engagement coming.
I love, love, love to hear fromyou, because a lot of times I
put these episodes out into thewild and it just kind of it's
out there, you don't know who'slistening, who's not listening,
(25:53):
and all that kind of fun stuff.
So please keep it coming.
We have some awesomeconversations coming up in the
next few weeks with some greatpeople that are directly
involved in digital Some not so,but tangentially, so I'll keep
you guessing on that.
But, you know, look forward tohaving you with this next series
(26:15):
of four episodes and then I'llsee you for episode 95, another
solo episode in a few weeks'time.
Have a great week ahead andthanks so much for listening.
Thank you for joining me forthis episode of the Digital CX
Podcast.
If you like what we're doing,consider leaving us a review on
your podcast platform of choice.
If you're watching on YouTube,leave a comment down below.
(26:37):
It really helps us to grow andprovide value to a broader
audience and get moreinformation about the show and
some of the other things thatwe're doing at
digitalcustomersuccesscom.
I'm Alex Trigovich.
Thanks so much for listening.
We'll talk to you next week.
Today we got another Q&A episode for you, heavily
focused on AI.
Stay tuned Once again.
Welcome to the Digital CustomerExperience Podcast with me,
alex Turkovich, so glad youcould join us here today and
every week as we explore howdigital can help enhance the
customer and employee experience.
My goal is to share what myguests and I have learned over
(00:25):
the years so that you can getthe insights that you need to
evolve your own digital programs.
If you'd like more info, needto get in touch or sign up for
the weekly companion newsletterthat has additional articles and
resources in it.
Go to digitalcustomersuccesscom.
For now, let's get started.
Greetings and welcome back tothe Digital CX Podcast, the show
(00:46):
where we talk about all thingsdigital and CX.
My name is Alex Tergovich.
You're listening to episode 90of the show and another solo
episode for you.
We've had an interesting coupleof weeks here just because we
had the holidays and I did acouple of additional solo shows
over the holidays where wetalked about like Notebook LM
(01:06):
and having that create podcastsfor you, which is super cool,
very fun, and I'd love to hearfrom you if you've played around
with it at all and if you'veused it.
You know one of the things thatI realized is that we didn't
really talk about you know,specific use cases of that
during that episode, so we maydig into that a little bit today
(01:30):
.
But first and foremost, I dowant to extend a really warm
welcome to our two show sponsorsfor the next little while here.
First off, I'd like to welcomeback Vitaly, appreciate the
sponsorship there and keepingthis content going.
Vitaly has been a sponsorbefore they're back for another
round, and so you'll be hearingfrom them throughout these
(01:54):
episodes.
Big difference between lasttime and this time is that I've
actually become a heavy Vitalyuser these days, and so I like
to talk about stuff that I canstand behind vitally user these
days, and so I'd like to talkabout stuff that I can stand
behind.
Also want to extend a bigwelcome to QueryPal as a second
show sponsor, who do amazingthings in terms of support,
(02:15):
ticket deflection and artificialintelligence.
We're actually testing themright now as well.
Welcome to our new showsponsors.
So today I have a couple ofthings that came to my inbox
after last week's episode that Ifelt were worth kind of
bringing up in this episode.
A couple of questions that camein that are AI related got a
couple of non-AI related thingsas well, but one of the
(02:38):
fundamental things that did comethrough was this question of
when is your program ready forartificial intelligence?
We talk a lot about thetechnology, we talk a lot about
the benefits that it brings toyou.
We talk a lot about the usecases and all that kind of fun
stuff, but what we don't reallyspend a lot of time on, and what
(03:04):
vendors also don't spend a tonof time on, is what you need in
place in order to be ready forartificial intelligence.
There are several use casesthat I have in mind as I'm
talking about this.
One would be implementing achatbot, a customer-facing
chatbot, for instance.
(03:24):
Implementing a chatbot, acustomer-facing chatbot, for
instance, you know.
Another might be implementing aco-pilot for your internal
teams, which are two kind ofinterrelated things.
Another one might be a chatbotto help you generate content,
and so the fundamental kind ofanswer here is it depends a
little bit on the type ofartificial intelligence tool
(03:44):
that you're looking for andshopping for, but fundamentally
speaking, one of the things thatyou are going to likely need in
order to be ready is the properdata.
