Artificial Intelligence in Modern Marketing (AI)

hello
i'm brian cardin i'm the chief marketing
officer
at envision and it's great to be with
you all
and uh very excited about our topic
today
which is uh all about ai and marketing
and uh i'm a marketing practitioner for
a long time
and uh we've of course all seen how ai
has touched so many things in our lives
and today we're going to be talking
about specific examples of how
artificial intelligence is being used in
modern marketing organizations
first a little bit about me i
began my career at the bottom of this
chart after business school i
joined a strategy consulting firm called
braxton which is part of deloitte
became a partner there in their consumer
marketing and strategy practice my
clients were
companies like heinz and campbell soup
and
ralph lauren and merc and uh
that was a great uh exciting time where
we started to see
um lots of uh digital transformation
going on
um and from there uh being a a
partner i uh we had uh our first
children were born twins
and uh the life of a consulting partner
is traveling all the time so i was
thrilled
that i got a job offer to be a cmo at
reed elsevier about a 5 billion
british dutch company they own things
like
lexisnexis the large online database
elsevier science which are subscription
scientific journals
trade shows a whole bunch of things i
was there for about six years
i joined forester research as head of
marketing and head of strategy
my first part of my career uh
in software was with eloqua marketing
automation and that was very exciting it
was a small company
when i joined about 15 million dollars a
year
we got up to about 100 million dollars a
year in about four years we took it
public and was acquired by oracle
lattice engines is actually a company
that does
artificial intelligence for sales and
marketing and i was there for four years
it was subsequently bought by donna
bradstreet to really marry the data
heritage and legacy and assets have done
a bradstreet with
predictive analytics and artificial
intelligence
and i've been at envision now for a
little over a year
um just a little bit about what envision
is we're a
uh digital product design platform it
allows
uh companies around the world to build
digital experiences for the customers we
have about seven million people around
the world
using our platform uh it includes a
hundred percent of the fortune 100
so it's very large companies but also
mid-size small
some of our accounts that we work with
are companies that were born digital
uh like uber and netflix and others
uh there was a really compelling reason
to become digital
uh companies uh like goldman sachs or
bank of america
one of the uh very interesting parts
about our company is we were founded
about 10 years ago
and we've always been a fully
distributed company we don't call it
remote we call it fully distributed
and we have zero offices we have uh
over 600 employees and we've never had
offices anywhere in the world and so
part of the idea was we help our
customers build digital experiences
and we wanted to see what it's like to
run a company
through digital experiences and not
face-to-face experiences
and we have a lot of resources on our
website that if you're thinking about
in a covered world about being fully
distributed forever
or a hybrid model there's lots of things
on our on our website that can help
help you through some of those issues as
i mentioned
we have 100 of the fortune 100 and lots
of different
kinds of customers
customers like disney use us for all the
applications for like fast track and
fast lane and uh in-park experiences
um they also uh you know use it for
um for different kinds of uh of
uh in animation customer experiences
um in the case of some of these service
providers like wpp or pwc
they use our platform on behalf of some
of their customers to build digital
experiences
for their customers
and as a platform that
allows people to build digital
experience we have the design piece
where you can design
experience you can share that prototype
with other people
then you can send it over to your
front-end developers who actually write
code and build it out
and you can scale it across the
organization and get other people to
comment
and talk about it and eventually ship
that digital experience
so let's talk a little bit about ai i
very
distinctly remember this moment do you
remember the name of the player in the
middle
that of course is watson um who
beat the player on the left ken jennings
who i believe to this day is still the
greatest human player
but uh for those of you who love
jeopardy it's a game that is very hard
to program it's not as simple as
searching for uh you know very objective
information
like how many humans are on the earth or
um
you know or or uh answers to equations
or specific answers it's a very
challenging
problem that ibm through watson has been
able to solve when i went to business
school
a lot of my colleagues wanted to be
traders and now we see that as a career
that human beings don't do but
[Music]
bots do so the algorithmic trading that
happens now the algorithms don't just
tell
what trades to make they actually
execute the trades as well
and so the human beings are very much
employed in this process but they're
writing the algorithms
and they're writing the code and using
artificial intelligence to
place optimal trades
has anyone ever been in an autonomous
car uh these are a lot of fun
to be in can be very frightening as well
i've i've been in autonomous car a
couple times all
