Generative AI for Real-World Business Productivity

foreign
I'm Chris Howard and welcome to top of
mind I'm thinking of this episode kind
of wrapping up our conversation about
generative AI now clearly that topic
will continue and it'll come up again
and so on but I wanted to round this out
today by thinking about what makes
generative AI productive going forward
now you may remember that I talked about
generative AI in the trough of
disillusionment and some people thought
wow that's really soon that's very early
well the truth is that this technology
is moving through the hype cycle faster
than any technology that I that I can
recall uh and the traffic
disillusionment simply means that this
is where the work is beginning where the
hard work starts and eventually what
happens when a technology comes up out
of the trough is you hit the plateau of
productivity by becoming more
enlightened about what works and what
doesn't and so on so I wanted to take a
few minutes today and just talk about
what we're seeing and really
articulating the productivity of
generative AI through use cases like as
you think about use cases are where
things get real or you're applying a
technology and you're achieving some
kinds of results from that
so let's start with an example it's a
healthcare example imagine that you were
a CIO at a hospital and what's happening
of course is this trend of
consumerization of Health Care and
access to knowledge and democratized
access to knowledge and this is
happening both for patients as well as
as caregivers Jeff cribs who's on my
Healthcare and Life Sciences team wrote
a document studying the impact of gpt4
on Healthcare and life sciences and he
chose gpt4 because it's more capable it
actually is has a deeper sense of logic
and zero shot learning and all those
types of things that differentiate for
from 3.5
so your CIO at hospital Jeff is writing
this for you so there's this
consumerization and access to
information but what's happening is
there are different perceptions of gpt4
that it makes it appear to be maybe more
empathetic or more correct perhaps and
certainly more accessible you think
about if you're waiting to get feedback
from your healthcare provider it can
take several days sometimes and so
people are using gpt4 to get better
answers or at least what they perceive
to be better answers it's also happening
with the practitioners themselves
they're using gpt4 for answers and
knowledge search and so on so clearly
you want to get some control over this
highly regulated environment there are a
lot of privacy issues and so on and so
Jeff walks through those opportunities
and risks for the practitioners for The
Regulators for the patients or everybody
that's touched by Health Care in the
hospital and then starts to say well
here are some things that you could do
the other thing that's making this even
more complex is that the vendors of
Health Care
applications like epic and others of
course are moving in the same direction
as building these capabilities into
their systems so think of what you might
be able to do with that maybe it's a
patient doctor consultation and that's
being recorded uh of like audio recorded
and then the gbt4 will create a summary
of that it's actually really interesting
and beneficial because that's one of the
things where errors tend to occur maybe
it happens after the visit is done maybe
it doesn't happen at all and so the use
of a technology that's sort of in an
ambient listening situation actually
creating a better record of the
interaction and then if you have
multiples of those interactions you can
start to see patterns across them even
for an individual patient or for
multiple patients with similar symptoms
that kind of thing and the way that Jeff
lays this out does to say okay well here
are the couple of design patterns you
would use to make that work so prompt
engineering some access to internal and
protected data the use of GPT Force the
conversational piece of this but then
building and embedding this all into an
application that the hospital in this
case would control the other thing
that's happening that's interesting in
healthcare space is the introduction of
domain specific models so models for
example that Mayo Clinic clinic is
experimenting with now that's really
specifically tuned to the healthcare
environment and the specific language
and patterns and parameters that
surround us so they're testing that now
and I think what this is is a good
representation of how an industry is
disrupted the CIO in their response to
it but also the emerging space within
large language models itself I think
very quickly what's going to happen here
is that you'll see the rise of these
very productive smaller large language
models that are that are domain specific
that look at a specific space in this
case the healthcare space but emerging
across all Industries and say two years
from now we're in a situation where you
have multiple models running
orchestration and integration across
them in the pursuit of business goals
but right now very beginning stages the
pilots that people are experimenting
with now are giving them some sense of
accuracy or lack thereof and how it
might be useful in a larger setting
so that example was a CIO example the
other thing of course that we're doing
at Gartner is looking at use cases
across all of the roles that we serve
and this is a place where you can help
me uh we're out collecting use cases
Pilots what are people doing with this
technology and ultimately going to
create that picture through research and
and show what's productive what's more
feasible what produces results perhaps
but is harder to do and so over the next
few months we're publishing a whole set
of these kinds of use cases if you're
doing something interesting with
generative Ai and it could be text or
not text would be anything I'd love you
to put that in the comments here and uh
and so we can see what you're doing and
also remember questions in the in the
chat as well because I do Circle back
around those and sometimes answer them
in line but we plan to do a couple of
episodes where it's really just
answering your questions
so back to roles what about roles like
HR let's say you want to maybe create a
summary of a talent review not that it
would be the final version but if you
actually want to go through all of the
data that you maybe have on an employee
and to create the draft of a performance