I know you've heard us talkabout data over and over and
over again.
We talk about it a lot becauseit's super important.
(04:06):
In this case, though, I want toraise two specific things in
terms of your data, and let'suse the use case of a chatbot.
The first is you know you don'treally need super, super clean
customer data, super, superclean customer data per se.
(04:30):
If you're implementing thatbase level of chatbot that
answers some product questions,right, you can literally throw
it out in a link or however thatplatform works, to your
customer so that they can answersome basic questions, and it
won't necessarily need to betrained on the customer-specific
data for that foundationallevel.
However, if you want to getmore sophisticated with the
(04:54):
artificial intelligence that iscustomer-facing, for instance,
you might want it to use somesemblance of personalization.
Use some semblance ofpersonalization.
You may want it to respond in away that's appropriate for the
kind of user the person is.
Are they an admin, are they anexecutive?
Are they an end user?
Whatever that may be, you maywant to personalize the
(05:19):
responses based on their role.
You may want to include theirfirst name in responses.
You may want the chatbot toknow their email address so it
can email things out.
Those kinds ofcustomer-specific
personalization elements needcustomer data.
But I think more fundamentally,if you're going to install a
(05:40):
chatbot, you got to have thecontent on the backend to train
the artificial intelligence toanswer the questions that are
coming in appropriately.
If you're expecting to installa chatbot and have it be able to
answer your product questionswithout doing much training,
(06:02):
then I think you've been ledastray by someone, because
ultimately, the more data andthe more quality data that you
can feed the chatbot, the moreaccurate and the more thorough
your responses are going to bewhen somebody goes to ask it a
(06:23):
question.
Now, what does that look like?
That can be any number ofthings and these are all kind of
platform specific.
Some do it, you know they alldo it a little bit differently.
Most chatbots will allow you totrain it using your knowledge
base, right?
(06:43):
So if you have written a bunchof articles, that is prime
territory for you know havingtraining a chatbot.
Another is your support tickets.
You can elect to have a chatbottrained on your support tickets
and what responses that you'veprovided to customers previously
that may not be reflected inyour knowledge base.
(07:06):
Some chatbots allow you totrain it on Slack channels and I
think the important call outhere is that you do have to be a
little bit careful about, um,what information you train your
chatbot and what you don't.
And you know, for instance, youdon't you don't want to release
(07:26):
the hounds, so to speak, on yourentire slack instance, because
then you'll effectively betraining that chatbot to respond
to things that you may not wantyour customers to know about.
So you do have to be prettyguarded in in what you're
training that chatbot.
So those are really the twofundamental things Like,
(07:47):
fundamentally speaking, whatcontent are you feeding it?
What content are you trainingit on?
The more the merrier, but notat the sacrifice of quality.
And so when people ask me aboutdata readiness or readiness to
start using AI especially ifit's content generation on a
(08:09):
product and those kinds ofthings my advice is always start
writing.
Start getting your team writingarticles.
If you're sitting on aknowledge base with 50 articles
in it, it's probably not goingto cut the mustard.
In fact, 100 may not be enough,200 may not be enough.
It just depends a little bit onyour product complexity as well
(08:29):
and how many different topicsthat are potentially you know
going to be asked by yourcustomers.
Now there are some pretty cooltools that can help you with
that sort of content generation.
A couple of things, a couple ofones that kind of stand out.
(08:51):
There are a few solutions outthere which help you
specifically with knowledge base.
One in particular is calledAriglad A-R-I-G-L-A-D.
We've been testing them out alittle bit as well.
So, basically, you train it onyour existing content, you train
(09:12):
it on your support tickets, youtrain it on specific Slack
channels, and what it will do isit will suggest updates to
existing articles as well assuggest new articles and write
them for you.
Very cool tech and does areally good job of uncovering
additional articles you may wantto write some things about.
(09:35):
And there are other solutionsthat do that.
In fact, some chatbot vendorsdo that intrinsically so, like
Maven.
Agi, for instance, has thiskind of feature where it will
suggest articles and suggestupdates to articles based on.
You know some of theconversations that have been
happening with customers usingthe chatbot.
(09:55):
So there's some cool things outthere that can help you
essentially maintain yourknowledge base.
