April 11, 2025 • 34 mins
Artificial Intelligence is reshaping the way businesses operate, and Large Language Models (LLMs) like ChatGPT are at the forefront of this transformation. In his presentation at the NextGen Healthcare Summit, James Paden explored how AI-powered tools can enhance productivity, streamline decision-making, and create a competitive edge.
Through this session, attendees gained valuable insights into how organizations can effectively integrate AI—starting with understanding the strategic risks and opportunities, and then evaluating the best approach to adoption: whether buying off-the-shelf solutions, building custom tools in-house, or partnering with AI specialists.
Key points from James’s talk include:
- AI is Your New Competitive Advantage – LLMs like ChatGPT are more than just tools; they act as highly skilled digital assistants, capable of research, writing, and decision-making.
- Adoption Strategies Matter – Choosing the right path—buy, build, or partner—can determine the success of your AI journey.
- Data Drives Success – The success of AI depends heavily on data availability, quality, security, and compliance.
- Learn from the Past – Just as the internet and smartphones transformed industries, AI is poised to do the same. Companies that fail to adapt risk being left behind.
- Infrastructure & Security – A strong technical foundation, privacy safeguards, and access controls are essential for responsible AI implementation.
This audio clip captures the highlights of James’s forward-thinking approach to AI in business—delivering a powerful message about the importance of embracing AI now to stay competitive in the future.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:01):
You're listening to Risk and Resolve, and now for
your hosts, ben Conner and ToddHufford.
Today I want to talk to you alittle bit about practical
strategies for leveraging AI.
I've been working in technology, building software companies,
for about 25 years now and I'venever been more excited about
technology than I am today.
(00:21):
What we're facing, as Benmentioned, is an enormous shift
in the capabilities oftechnology, what it can do for
us, what it can do for ourbusinesses, and today I want to
share a little bit about that.
So we're going to go over whyAI matters now.
Why now?
Why not before?
Why is everyone talking aboutit?
(00:42):
We'll talk about what it can do, a little bit about what it
can't do, and we'll spend mostof our time talking about how to
adopt it in our orgs, where arethe opportunities to use it and
how do we go about doing thatstrategically, intelligently,
not just, you know, willy-nillyall over the place, but how do
we do it with some thought, andthen we'll end with just a tad
(01:02):
on where to go from here.
When you walk out of this roomtoday, what should you be
thinking about?
What are the next steps?
So I'll start with a littlebrief history lesson.
You've probably heard about AIfor decades now.
Right, it's been a buzzword,everyone's making it.
It's been around.
Why is everyone talking aboutit?
(01:23):
Why is it, you know inpresentations?
Why are we giving speeches onit?
Why is everybody raisinghundreds of millions of dollars?
So in the past, we had AI models.
We used computers, we trainedthem.
They could learn things.
They were fantastic atidentifying a stop sign or
identifying a mustache oridentifying if someone's
(01:45):
carrying a weapon or not.
They're sort of single purposemodels.
They were good for one or twothings and they took millions of
dollars to build.
They took engineers, phds.
They were sort of out of thegrasp of companies like yours
and mine.
That's not true anymore.
(02:06):
The current AI models arecompletely different in their
capabilities and in their coststructure.
So about 2017, google inventeda new way to train an AI model
called Transformers.
This model got combined withchips from NVIDIA that were used
(02:28):
for Bitcoin mining.
So these two sort ofdevelopments basically made
possible for the first time toimagine not a small model, not a
model that knows what amustache is or where a stop sign
is, but a model that is largeand that created this thing
called a large language model.
And so to build these largelanguage models, they sucked up
the entire knowledge of humanity, everything that humanity has
(02:51):
ever created in any digital form, and even some non-digital
forms.
They sucked it all into thisbig model and trained it on
everything.
So the word large implies,contrary to the old models that
were good at small tasks andtrained on small sources of data
, this was trained on a largesource of data, the entire
Internet.
So all the books, wikipedia,articles, movies, youtube,
(03:17):
research papers it has it all inthis giant model.
The word language here isimportant as well.
It's not language in terms ofEnglish or grammar.
It's sort of at that rootdefinition of what is a language
.
It's for communicating conceptsand understanding for us
working as humans.
And the large language model,or LLM, does the exact same
(03:41):
thing.
It understands concepts, so itcomputes, takes all these
concepts, all this knowledge,and boils it down to math and at
a mathematical level it's sortof like taking the number for
the word king and you add thenumber for the word woman and it
understands that you're talkingabout a queen.
So before now this wasn'tpossible at this scale.
(04:04):
We didn't have theunderstanding of these concepts.
You take the word child and youadd the number for the word
time and out pops the number forthe word adult.
So for the first time, we haveuniversal knowledge and
understanding that can beapplied in any concept at any
(04:25):
business.
It's not something that has tobe specially trained for your
business.
It's available off the shelf.
It did require tens or hundredsof millions of dollars to train
, but now it is so generallyapplicable that I can access it
for pennies.
You can access it for pennies.
So this is the foundationalshift is that it doesn't take
(04:47):
millions of dollars to make aspecialized model, but for
pennies and for dollars we canall do things that were
previously impossible.
So this opens up an enormousamount of opportunity for all of
us.
On the capability side, what doyou do with a model that knows
(05:07):
everything and understands whatyou tell it?
What can you do with it?
Right?
What can it do for our orgs?
I think, first and foremost,you can take nothing and turn it
into something.
You can do research.
You can ask a question.
We've been using Google foryears, right?
Google is keyword-based.
You type in you know what moviedid Matt Damon star in and you
(05:35):
get a list of Matt Damon movies.
These large language models.
Tools like ChatGPT can beconcept-based.
You can just talk to it aboutsomething.
You vaguely remember what theprotagonist did, and it can find
that information.
So at a business level, we canresearch and find all sorts of
things.
We can look up research papers.
We can find the latest trendsin HR.
We can create something out ofnothing.
(05:58):
Then we can automate that.
So all of that is possible.
We can take a giant mess andturn it into something nice and
clean and organized Again.
This wasn't possible beforewithout a lot of money and a lot
of time specialized to yourparticular use case.
But this strategy can beapplied to any use case.
(06:19):
Ai, current AI models,understand those concepts, that
language within our data.
So if you have a giant pile ofsomething, it previously may not
have had a lot of value and nowit can have a lot of value.
So if you have a giant pile ofresumes or employee descriptions
(06:41):
, you can analyze those andextract information from those
notes, from that free formdocument, maybe skills.
If you've got medical recordsor doctor's notes, you can
extract coding from that rightAnytime.
You have information that's notorganized now we can organize
(07:03):
it.
We can take audio records andtranscripts and organize those
at scale.
All because the modelunderstands the language being
used, understands what we'resaying.
You can take a bunch ofsomething and make it into a
smaller something.
You can condense information.
(07:24):
You can extract the essence ofmeaning out of a document.
So if you have a bunch ofemployee feedback surveys right,
you could squeeze thatinformation down into the
sentiment analysis.
What is that person thinking orfeeling when they say this
large block of text, right?
What is the core meaning thatwe're extracting?
(07:47):
Large block of text, right?