review that's an interesting use case a
lot of HR departments also do Talent
intelligence like out in the market what
are competitors doing what other job
postings look like that kind of thing
that's a great use case for for gdt or
for for large language models in general
what about in legal legal of course is
looking at all the regulations that
affect the company and its operations so
using large language models is a great
way to maybe get summaries of
regulations and looking at how things
change or how they're different from one
place to another or even just generating
policy documentation or summarizing
policy documentation
uh marketing marketing has been a heavy
user of generative AI for for a while
now prior to all of this chat GPT stuff
that happened at the end of 22 they're
using it to generate ad copy you know
again summarization of activities maybe
even things like emotion sensing in
detection
in sales sales is constantly generating
material sales materials maybe it's
pitch decks or it's you know calls or
post call follow-up again great use
cases for this kind of Technology
what about customer service customer
service we're talking a lot about the
augmented service rep so this is a
person that's Fielding incoming calls uh
usually you have multiple screens that
they're working with now imagine
generative AI being popped in there to
create easy access to knowledge to help
solve the hard problems that come in to
a call center
Finance anomaly detection is a great use
case there the thing that we're watching
in finance is also conversational
reporting so using conversational
interfaces to context reporting and then
letting the engines work behind the
scenes to actually bring that to you
something about supply chain so it's one
of the challenges with supply chain is
keeping workers in place the turnover
rate tends to be very high and so if you
can create for them a better digital
experience through access to knowledge
maybe working in a warehouse you know
and and maybe even combining things like
mixed reality together with generative
AI
creates an environment that is really
knowledge rich but also better to work
in from a design point of view so you
can see their use case I've just given
you the tip of the iceberg we're
collecting hundreds of these from across
all Industries to see who's impacted I
gave you a healthcare example and that's
one that's easier for us to understand
because we you know we interact with
Healthcare but I can find use cases
across multiple Industries where you
might not expect them so for example
mining sort of heavy industry and money
you think well what's what's you know
the what's the use there turns out that
you can actually ingest Blueprints and
then create interesting designs for
extending a mine or creating new ones or
making them more efficient or actually
making them perhaps even more
environmentally friendly to the extent
that you could do that and so whatever
industry you're in I want you to be
thinking creatively about what you can
do and share that here because we've got
a lot of people watching and it would be
great to sort of foster some
communication about that
let me give you a couple of concrete
examples just a couple to show that
people are really doing this I was on a
board meeting with a very large grocery
company so think about packaged
groceries and they are using generative
AI to help employees understand the
employee policies like travel policies
or maybe data retention policies all
those kinds of things which tend to be
spread out across multiple documents and
standing operating procedures they've
built an interface into that so you can
ask very simple questions like what's my
per diem if I'm going to New York City
next week things like that
uh there's a large financial institution
that I work with that's using this to
experiment with a code modernization
which is something I haven't talked a
lot about but generative AI for code
generation is one of the really great
productivity use cases the numbers that
we're seeing for productivity
exceptionally high now they're using
this to test movement from Cobalt 2 to
Cobalt 6 or even out of cobal and into
more modern languages like Python and
things like that so they're
experimenting with this and checking
again accuracy and so on one thing I'll
say though about code generation is that
it increases the amount of time that you
need to do testing the output and when
it comes to modernization which is
something that Gartner is going to be
studying really closely over the next
couple of months it translates the code
but it doesn't at this point it doesn't
actually restructure it or modernize the
code itself it changes into the
languages and so on so one of the bank
system is doing this
so generative AI moving into the trough
of disillusionment Simply means that
this is where the hard work starts and
we'll find stuff that works well and
stuff that doesn't work well and we just
we work through that and what happens
ultimately is the generative AI takes
its place alongside the other
capabilities of artificial intelligence
that have been around for a decade or
more
so you may remember the first episode
that I recorded I told a story about the
first automobiles in Yarmouth Nova
Scotia and that was on my mom's side of
the family I remember my father telling
me about early car trips though so this
would have been in the 40s the late 40s
maybe the early 50s and how the roads
were so bad that you could be sure that
you would blow a tire on a trip even of
just a short distance of maybe 40 or 50
miles roads weren't great the tires
weren't sort of industrial strength yet
and so you always set spares and you
would blow your tires that you could be
short of it
actually turned people off of
automobiles it said no this is never
going to work so you know let's let's do
something else it actually encouraged a
solution to those problems so roads got
better tires got better that kind of
thing to the point now where we have
ride flat tires and so even still what
we're progressing from that that's the
nature of change in technology and
learning how to use something and to
make it productive
and so again share your stories what are
you using gen AI for to become
productive share those stories here
share your questions and welcome back to
them
this has been top of mind thanks for
joining and we'll see you again in a
couple of weeks
foreign