Some others that are worthmentioning something like Clueso
is really cool.
We've been checking out Cluesorecently as well.
Clueso I could say Clueso onemore time.
Clueso is a platform thatallows you to either record
(10:16):
video using the platform orupload video of like product
demos and screen shares andthose kinds of things, and it'll
turn it into basically AIregenerated, rewritten video
snippets with professionalsounding voiceovers, as well as
articles.
So by uploading one kind ofkarate demo, you can spit out
(10:38):
videos and articles withouthaving to write them all from
scratch.
You can even do things like youknow create GIFs out of the
images and things like that thatit creates.
So there's all kinds of cooltools.
I'm just scratching the surface.
There is a tech stack page onthe website that you can go
check out if you want to go seewhat other platforms I've
(11:00):
recommended in the past or youknow are out there to help you
with some of that contentcreation.
Another thing that came out oflast week's episode is Mike and
I had a somewhat lengthy backand forth about being behind on
AI and if you are behind on AI,what you should kind of do about
(11:21):
it.
We didn't really get into thespecifics there, so I wanted to
provide a couple of specificsoff the back of that
conversation, because I did geta couple of questions about okay
, well, you know, what should Ido to really increase my, I
guess language in artificialintelligence.
The first thing I would justrecommend blanket you know
(11:43):
blanket statement is start usingchat, chat GBT, start using
perplexity and I've talked aboutthis, the difference between
the two on the, on the channelor on the podcast before.
But essentially chat GBT it'schanging a little bit just in
terms of its access to theinternet, but ChatGPT tends to
(12:04):
be kind of like a companion,more so than Perplexity, which
is basically a research engine.
It's kind of like you know, alot of times I'll use Perplexity
over Google just because itdoes, you know, really good
internet research and summarizesthings for you really nicely,
kind of like Google searchresults do today, but just
better.
Kind of like Google searchresults do today, but just
(12:28):
better.
So my first piece of advicewould be just start using that
stuff on a regular basis.
Anything you would Googlenormally, just pulling up Google
real quick, use ChatGPT orPerplexity to do that and start
getting in the habit of using itand seeing what it comes back
with, because one of the otherbenefits there is that you can
follow up with it andessentially have a conversation
(12:50):
with artificial intelligencethat way, so that it learns a
little bit about how you want tolearn and how you think, and
then you know you in turn getthe details that you're looking
for and get the accurate answersthat you're looking for,
because we all know a Googlesearch typically doesn't return
what you want it to the firsttime around.
(13:11):
These next few are a couple ofsuggestions just for leadership
in general and some of thetooling that you might want to
look at implementing if you arelooking to essentially get your
organization into AI when it'sany number of like ticket
deflection, ticket co-pilotsystems the sponsor of the show,
for instance, querypal, is agood example of that but
(13:34):
essentially systems that willsuggest answers to questions
that come in in real time sothat when your agents or when
your CSMs go to respond, they'llhave a baseline or maybe even a
draft of an answer ready to gothat they can just edit and fire
off.
That can save a ton of time andreally increase the amount of
(14:00):
throughput that your teams canhandle, just because some of
that grunt work is being donefor them and research is being
done for them.
So, basically, like a co-pilotsituation.
Another is just in the realm ofcontent creation.
There are so many cool toolslike HeyGen, like Cluso that are
(14:23):
really focused on contentcreation and especially video
content creation.
That has come a long way in avery, very short amount of time
and there really isn't an excuseanymore for your knowledge base
articles to not have videos inthem and for you not to have
really thorough productwalkthroughs and demos and
(14:44):
things like that that you canthen go and install either in
product or have, you know, sendout via guides or there's any
number of things that you can doonce you have that content
created, but a lot of times, asyou know, content creation tends
to be, you know, the root of ofa lot of problems, just in
terms of getting off the groundwith self-serve motions.
(15:07):
So look for tools, look forartificial intelligence tools
that can help you create thatcontent.
And I'm not saying that thecontent should be solely created
by artificial intelligence,because while it's good, it's
not there yet.
I can tell you right now Iwould not want to sit down and
(15:28):
listen to 30 minutes ofAI-generated video, because the
voiceover still you can tell,you can just tell it is
AI-generated.
Do I mind listening to that fora minute or two or three, for a
quick walkthrough of something?