What is the core meaning thatwe're extracting?
If you take, you know, you cansummarize a transcript of a call
.
If you have a sales call or acustomer service call, condense
that down.
I use this a lot.
Where I'll take a sales calland I'll be hey, what are the
key pain points that werereferenced there?
(08:08):
What are the stories, what arethe things that were listed and
said?
What are the action items thoseof you who may have experienced
a lot of the meeting bots onZoom or Fathom I've probably
seen this in action where it cantake a call and then outline
here's the topics discussed,here's the action items.
Right, you can take that sameconcept of extracting and
(08:32):
condensing information down atany part of your business.
Now and again it's availablefor pennies, for dollars.
You go the other way.
You can take a little bit ofsomething and you make it into
something bigger.
You can take an idea, somethingyou just built and you want to
promote it.
Take that idea, that concept,and turn that concept into a
(08:52):
tweet.
Turn that tweet into a LinkedInpost.
Turn that LinkedIn post into aseries of blog posts.
You can generate HR policiesand documentation, all with just
a few sentences.
So you can take a little bit ofsomething, combining it again.
You have a piece of softwarebehind you with the universal
(09:13):
knowledge of the world.
It knows what a blog post lookslike or HR documentation looks
like.
It has a thousand examplessitting in its math code.
Right, you can make thesethings.
You can generate something andexpand upon it.
My favorite you can transform itinto something completely new
(09:35):
and different.
So if you've got a documentthat's full of medical jargon,
you can be like hey, get rid ofthat medical jargon, give it to
me in plain English now, right,like a normal person would.
If you haven't used ChatGPT andasked it to talk like a normal
person, try that.
It's very good at just talkinglike a normal person.
I think we could use more ofthat in the business world.
(09:56):
You can take and translate,obviously, into different
languages, but you could alsotranslate all sorts of things
and transform documents.
One of my favorite examples isyou take a piece of support
documentation and transform thatinto sales material.
You can take the same corenuggets of information but
(10:16):
change the intent of how we'recommunicating with them.
Are we communicating to sell,to convince, to train, to
support?
Do we want to do it in aconcise, easy-to-read format
that's scannable, or do we wantto use more words?
That would be more explanatory?
So all of these things areavailable again at scale across
(10:36):
a wide variety of use cases,wide variety of industries.
We can also use AI to help usmake decisions.
To go, what should we do aboutthat?
And again, we have a giant pile.
That's the history of the world, all of humanity's knowledge
Doesn't mean it's perfect,doesn't mean it's going to do
(10:57):
exactly what we would do,doesn't mean we can't have a
human in the loop, but we canask it.
Hey, pretend to be an expert inthis subject matter area.
What would you do?
Pretend to be an angry customer.
How would you react.
What would you do in thisscenario?
Give me a list of fivedifferent frameworks I can use
(11:20):
to analyze a decision.
It's like having an MBA in yourback pocket, ready to help you
analyze any decision, and thenyou can build automations on top
of that.
You can connect all of theseconcepts together, or you can
start with a giant pile of data,clean it up and then transform
it, then build a decision.
(11:41):
It can loop on this over andover and over again.
Never before have we been ableto do this.
I've worked on, you know, like Isaid, technology for 25 years
now, and I've never had thiskind of capability at my
fingertips where, in seconds, Ican take any document, any piece
(12:02):
of data, and make it intosomething else, and then I can
automate that, and it cost mepractically nothing.
And then I can automate that,and it costs me practically
nothing.
So this is the sort of thetransformative nature of AI is.
It allows all of us to do thisand we all have the same exact
opportunity, right?
Every business, everydepartment is looking at this
and thinking what should I doabout it?
(12:25):
What can I do about it?
What are the opportunities thatexist?
So that's why everyone'stalking about AI right now is
that we're all facing these samequestions and these same
opportunities, the samechallenges.
So hopefully I give you a fewmore answers here as we go along
.
But this is new.
We've never had this capabilitybefore at this scale or at this
(12:47):
price.
So a few case studies of peoplewho have done things with it
Commerce Bank in Germany.
They're a bank and when theytalk to you, they have to.
You know they do some moneytransfer for you over the phone
and something like that.
They have to log that.
They take all those notes,write all that down.
What did they do?
What did they talk about?
What happened?
They're a bank.
(13:08):
They have an audit trail.
Well, now they put thattranscript into AI and the AI
generates all that.
They still have an employeereview it.
They're still an employee inthe loop, but now it's minutes
instead of lots of minutes or anhour.
Details are available on Google.
Bloomreach, big e-commercecompany, doubled their blog
posts, with a 40% increase intraffic, by using an AI tool
(13:31):
called Jasper for their contentmarketing efforts.
So expanded their efforts, 40%increase in traffic.
A lot of these case studies arebased off, let's say, a year
ago's AI capabilities as well.
So companies today are doingeven more powerful things,
getting even more results.
I used AI to find all thesecase studies.
(13:52):
You can too.
So a company called 123RFreduced by 95% their content
translation.
So they're a multilingualcompany and they translated it
all now using generative AItools, and it's not like they
weren't using computertranslation before.
(14:13):
They were right, but the newermodels are just that much
cheaper, that much faster.
So when you think about at yourorg, your department, your team
, what could you do if youreduced the cost of doing
something by 95%?
That's what we can do with alot of these tools and a lot of
these use cases.
Thinkbridge, as a consultingcompany, it turns out they hire
(14:35):
a lot of people, so they build achatbot system to help them I
think mainly internally accesstheir recruiting and HR
information better, access theirrecruiting and HR information
better.
So that system led to a 35%increase in recruiter
productivity and that led to a40% increase in candidate
engagement.
(14:56):
By having a better, fastercommunication path, they
increased their ultimate goal,which was engagement.
So those are again just a brief.
There are tens, hundreds ofthese kinds of case studies out
there.
You can ask ChatGPT to help youresearch them and you'll find a
bunch more, perhaps moreapplicable to your industry.
(15:18):
But if that's what's possible,right?
How do we adopt itstrategically?
How do we think about it at ahigher level and make smart
decisions?
So, first thing you can do isyou can just sprinkle a little
AI on top.
Nope, I'm lying, don't do that.
Do not do that.
Do not be the person in theroom who says, just well, just
AI can do it.
(15:38):
Right, let's think smart.
The first way to think smart isto understand how they work, to
use it at a personal level.
Right, if you're not using chat, gpt, the tools provided by
your organization, whatever itmight be, get some experience
with these tools.
They're fallible.
They make mistakes At theircore level.
We're skipping a lot of thetechnical details here, but this
(16:00):
core level, generative AI is aguessing machine.
It gets better and smarter atguessing the right answer with
every iteration, but it's justguessing, and sometimes it
guesses with 90% accuracy andsometimes it guesses with 99.9%
accuracy, but ultimately it's aguess.
There's an inherent randomnessin the system so that it sounds
(16:22):
more natural and human, becausethat's how we talk.
What I say, the way I say asentence, is going to be
different than the way Jason,when he speaks next, says that
same sentence, and AI does thesame thing.
So it makes mistakes.
So if you want to know how itmakes mistakes, what its
limitations are, practice, getused to it.
Work on these at a personallevel so that you're able to
(16:43):
apply these automations at anorganizational level, so that
you can have smarter,intelligent conversations when
it comes time to guess well,what can we use AI for?
How would this act on thisdocument?
It's not just sprinkle AI magicon top, but it's.
Well, let's apply thistransformation.
Let's use this prompt it canhandle this use case.
(17:05):
Let's use this prompt it canhandle this use case.
So, at a deeper level, though,when I think about when I work
with clients and organizationsimplementing AI, we talk about
sort of a three level pyramid.
So, at the base level, we haveour supporting functions.
This is the departments in yourorganizations.
This is HR, engineering,finance.
(17:26):
They support the organization.
They help the organization dowhatever the organization does.
In the middle, you have yourcore processes and experience.
This is what you do.
You know the classic moviequote.
You know what would you say youdo here?
What is your company's answerfor this question?
This is what you do how datamoves through your company, how
services and products movethrough your company.
(17:48):
What is it you're selling?
That's your core.
And then, at the top level,there's that strategic
differentiation.
What do you do better thananybody else?
Why do customers pick yourcompany and not your
competitor's company?
What is that unique aspect?
So, at each of these threelevels, ai can assist and do
(18:10):
different things right.
And so what we want to think issort of level by level where are
the opportunities to apply AIin my business?
And obviously they have greaterimpacts the higher you go up.
So step one is identify somehigh-value use cases right.
Each one of these levels.
We're going to sort of gothrough this exercise.
(18:32):
You're going to think aboutwhat is the business impact if I
had AI here, if I couldtransform data or clean data or
generate documentation ortranslation at massive scale and
minimal cost?
What would the business impactof a shift like that be?
Just a theoretical level, right?
Do I have any data that gives mean advantage in doing that
(18:54):
thing, that makes it easier ormaybe harder to do that thing?
What do I have?
How long will it take toimplement?
How ready am I to implementthat?
Do I have people that can helpme?
Do I know consultants that canhelp me?
Do I have an IT team?
How ready am I to implementthese things?
(19:16):
And then, where can they have,apply and create that
competitive differentiation?
Do you have access to a teamthat your competitor doesn't, or
data that your competitordoesn't, or are you in better
shape?
So the joke among AIconsultants is that every AI
project is actually a dataproject.
(19:39):
In order to do any of thesethings, you have to be able to
give the machines your data andsay, hey, here's a something.
Whatever that something iscould be a record, notes,
transcripts, marketing materialyou have to be able to give it
to the machine right, and dothat in a secure and safe format
.
So when you're talking aboutyour data, first off think about
(20:00):
availability.
Is it sitting on Joe's computerand when Joe goes home for the
day, that's it.
The data's gone.
Or is it in the cloud somewhere?
Is it accessible and availableto the AI in a way that can be
connected?
Is your data of high quality,or is it a mess with lots of
duplicate data and incorrectrecords and some mixture of old
(20:23):
stuff and new stuff?
Ai can help clean up a lot ofthis right, but that's sort of a
separate process than justactually using it.
So what state is your data andyour documents in?
Are they accessible?
How secure?
Are they right?
If you've got a whole list ofcustomer records or financial
records, you'll want someemployees to have access to some
(20:46):
of those documents.
You probably have a lot ofthese controls available today,
but if you upload all thosedocuments into ChatGPT or
various tools, does it offerthat same level of data control?
I know we're talking in thehealthcare industry.
There's probably a lot ofcompliance privacy regulations
at play.
(21:06):
How do you continue to followthose regulations within these
new tools and frameworks?
So, and lastly, do you have thetechnical infrastructure to
manage all this?
Is it again in the cloud?
Is it available?
Where is it at?
So these are kind of questionsto be thinking about as you
process all this.
You can ask AI to help you withany of these.
(21:27):
You know you don't always haveto work with a consultant.
You can start with just askingAI.
It again sort of knowseverything.
It has all that knowledge.
If you use a paid version ofChatGPT or some of these other
tools to get started at apersonal level.
They have a lot of researchpower.
They are up to date.
I used to have a slide in mydecks that says here's the four
(21:49):
or five things AI can't do.
Here's what you need to watchout for.
I had to take that slide outbecause they've changed so much.
Right, they can do most ofthose things.
So if you start using ChatGPTat that personal executive level
, sign up for a paid plan.
You get access to the lateststuff and then you can just ask
it to help you.
So the ultimate here is thatyou sort of get this
(22:13):
cross-section between impact andeffort, and in the top right
you have your strategicinvestments this cross-section
between impact and effort, andin the top right you have your
strategic investments.
This slide's probably a littleboring for most of you because
you've seen a thousand examplesof it.
Which is the point?
This isn't anything new.
The way we evaluate theseopportunities is not new.
It's not groundbreaking.
The way we implement them isrelatively standard software
(22:36):
practices.
Jason's been my next speaker.
He's been working on this fordecades.
Right, ai gives him new toolsand new advantages.
With generative AI, the core wayit works is the same.
The same way we make decisionsis the same.
I had a potential client I wasworking with and they asked for
(22:57):
a governance framework and wehad this whole 10-minute
conversation where I tried toget them to tell me what they
wanted and I didn't understand.
Ultimately, what they wantedwas a way to figure out how to
make decisions and prioritize AIprojects.
How do you prioritize projectstoday?
It's the same.
What's new is that we can do alot more things that we could
(23:21):
not do yesterday.
Right, the possibilities arenew and there's a lot more of
them coming at us faster thanever before.
Back when, you know, appleinvented the smartphone, over
the next few years, everycompany had to ask themselves do
we build an app?
Does it make sense for us tohave an app?
It was sort of a singulardecision and we had some years
to make it, as everyone gotsmartphones.
(23:43):
That's not the case for AI.
There's a lot more decisions tobe made.
There's a lot moreopportunities to be had, to be
seized by forward-lookingcompanies, by forward-looking
individuals to further theircareer.
There's a lot of opportunitieson the table in front of us and
(24:04):
everyone has access to them allat the same time.
It's not like the smartphone,where it gradually got rolled
out to 10% of the population,and now everyone's got a
smartphone in their pocket.
This is a slightly differentballgame in terms of the
opportunities in front of us,but the way we make decisions,
the way we think about it andlook at those opportunities is
the same way we look at businessopportunities today.
So, for each project, how areyou going to implement it?
(24:26):
Right?
Your base options roughly fallinto the same three categories
most projects fall into.
Are you going to buy it off theshelf?
Are you going to build in-house?
Are you going to partner with aspecialist or consultant, such
as myself?
When we think about the threelevels, they roughly break down
and this is fuzzy.
(24:46):
It's labels, categories, it'sall fuzzy, but roughly.
What I tend to recommend mostclients is if you're talking
about the base functions, ifyou're talking about something
that your marketing team needs,your HR team, your engineering
team, try to find off-the-shelfsolutions for those.
They're already using softwaresuites, software tools, wait, or
(25:09):
look for solutions within thoseexisting suite of tools.
I guarantee you, every softwarecompany in the world right now
is trying to figure out how AImakes sense within their product
.