Absolutely not.
I think that is a prime usecase for AI video generation,
(15:52):
those really short walkthroughsthat you can just pump out one
to the next Great use case forAI.
Another place to look atimplementing artificial
intelligence is via yourintegrations.
I will use Zapier as an examplefor right now.
Makecom also has somethingsimilar.
(16:15):
A lot of tools are nowlaunching artificial
intelligence configuration help.
In other words, you prompt whatyou want to accomplish and the
tool will then spit out aconfiguration suggestion based
on what it is you wanted to do.
Useful, especially when you'rebuilding very complex
(16:43):
integrations and when you'rewanting to do a lot in terms of
connecting different tools anddata sets and those kinds of
things.
It can really take the gruntwork out of building
integrations between systems.
So definitely look into youknow how AI can help you with
those kinds of things.
And then you know.
Another thing that we mentionedin the last question was just
around chatbots in general.
(17:04):
There are tons of vendors outthere.
Now A couple that come to mindMaven, agi, intercom is
interesting because Intercom hasa standalone AI chatbot that
you whether you use intercom forsupport tickets or Zendesk for
support tickets or whatever youuse for support tickets um,
(17:27):
intercom and their AI chatbotcan actually um, you know it can
.
It can plug into any any systemthere.
So that is definitely somethingthat is relatively easy to
stand up, but again, you need tohave the content to train the
chatbot.
It's kind of a vicious cycle and, as I mentioned at the
(17:50):
beginning of the episode,another thing I wanted to make
sure that we touched on todaywas in response to the two
holiday episodes that Ipublished, all about using
Google's Notebook LM to promptand then to have it pump out
podcast episodes for you.
Now, there were a couple ofresponses that I got that were
(18:15):
like well, why would you wannado that?
And it's a fair question,because it seems, at first
glance it seems a little bitrandom that you would want to
produce a podcast episode fromtwo personas that seemingly had
nothing to do with you.
You'd never like.
These are just, you know, afemale and a male persona.
(18:36):
They're chatting about whatever.
Whatever it is you want them tochat about via your prompt
persona.
They're chatting about whateverit is you want them to chat
about via your prompt.
But a couple of things did cometo mind as I was going through
that exercise.
The first is if you are drivinga specific type of let's call
it thought leadership.
(18:57):
One example would be if you arein vertical SaaS and you serve
a very specific industry, it isquite likely that you have a
blog out there with articlesrelated to that industry.
But maybe the industry isunderserved in podcast land and
so you might want to produce apodcast that is specific to that
(19:19):
industry.
It might get a little bit oldif you're producing this podcast
using solely Notebook LM, butif you want to do you know you
(19:39):
were writing that could be areally good use case of using
Notebook LM to produce thatprofessional sounding podcast
for your industry.
And very similarly, if you'renot serving a particular
industry you're serving multipleindustries, but you a software
platform that is focused on Idon't know e-commerce or
(20:02):
accounting or whatever acrossyou know multiple industries.
That could be another instancewhere you do want to produce
some thought leadership aboutthe services and the space that
you are in.
So considering using NotebookLM for that sort of thing and
you know you might even toy withhaving it answer questions,
(20:23):
answer customer questions viaaudio that could be an
interesting use case, though youknow you may want to vet that
it is actually answering thequestion correctly.
I think, ultimately, when itcomes to all of these artificial
intelligence tools, the thingthat you're going to want to
watch out for is just kind ofover-rotating on it, because you
(20:46):
know, as I mentioned in thatepisode with Mike, you know
there are people out there, Iknow, that are listening, where
you feel kind of behind thetimes and you kind of feel
guilty about that a little bit.
And I'm here to tell you thatit doesn't really take a whole
lot to get yourself fullycomfortable with what artificial
(21:07):
intelligence is today versuswhat it can also be, because it
is really only limited by ourimagination, honestly, and you
know we'll see where all of thatgoes.
But fundamentally, it can beequally, I guess, intimidating
as well and it can be confusingas well.
(21:27):
So I would, I would, I wouldstrongly encourage you to be
super crisp on what your usecases are going into it.
If you know that you want toleverage artificial intelligence
for helping your support agents, or for a phone tree or
something like that, or for achatbot or a content creation or
(21:51):
process automation, whateverthe use case is, my
recommendation is to go in witha really strong use case and
then work backwards from there.