You don't have to build itcustom.
You can use a lot of yourexisting technical stack to take
advantage of the opportunitiesof AI at sort of that
foundational level.
Your finance department doesnot need to reinvent the wheel.
(25:31):
It's coming.
So if it's not there yet, youcan be patient.
You can look for a new softwarevendor.
But a lot of the stuff isavailable off the shelf or with
existing software packages.
You don't have to build it, youcan just go buy it.
The middle layer, on the otherhand, I'd argue that that is the
opportunity to work with aconsultant or a specialist,
(25:52):
someone who does have thegroundbreaking latest knowledge
of what AI can do, where it cando it, where it can do it most
cost effectively.
Every few months, ai changes.
I mentioned that slide I had totake out of the deck.
Every couple months there's anew model, a new way of doing
things, and so working with aspecialist who has the latest
(26:12):
knowledge can get you therefaster.
Right, if you don't have thein-house team or in-house
knowledge to build these things.
This is where we believe thatthere's a lot of opportunity to
move.
Fast is in that core.
What does your company do here?
How can you accelerate thatchange, that with AI At the top
level, where we're building thatstrategic differentiation.
(26:35):
That's where I'd argue.
You make those investments, youbuild it in-house.
What can you do that no oneelse can do At the AI level?
That's a lot about what data doyou have that no one else has?
We all have access to chat, gptand the base large language
models.
That's not unique to my companyor your company.
They're practically giving itaway out there.
(26:56):
That's not unique to my companyor your company.
They're practically giving itaway out there.
But what do you have insideyour walls of your organization
that no one else has?
And how can you leverage AI tomake that more valuable, more
effective?
And that's where I thinkthere's a lot of opportunities
to make some investment to buildit in-house.
Again, you can work withspecialists at any of these
levels.
If you've got an in-house ITteam with spare cycles because I
(27:17):
know we're all sitting aroundbored all day Software guys they
can build any of this at anylevel.
But most of the time, I seeprojects sort of broken down by
these categories.
(27:42):
Lastly, I'll talk just brieflyhere about where to go from here
how to lead, be a leader inyour organization, how to equip
the rest of your team to helpyou on this journey to to be
forward thinking as well.
So, first and foremost, if youdon't have an AI policy within
your team, within yourdepartment, your org, make one
Right.
Tell your team what they canand cannot do, starting with
(28:03):
who's in charge of this thing.
Who do they go to when theyhave a question or want to make
an exception?
And they will want to makeexceptions.
There'll be lots of exceptions,of people trying to ideally,
push the boundaries, find newtools, to use new ways to do
things, but make sure everyoneknows who is the decision maker
(28:23):
on AI policy in yourorganization.
And then, what can they do?
What tools can they use?
Can they use ChatGPT?
Can they use Cloud or Copilot?
Can they use something that'sbaked in their software?
Can they turn on that AI togglein the settings?
What can they do?
You need to evaluate the risksinvolved in that and tell them
(28:46):
what they can use.
I will say if you tell themthey can't use anything, they're
just going to go to ChatGPTbehind your back and start using
those tools right, guarantee it.
So figure out what they'reallowed to use, how they can use
it.
If you don't know, chatgpt willtake your data and use it for
training the next model bydefault.
(29:07):
The default setting on ChatGPTis we're going to use your data
in our next training model, soit is important to have trained
employees, to be trainedyourselves and know how do I use
these tools securely, so theyprotect privacy.
When configured properly, Ibelieve these tools are just as
(29:29):
safe as using Salesforce,hubspot any number of the
software systems we use today.
Right, we treat those, weupload all sorts of documents
into those systems, and I thinkwe can do the same with many AI
tools.
But it's only if they'reconfigured properly and you've
had the training and educationto know how that works.
What data can be uploadedwithin these tools?
(29:51):
Is it anything?
Is it financial data?
Is customer data okay?
What if it's anonymized?
Be specific, provide guidanceto your team.
And then, is it necessary todisclose the AI usage?
Generally, I'd argue no.
Maybe if the AI built thisthing completely with an
automation, maybe you put adisclaimer that AI made that,
(30:13):
but this is one of these rapidlychanging systems, and why you
need to update your AI policyall the time is because people's
expectations around AI usagehas shifted so much over the
last couple of years.
I would expect it's rare in aprofessional setting for me to
see a document that hasn't beentouched by an AI model.
I certainly used it to helpmake my presentations, to write
(30:35):
my proposals, to review all mylegal documents, right.
But you know you decide what'sbest for your business and your
industry.
The important thing is that youcommunicate and again, this
isn't groundbreaking.
This isn't a new way of doingbusiness.
It's the same old business.
It's the same old.
You know change managementtechniques.
Same old communicationtechniques.
It's just a whole new areathat's changing rapidly and we
(30:58):
need to stay on top of it.
You can lead by example.
Be public about your usage, thefact that you're using AI, that
it's okay to use AI.
I think the ideal situation isyou have a whole team of people
who feel empowered and excitedby the opportunities in front of
them with AI.
That's the whole thing we'vetalked about today is that, at a
(31:18):
strategic level, this isentirely new capabilities.
We never could do this before.
So lead, show them it's okay.
Show them it's okay to do thesethings.
Get them trained.
We talked about how that'simportant.
There's a big difference betweensomeone who's using AI and not
trained and someone who istrained.
The prompts, the instructionswe provide AI.
(31:41):
There's specific techniques.
You can use a lot of tips andtricks you'll get from either
personal experimentation or withtraining, so make sure your
people are equipped.
It's the same as anything else.
Right?
You want to have trained,educated people, that's just.
The entire world is having tolearn at the same time how to
(32:02):
use these tools.
So don't leave your peoplebehind.
Get them trained, standardizeand share best practices.
If you find something thatworks well for your organization
, a use case that works well,share that.
Show your own internal casestudies.
The most impact comes when westart to automate processes,
(32:23):
when we build these things intoour core processes.
But that starts by saying hey,I uploaded this survey into
ChatGPT and I was able to findinsight, I was able to condense
results, I was able to dosomething five times faster by
using these tools and when youshare and document these things
with your org, that's how youfind these opportunities and
these use cases across the org,across departments, by
(32:45):
standardizing and sharing.
So, that said, I'll wrap uphere Points.
Main points are AI has unlockednew capabilities.
Main points are AI has unlockednew capabilities.
Never before.
I keep saying it, this is whywe're talking about it.
It is a monumental shift inwhat is possible and what we
(33:08):
could do.
Consider the use cases for eachlevel.
You have your supportingfunctions, your supporting teams
, you have your core processesand teams and then you have that
little nugget of what makes mycompany unique.
So take the time to think abouthow AI can be applied at each
of those levels within yourorganization.
Decide how to develop that usecase.
(33:28):
Does it make sense to build?
What's the ROI impact standard?
How do I prioritize thisproject kind of stuff?
Is this something I'm going towork with a consultant on?
Am I going to try and build itin-house?
What's my capacity for buildingthese projects in-house?
How many can I tackle?
And then, lastly, lead yourorganization into this future.
You know I'm excited about it.
(33:48):
I do this all day, every day.
This is the most fun I've everhad in technology.
Like it's like a kid in a candyshop out here for an engineer
uh, all sorts of groundbreakingnew things that were not
possible, um.
But I think the same applies atthe organizational level.
As you're leading a team,there's all sorts of new
possibilities, new ways of doingthings, and by equipping
(34:12):
yourself to lead into thatfuture, your team will see the
most results.
So thank you guys.
Thanks for tuning in to Riskand Resolve.
See you next time.
You're listening to Risk and Resolve, and now for
your hosts, ben Conner and ToddHufford.
Today I want to talk to you alittle bit about practical
strategies for leveraging AI.
I've been working in technology, building software companies,
for about 25 years now and I'venever been more excited about
technology than I am today.
(00:21):
What we're facing, as Benmentioned, is an enormous shift
in the capabilities oftechnology, what it can do for
us, what it can do for ourbusinesses, and today I want to
share a little bit about that.
So we're going to go over whyAI matters now.
Why now?
Why not before?
Why is everyone talking aboutit?
(00:42):
We'll talk about what it can do, a little bit about what it
can't do, and we'll spend mostof our time talking about how to
adopt it in our orgs, where arethe opportunities to use it and
how do we go about doing thatstrategically, intelligently,
not just, you know, willy-nillyall over the place, but how do
we do it with some thought, andthen we'll end with just a tad
(01:02):
on where to go from here.
When you walk out of this roomtoday, what should you be
thinking about?
What are the next steps?
So I'll start with a littlebrief history lesson.
You've probably heard about AIfor decades now.
Right, it's been a buzzword,everyone's making it.
It's been around.
Why is everyone talking aboutit?
(01:23):
Why is it, you know inpresentations?
Why are we giving speeches onit?
Why is everybody raisinghundreds of millions of dollars?
So in the past, we had AI models.
We used computers, we trainedthem.
They could learn things.
They were fantastic atidentifying a stop sign or
identifying a mustache oridentifying if someone's
(01:45):
carrying a weapon or not.
They're sort of single purposemodels.
They were good for one or twothings and they took millions of
dollars to build.
They took engineers, phds.
They were sort of out of thegrasp of companies like yours
and mine.
That's not true anymore.
(02:06):
The current AI models arecompletely different in their
capabilities and in their coststructure.
So about 2017, google inventeda new way to train an AI model
called Transformers.
This model got combined withchips from NVIDIA that were used
(02:28):
for Bitcoin mining.
So these two sort ofdevelopments basically made
possible for the first time toimagine not a small model, not a
model that knows what amustache is or where a stop sign
is, but a model that is largeand that created this thing
called a large language model.
And so to build these largelanguage models, they sucked up
the entire knowledge of humanity, everything that humanity has
(02:51):
ever created in any digital form, and even some non-digital
forms.
They sucked it all into thisbig model and trained it on
everything.
So the word large implies,contrary to the old models that
were good at small tasks andtrained on small sources of data
, this was trained on a largesource of data, the entire
Internet.
So all the books, wikipedia,articles, movies, youtube,
(03:17):
research papers it has it all inthis giant model.
The word language here isimportant as well.
It's not language in terms ofEnglish or grammar.
It's sort of at that rootdefinition of what is a language
.
It's for communicating conceptsand understanding for us
working as humans.
And the large language model,or LLM, does the exact same
(03:41):
thing.
It understands concepts, so itcomputes, takes all these
concepts, all this knowledge,and boils it down to math and at
a mathematical level it's sortof like taking the number for
the word king and you add thenumber for the word woman and it
understands that you're talkingabout a queen.
So before now this wasn'tpossible at this scale.
(04:04):
We didn't have theunderstanding of these concepts.
You take the word child and youadd the number for the word
time and out pops the number forthe word adult.
So for the first time, we haveuniversal knowledge and
understanding that can beapplied in any concept at any
(04:25):
business.
It's not something that has tobe specially trained for your
business.
It's available off the shelf.
It did require tens or hundredsof millions of dollars to train
, but now it is so generallyapplicable that I can access it
for pennies.
You can access it for pennies.
So this is the foundationalshift is that it doesn't take
(04:47):
millions of dollars to make aspecialized model, but for
pennies and for dollars we canall do things that were
previously impossible.
So this opens up an enormousamount of opportunity for all of
us.
On the capability side, what doyou do with a model that knows
(05:07):
everything and understands whatyou tell it?
What can you do with it?
Right?
What can it do for our orgs?
I think, first and foremost,you can take nothing and turn it
into something.
You can do research.
You can ask a question.
We've been using Google foryears, right?
Google is keyword-based.
You type in you know what moviedid Matt Damon star in and you
(05:35):
get a list of Matt Damon movies.
These large language models.
Tools like ChatGPT can beconcept-based.
You can just talk to it aboutsomething.
You vaguely remember what theprotagonist did, and it can find
that information.
So at a business level, we canresearch and find all sorts of
things.
We can look up research papers.
We can find the latest trendsin HR.
We can create something out ofnothing.
(05:58):
Then we can automate that.
So all of that is possible.
We can take a giant mess andturn it into something nice and
clean and organized Again.
This wasn't possible beforewithout a lot of money and a lot
of time specialized to yourparticular use case.
But this strategy can beapplied to any use case.
(06:19):
Ai, current AI models,understand those concepts, that
language within our data.
So if you have a giant pile ofsomething, it previously may not
have had a lot of value and nowit can have a lot of value.
So if you have a giant pile ofresumes or employee descriptions
(06:41):
, you can analyze those andextract information from those
notes, from that free formdocument, maybe skills.
If you've got medical recordsor doctor's notes, you can
extract coding from that rightAnytime.
You have information that's notorganized now we can organize
(07:03):
it.
We can take audio records andtranscripts and organize those
at scale.
All because the modelunderstands the language being
used, understands what we'resaying.
You can take a bunch ofsomething and make it into a
smaller something.
You can condense information.
(07:24):
You can extract the essence ofmeaning out of a document.
So if you have a bunch ofemployee feedback surveys right,
you could squeeze thatinformation down into the
sentiment analysis.
What is that person thinking orfeeling when they say this
large block of text, right?
What is the core meaning thatwe're extracting?
(07:47):
Large block of text, right?
What is the core meaning thatwe're extracting?
If you take, you know, you cansummarize a transcript of a call
.
If you have a sales call or acustomer service call, condense
that down.
I use this a lot.
Where I'll take a sales calland I'll be hey, what are the
key pain points that werereferenced there?
(08:08):
What are the stories, what arethe things that were listed and
said?
What are the action items thoseof you who may have experienced
a lot of the meeting bots onZoom or Fathom I've probably
seen this in action where it cantake a call and then outline
here's the topics discussed,here's the action items.
Right, you can take that sameconcept of extracting and
(08:32):
condensing information down atany part of your business.
Now and again it's availablefor pennies, for dollars.
You go the other way.
You can take a little bit ofsomething and you make it into
something bigger.
You can take an idea, somethingyou just built and you want to
promote it.
Take that idea, that concept,and turn that concept into a
(08:52):
tweet.
Turn that tweet into a LinkedInpost.