A lot of people make the mistakeof purchasing the tool or
purchasing the solution, orgetting into a tool only to find
out that it kind of doesn'treally do what you imagined it
(22:14):
might do.
But if it's you just likeliterally wanting to explore
what the possibilities ofartificial intelligence are, my
recommendation is just to dig in.
Do some chat, gpt, do someperplexity.
Replace your Googling withthose kinds of tools.
There are also some reallygreat courses out there.
Coursive is one, but it's kindof like clickbaity.
(22:36):
I don't know.
I've been through some of itC-O-U-R-S-I-V.
You've probably seen it in yourTikTok feed or something like
that.
If you're on TikTok, linkedinLearning, there's a lot of good
stuff on there.
Coursera has some great stuffaround artificial intelligence.
Just educate yourself on thisstuff and I guarantee you if you
(22:57):
spend a concerted week or twoon this stuff, and I guarantee
you if you spend a concertedweek or two on this stuff,
you'll be in good shape.
So don't let your lack ofcurrent knowledge detract you
from adopting it tomorrow To dosimple things, because there are
very simple and powerful thingsthat you can do.
A couple of things that I useChatGPT for constantly is if I
(23:20):
have a long document and I don'thave time to read through it.
I'm going to upload it toChatGPT and I'm going to ask it
to summarize it for me.
Images, things like that.
I have it help me with my son'shomework.
My son is in algebra and it'sbeen 30 years since I've done 20
, 25 years I'm not that old 25years or so since I've done 25
(23:41):
years.
I'm not that old 25 years or sosince I've done algebra.
And so literally I'll pick upChatGPT, I'll take a picture of
the problem that he's working onand it'll solve it for me and
explain how it's done, so thatthen I can explain how it's done
.
Like you know, the use casesare just insane.
(24:02):
The other favorite, favorite,favorite thing that I love to do
and I'll leave you with thisone is I've been walking a ton.
I've been trying to do like 10,you know the 10,000 steps a day
, three miles kind of situation,some days more successfully
than others.
But sometimes, when I'm workingon a specific problem or
(24:25):
thinking through somethingspecific and I'm just a little
bit stuck, what I will do ispull up ChatGPT and go into
conversation mode, where you canliterally talk to it and just
have a conversation aboutwhatever it is I'm talking about
and it will help me organize mythoughts and it'll have a
conversation with me about thatthing until I have a clear
(24:48):
mental image or you know apicture of what that is and what
my next steps should be.
I've used it so many times tounblock myself about certain
things where either I justdidn't have that little spark of
creativity or that little bitof knowledge, and that's what
one of the things that ChatGPTis really really good at.
(25:10):
It's great at unlocking thatkind of thought process.
So I guess that's your top tipfor the episode go talk to chat
gpt literally.
Uh, you won't regret it, uh,though it gets a little creepy
sometimes.
Anyway, um, this episode hasbeen a little bit all over the
(25:31):
place, but I hope you've taken,uh, something away and I've hope
I I hope I've answered some ofthe questions that I've gotten
in from you from recent episodes.
Please keep the questions andthe engagement coming.
I love, love, love to hear fromyou, because a lot of times I
put these episodes out into thewild and it just kind of it's
out there, you don't know who'slistening, who's not listening,
(25:53):
and all that kind of fun stuff.
So please keep it coming.
We have some awesomeconversations coming up in the
next few weeks with some greatpeople that are directly
involved in digital Some not so,but tangentially, so I'll keep
you guessing on that.
But, you know, look forward tohaving you with this next series
(26:15):
of four episodes and then I'llsee you for episode 95, another
solo episode in a few weeks'time.
Have a great week ahead andthanks so much for listening.
Thank you for joining me forthis episode of the Digital CX
Podcast.
If you like what we're doing,consider leaving us a review on
your podcast platform of choice.
If you're watching on YouTube,leave a comment down below.
(26:37):
It really helps us to grow andprovide value to a broader
audience and get moreinformation about the show and
some of the other things thatwe're doing at
digitalcustomersuccesscom.
I'm Alex Trigovich.
Thanks so much for listening.
We'll talk to you next week.
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.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.