Turn that LinkedIn post into aseries of blog posts.
You can generate HR policiesand documentation, all with just
a few sentences.
So you can take a little bit ofsomething, combining it again.
You have a piece of softwarebehind you with the universal
(09:13):
knowledge of the world.
It knows what a blog post lookslike or HR documentation looks
like.
It has a thousand examplessitting in its math code.
Right, you can make thesethings.
You can generate something andexpand upon it.
My favorite you can transform itinto something completely new
(09:35):
and different.
So if you've got a documentthat's full of medical jargon,
you can be like hey, get rid ofthat medical jargon, give it to
me in plain English now, right,like a normal person would.
If you haven't used ChatGPT andasked it to talk like a normal
person, try that.
It's very good at just talkinglike a normal person.
I think we could use more ofthat in the business world.
(09:56):
You can take and translate,obviously, into different
languages, but you could alsotranslate all sorts of things
and transform documents.
One of my favorite examples isyou take a piece of support
documentation and transform thatinto sales material.
You can take the same corenuggets of information but
(10:16):
change the intent of how we'recommunicating with them.
Are we communicating to sell,to convince, to train, to
support?
Do we want to do it in aconcise, easy-to-read format
that's scannable, or do we wantto use more words?
That would be more explanatory?
So all of these things areavailable again at scale across
(10:36):
a wide variety of use cases,wide variety of industries.
We can also use AI to help usmake decisions.
To go, what should we do aboutthat?
And again, we have a giant pile.
That's the history of the world, all of humanity's knowledge
Doesn't mean it's perfect,doesn't mean it's going to do
(10:57):
exactly what we would do,doesn't mean we can't have a
human in the loop, but we canask it.
Hey, pretend to be an expert inthis subject matter area.
What would you do?
Pretend to be an angry customer.
How would you react.
What would you do in thisscenario?
Give me a list of fivedifferent frameworks I can use
(11:20):
to analyze a decision.
It's like having an MBA in yourback pocket, ready to help you
analyze any decision, and thenyou can build automations on top
of that.
You can connect all of theseconcepts together, or you can
start with a giant pile of data,clean it up and then transform
it, then build a decision.
(11:41):
It can loop on this over andover and over again.
Never before have we been ableto do this.
I've worked on, you know, like Isaid, technology for 25 years
now, and I've never had thiskind of capability at my
fingertips where, in seconds, Ican take any document, any piece
(12:02):
of data, and make it intosomething else, and then I can
automate that, and it cost mepractically nothing.
And then I can automate that,and it costs me practically
nothing.
So this is the sort of thetransformative nature of AI is.
It allows all of us to do thisand we all have the same exact
opportunity, right?
Every business, everydepartment is looking at this
and thinking what should I doabout it?
(12:25):
What can I do about it?
What are the opportunities thatexist?
So that's why everyone'stalking about AI right now is
that we're all facing these samequestions and these same
opportunities, the samechallenges.
So hopefully I give you a fewmore answers here as we go along
.
But this is new.
We've never had this capabilitybefore at this scale or at this
(12:47):
price.
So a few case studies of peoplewho have done things with it
Commerce Bank in Germany.
They're a bank and when theytalk to you, they have to.
You know they do some moneytransfer for you over the phone
and something like that.
They have to log that.
They take all those notes,write all that down.
What did they do?
What did they talk about?
What happened?
They're a bank.
(13:08):
They have an audit trail.
Well, now they put thattranscript into AI and the AI
generates all that.
They still have an employeereview it.
They're still an employee inthe loop, but now it's minutes
instead of lots of minutes or anhour.
Details are available on Google.
Bloomreach, big e-commercecompany, doubled their blog
posts, with a 40% increase intraffic, by using an AI tool
(13:31):
called Jasper for their contentmarketing efforts.
So expanded their efforts, 40%increase in traffic.
A lot of these case studies arebased off, let's say, a year
ago's AI capabilities as well.
So companies today are doingeven more powerful things,
getting even more results.
I used AI to find all thesecase studies.
(13:52):
You can too.
So a company called 123RFreduced by 95% their content
translation.
So they're a multilingualcompany and they translated it
all now using generative AItools, and it's not like they
weren't using computertranslation before.
(14:13):
They were right, but the newermodels are just that much
cheaper, that much faster.
So when you think about at yourorg, your department, your team
, what could you do if youreduced the cost of doing
something by 95%?
That's what we can do with alot of these tools and a lot of
these use cases.
Thinkbridge, as a consultingcompany, it turns out they hire
(14:35):
a lot of people, so they build achatbot system to help them I
think mainly internally accesstheir recruiting and HR
information better, access theirrecruiting and HR information
better.
So that system led to a 35%increase in recruiter
productivity and that led to a40% increase in candidate
engagement.
(14:56):
By having a better, fastercommunication path, they
increased their ultimate goal,which was engagement.
So those are again just a brief.
There are tens, hundreds ofthese kinds of case studies out
there.
You can ask ChatGPT to help youresearch them and you'll find a
bunch more, perhaps moreapplicable to your industry.
(15:18):
But if that's what's possible,right?
How do we adopt itstrategically?
How do we think about it at ahigher level and make smart
decisions?
So, first thing you can do isyou can just sprinkle a little
AI on top.
Nope, I'm lying, don't do that.
Do not do that.
Do not be the person in theroom who says, just well, just
AI can do it.
(15:38):
Right, let's think smart.
The first way to think smart isto understand how they work, to
use it at a personal level.
Right, if you're not using chat, gpt, the tools provided by
your organization, whatever itmight be, get some experience
with these tools.
They're fallible.
They make mistakes At theircore level.
We're skipping a lot of thetechnical details here, but this
(16:00):
core level, generative AI is aguessing machine.
It gets better and smarter atguessing the right answer with
every iteration, but it's justguessing, and sometimes it
guesses with 90% accuracy andsometimes it guesses with 99.9%
accuracy, but ultimately it's aguess.
There's an inherent randomnessin the system so that it sounds
(16:22):
more natural and human, becausethat's how we talk.
What I say, the way I say asentence, is going to be
different than the way Jason,when he speaks next, says that
same sentence, and AI does thesame thing.
So it makes mistakes.
So if you want to know how itmakes mistakes, what its
limitations are, practice, getused to it.
Work on these at a personallevel so that you're able to
(16:43):
apply these automations at anorganizational level, so that
you can have smarter,intelligent conversations when
it comes time to guess well,what can we use AI for?
How would this act on thisdocument?
It's not just sprinkle AI magicon top, but it's.
Well, let's apply thistransformation.
Let's use this prompt it canhandle this use case.
(17:05):
Let's use this prompt it canhandle this use case.
So, at a deeper level, though,when I think about when I work
with clients and organizationsimplementing AI, we talk about
sort of a three level pyramid.
So, at the base level, we haveour supporting functions.
This is the departments in yourorganizations.
This is HR, engineering,finance.
(17:26):
They support the organization.
They help the organization dowhatever the organization does.
In the middle, you have yourcore processes and experience.
This is what you do.
You know the classic moviequote.
You know what would you say youdo here?
What is your company's answerfor this question?
This is what you do how datamoves through your company, how
services and products movethrough your company.
(17:48):
What is it you're selling?
That's your core.
And then, at the top level,there's that strategic
differentiation.
What do you do better thananybody else?
Why do customers pick yourcompany and not your
competitor's company?
What is that unique aspect?
So, at each of these threelevels, ai can assist and do
(18:10):
different things right.
And so what we want to think issort of level by level where are
the opportunities to apply AIin my business?
And obviously they have greaterimpacts the higher you go up.
So step one is identify somehigh-value use cases right.
Each one of these levels.
We're going to sort of gothrough this exercise.
(18:32):
You're going to think aboutwhat is the business impact if I
had AI here, if I couldtransform data or clean data or
generate documentation ortranslation at massive scale and
minimal cost?
What would the business impactof a shift like that be?
Just a theoretical level, right?
Do I have any data that gives mean advantage in doing that
(18:54):
thing, that makes it easier ormaybe harder to do that thing?
What do I have?
How long will it take toimplement?
How ready am I to implementthat?
Do I have people that can helpme?
Do I know consultants that canhelp me?
Do I have an IT team?
How ready am I to implementthese things?
(19:16):
And then, where can they have,apply and create that
competitive differentiation?
Do you have access to a teamthat your competitor doesn't, or
data that your competitordoesn't, or are you in better
shape?
So the joke among AIconsultants is that every AI
project is actually a dataproject.
(19:39):
In order to do any of thesethings, you have to be able to
give the machines your data andsay, hey, here's a something.
Whatever that something iscould be a record, notes,
transcripts, marketing materialyou have to be able to give it
to the machine right, and dothat in a secure and safe format
.
So when you're talking aboutyour data, first off think about
(20:00):
availability.
Is it sitting on Joe's computerand when Joe goes home for the
day, that's it.
The data's gone.
Or is it in the cloud somewhere?
Is it accessible and availableto the AI in a way that can be
connected?
Is your data of high quality,or is it a mess with lots of
duplicate data and incorrectrecords and some mixture of old
(20:23):
stuff and new stuff?
Ai can help clean up a lot ofthis right, but that's sort of a
separate process than justactually using it.
So what state is your data andyour documents in?
Are they accessible?
How secure?
Are they right?
If you've got a whole list ofcustomer records or financial
records, you'll want someemployees to have access to some
(20:46):
of those documents.
You probably have a lot ofthese controls available today,
but if you upload all thosedocuments into ChatGPT or
various tools, does it offerthat same level of data control?
I know we're talking in thehealthcare industry.
There's probably a lot ofcompliance privacy regulations
at play.
(21:06):
How do you continue to followthose regulations within these
new tools and frameworks?
So, and lastly, do you have thetechnical infrastructure to
manage all this?
Is it again in the cloud?
Is it available?
Where is it at?
So these are kind of questionsto be thinking about as you
process all this.
You can ask AI to help you withany of these.
(21:27):
You know you don't always haveto work with a consultant.
You can start with just askingAI.
It again sort of knowseverything.
It has all that knowledge.
If you use a paid version ofChatGPT or some of these other
tools to get started at apersonal level.
They have a lot of researchpower.
They are up to date.
I used to have a slide in mydecks that says here's the four
(21:49):
or five things AI can't do.
Here's what you need to watchout for.
I had to take that slide outbecause they've changed so much.
Right, they can do most ofthose things.
So if you start using ChatGPTat that personal executive level
, sign up for a paid plan.
You get access to the lateststuff and then you can just ask
it to help you.
So the ultimate here is thatyou sort of get this
(22:13):
cross-section between impact andeffort, and in the top right
you have your strategicinvestments this cross-section
between impact and effort, andin the top right you have your
strategic investments.
This slide's probably a littleboring for most of you because
you've seen a thousand examplesof it.
Which is the point?
This isn't anything new.
The way we evaluate theseopportunities is not new.
It's not groundbreaking.
The way we implement them isrelatively standard software
(22:36):
practices.
Jason's been my next speaker.
He's been working on this fordecades.
Right, ai gives him new toolsand new advantages.
With generative AI, the core wayit works is the same.
The same way we make decisionsis the same.
I had a potential client I wasworking with and they asked for
(22:57):
a governance framework and wehad this whole 10-minute
conversation where I tried toget them to tell me what they
wanted and I didn't understand.
Ultimately, what they wantedwas a way to figure out how to
make decisions and prioritize AIprojects.
How do you prioritize projectstoday?
It's the same.
What's new is that we can do alot more things that we could
(23:21):
not do yesterday.
Right, the possibilities arenew and there's a lot more of
them coming at us faster thanever before.
Back when, you know, appleinvented the smartphone, over
the next few years, everycompany had to ask themselves do
we build an app?
Does it make sense for us tohave an app?
It was sort of a singulardecision and we had some years
to make it, as everyone gotsmartphones.
(23:43):
That's not the case for AI.
There's a lot more decisions tobe made.
There's a lot moreopportunities to be had, to be
seized by forward-lookingcompanies, by forward-looking
individuals to further theircareer.
There's a lot of opportunitieson the table in front of us and
(24:04):
everyone has access to them allat the same time.
It's not like the smartphone,where it gradually got rolled
out to 10% of the population,and now everyone's got a
smartphone in their pocket.
This is a slightly differentballgame in terms of the
opportunities in front of us,but the way we make decisions,
the way we think about it andlook at those opportunities is
the same way we look at businessopportunities today.
So, for each project, how areyou going to implement it?
(24:26):
Right?
Your base options roughly fallinto the same three categories
most projects fall into.
Are you going to buy it off theshelf?
Are you going to build in-house?
Are you going to partner with aspecialist or consultant, such
as myself?
When we think about the threelevels, they roughly break down
and this is fuzzy.
(24:46):
It's labels, categories, it'sall fuzzy, but roughly.
What I tend to recommend mostclients is if you're talking
about the base functions, ifyou're talking about something
that your marketing team needs,your HR team, your engineering
team, try to find off-the-shelfsolutions for those.
They're already using softwaresuites, software tools, wait, or
(25:09):
look for solutions within thoseexisting suite of tools.
I guarantee you, every softwarecompany in the world right now
is trying to figure out how AImakes sense within their product
.
You don't have to build itcustom.
You can use a lot of yourexisting technical stack to take
advantage of the opportunitiesof AI at sort of that
foundational level.
Your finance department doesnot need to reinvent the wheel.
(25:31):
It's coming.
So if it's not there yet, youcan be patient.
You can look for a new softwarevendor.
But a lot of the stuff isavailable off the shelf or with
existing software packages.
You don't have to build it, youcan just go buy it.
The middle layer, on the otherhand, I'd argue that that is the
opportunity to work with aconsultant or a specialist,
(25:52):
someone who does have thegroundbreaking latest knowledge
of what AI can do, where it cando it, where it can do it most
cost effectively.
Every few months, ai changes.
I mentioned that slide I had totake out of the deck.
Every couple months there's anew model, a new way of doing
things, and so working with aspecialist who has the latest
(26:12):
knowledge can get you therefaster.
Right, if you don't have thein-house team or in-house
knowledge to build these things.
This is where we believe thatthere's a lot of opportunity to
move.
Fast is in that core.
What does your company do here?
How can you accelerate thatchange, that with AI At the top
level, where we're building thatstrategic differentiation.
(26:35):
That's where I'd argue.
You make those investments, youbuild it in-house.
What can you do that no oneelse can do At the AI level?
That's a lot about what data doyou have that no one else has?
We all have access to chat, gptand the base large language
models.
That's not unique to my companyor your company.
They're practically giving itaway out there.
(26:56):
That's not unique to my companyor your company.
They're practically giving itaway out there.
But what do you have insideyour walls of your organization
that no one else has?
And how can you leverage AI tomake that more valuable, more
effective?
And that's where I thinkthere's a lot of opportunities
to make some investment to buildit in-house.
Again, you can work withspecialists at any of these
levels.
If you've got an in-house ITteam with spare cycles because I
(27:17):
know we're all sitting aroundbored all day Software guys they
can build any of this at anylevel.
But most of the time, I seeprojects sort of broken down by
these categories.
(27:42):
Lastly, I'll talk just brieflyhere about where to go from here
how to lead, be a leader inyour organization, how to equip
the rest of your team to helpyou on this journey to to be
forward thinking as well.
So, first and foremost, if youdon't have an AI policy within
your team, within yourdepartment, your org, make one
Right.
Tell your team what they canand cannot do, starting with
(28:03):
who's in charge of this thing.
Who do they go to when theyhave a question or want to make
an exception?
And they will want to makeexceptions.
There'll be lots of exceptions,of people trying to ideally,
push the boundaries, find newtools, to use new ways to do
things, but make sure everyoneknows who is the decision maker
(28:23):
on AI policy in yourorganization.
And then, what can they do?
What tools can they use?
Can they use ChatGPT?
Can they use Cloud or Copilot?
Can they use something that'sbaked in their software?
Can they turn on that AI togglein the settings?
What can they do?
You need to evaluate the risksinvolved in that and tell them
(28:46):
what they can use.
I will say if you tell themthey can't use anything, they're
just going to go to ChatGPTbehind your back and start using
those tools right, guarantee it.
So figure out what they'reallowed to use, how they can use
it.
If you don't know, chatgpt willtake your data and use it for
training the next model bydefault.
(29:07):
The default setting on ChatGPTis we're going to use your data
in our next training model, soit is important to have trained
employees, to be trainedyourselves and know how do I use
these tools securely, so theyprotect privacy.
When configured properly, Ibelieve these tools are just as
(29:29):
safe as using Salesforce,hubspot any number of the
software systems we use today.
Right, we treat those, weupload all sorts of documents
into those systems, and I thinkwe can do the same with many AI
tools.
But it's only if they'reconfigured properly and you've
had the training and educationto know how that works.
What data can be uploadedwithin these tools?
(29:51):
Is it anything?
Is it financial data?
Is customer data okay?
What if it's anonymized?
Be specific, provide guidanceto your team.
And then, is it necessary todisclose the AI usage?
Generally, I'd argue no.
Maybe if the AI built thisthing completely with an
automation, maybe you put adisclaimer that AI made that,
(30:13):
but this is one of these rapidlychanging systems, and why you
need to update your AI policyall the time is because people's
expectations around AI usagehas shifted so much over the
last couple of years.
I would expect it's rare in aprofessional setting for me to
see a document that hasn't beentouched by an AI model.
I certainly used it to helpmake my presentations, to write
(30:35):
my proposals, to review all mylegal documents, right.
But you know you decide what'sbest for your business and your
industry.
The important thing is that youcommunicate and again, this
isn't groundbreaking.
This isn't a new way of doingbusiness.
It's the same old business.
It's the same old.
You know change managementtechniques.
Same old communicationtechniques.
It's just a whole new areathat's changing rapidly and we
(30:58):
need to stay on top of it.
You can lead by example.
Be public about your usage, thefact that you're using AI, that
it's okay to use AI.
I think the ideal situation isyou have a whole team of people
who feel empowered and excitedby the opportunities in front of
them with AI.
That's the whole thing we'vetalked about today is that, at a
(31:18):
strategic level, this isentirely new capabilities.
We never could do this before.
So lead, show them it's okay.
Show them it's okay to do thesethings.
Get them trained.
We talked about how that'simportant.
There's a big difference betweensomeone who's using AI and not
trained and someone who istrained.
The prompts, the instructionswe provide AI.
(31:41):
There's specific techniques.
You can use a lot of tips andtricks you'll get from either
personal experimentation or withtraining, so make sure your
people are equipped.
It's the same as anything else.
Right?
You want to have trained,educated people, that's just.
The entire world is having tolearn at the same time how to
(32:02):
use these tools.
So don't leave your peoplebehind.
Get them trained, standardizeand share best practices.
If you find something thatworks well for your organization
, a use case that works well,share that.
Show your own internal casestudies.
The most impact comes when westart to automate processes,
(32:23):
when we build these things intoour core processes.
But that starts by saying hey,I uploaded this survey into
ChatGPT and I was able to findinsight, I was able to condense
results, I was able to dosomething five times faster by
using these tools and when youshare and document these things
with your org, that's how youfind these opportunities and
these use cases across the org,across departments, by
(32:45):
standardizing and sharing.
So, that said, I'll wrap uphere Points.
Main points are AI has unlockednew capabilities.
Main points are AI has unlockednew capabilities.
Never before.
I keep saying it, this is whywe're talking about it.
It is a monumental shift inwhat is possible and what we
(33:08):
could do.
Consider the use cases for eachlevel.
You have your supportingfunctions, your supporting teams
, you have your core processesand teams and then you have that
little nugget of what makes mycompany unique.
So take the time to think abouthow AI can be applied at each
of those levels within yourorganization.
Decide how to develop that usecase.
(33:28):
Does it make sense to build?
What's the ROI impact standard?
How do I prioritize thisproject kind of stuff?
Is this something I'm going towork with a consultant on?
Am I going to try and build itin-house?
What's my capacity for buildingthese projects in-house?
How many can I tackle?
And then, lastly, lead yourorganization into this future.
You know I'm excited about it.
(33:48):
I do this all day, every day.
This is the most fun I've everhad in technology.
Like it's like a kid in a candyshop out here for an engineer
uh, all sorts of groundbreakingnew things that were not
possible, um.
But I think the same applies atthe organizational level.
As you're leading a team,there's all sorts of new
possibilities, new ways of doingthings, and by equipping
(34:12):
yourself to lead into thatfuture, your team will see the
most results.
So thank you guys.
Thanks for tuning in to Riskand Resolve.
See you next time.
Popular Podcasts
Cold Case Files: Miami
Joyce Sapp, 76; Bryan Herrera, 16; and Laurance Webb, 32—three Miami residents whose lives were stolen in brutal, unsolved homicides. Cold Case Files: Miami follows award‑winning radio host and City of Miami Police reserve officer Enrique Santos as he partners with the department’s Cold Case Homicide Unit, determined family members, and the advocates who spend their lives fighting for justice for the victims who can no longer fight for themselves.
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
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.