in san francisco and it's just uh
really fun we went on the the 405 so i
remember
leaving google's offices and getting on
a ramp and getting on
the freeway there and the car performed
flawlessly
curiously when they brought the car to
the east coast i live in boston
and they brought it to the east coast um
in january
and uh despite having hundreds of
sensors and cameras
the car did not know what snow was so
when it came to a little wall of snow a
couple of inches the car would stop
assuming that it was uh not a material
that you could
drive over and crush uh so the car had
to learn about different weather
conditions
which i find really interesting all of
our lives have digital assistants alexa
and google and siri that are parts of
our lives naturally
one of my favorite examples is there's a
sundar from google
and he wants to make a uh a haircut
appointment
and it's not simply a matter of asking
his
digital assistant to put it on his
calendar but the digital assistant
actually
finds the hair salon goes through his
contacts finds a hair salon that he
always goes to
finds an available time places the call
and uses a voice to actually
interact with a human being at the hair
salon and book an appointment put it on
his calendar
and confirm it it's quite striking
there's the url if you want to watch the
youtube but it's a great example of
humans and ai working together
in a very effective way and we of course
we live in a world powered by a.i
in all sorts of areas like healthcare
insurance credit cards retail
it just seems to permeate almost every
field
today we're going to talk about
marketing and sales
so let me give you a couple of examples
that
we believed five or ten years ago were
going to be bright spots for artificial
intelligence and marketing
the first area is smart lead routing
so lead routing is when you have a lead
and the idea was i'll route it to the
rep
who has demonstrated the highest ability
to close
a lead like that so let's say it's a
health care
company that is a lead and you know that
nancy
has the highest close rate in the health
care sector
so the leads would be routed to the rep
there
except the second example would be
content creation
and uh can artificial intelligence
identify
the kinds of content that you should be
creating on your website
to drive engagement what are the topics
of most interest
and the third area that held a lot of
promise was predictive email
as you might expect we have lots and
lots of history of when people
open email when they click through is
there an
optimal day and time of day
that you send email not to everyone but
to a particular person based on their
history
can you build artificial intelligence
that will yield
higher open rates based on when people
historically have opened emails
and so in all three cases they have
failed
so i'm gonna in a couple of minutes give
you some examples of things that have
succeeded and these have failed for a
number of reasons
in most cases it's lack of enough data
to
build a model that is robust enough
that it would yield higher results that
would happen
happen randomly so these three examples
while
people believed eight or ten years ago
were very promising have not proven
to be very helpful
so let me talk a little bit about what a
lot of marketers say they say that we're
now
modern marketers i'm particularly
talking about b2b business to business
marketers
but the data would suggest otherwise
because the results are quite poor
so what's wrong here mql stands for
marketing qualified leads
so the marketing team spends a lot of
time and gets a lead someone comes to
the website
engages with some content they pass that
lead over to
a bdr a business development rep or a
sales development rep in sdr
but in fact from all the data across
thousands of companies we see that
94 of all quote marketing qualified
leads
will never close and so if i was on the
receiving end of those leads
i probably would not follow up very well
because
only six percent are going to close that
doesn't seem like a very healthy rate
and that rate has been pretty steady
over the last eight years
so we don't see a lot of proven despite
all these marketers being modern
marketers
the second area in a related function is
sales
is that according to the data from
gartner and serious decisions and
foresters 52
of sales reps will not make quota so
here we are i do a lot of planning
and yet barely half of your reps hit the
number that they're assigned
huge failure so both the mqls are not
converting at a healthy rate most will
never close
and just about half the reps are not
hitting quota
so i'm going to suggest today that
there's a better way to do things
sort of next gen marketing this era of
modern marketing happened because of
marketing automation and scaling
things like eloqua where i worked or
marketo
but there were new marketing tools that
are marketers to automate and scale
and there are additional tools now with
ai that allows marketers to move
from poor performance to extraordinary
performance
i'm gonna go into some detail about
these
um william gibson the writer uh
said famously that the future is already
here it's just not evenly distributed
the point being that um we're seeing
some organizations applying ai very
effectively
in marketing and other organizations
either not applying it or
spending a lot of time and cycles and
effort and money
and it's had disastrous results
what we're seeing is that if you
are very clear about what job you want
to get done
there are very specific ai powered tools
that will allow you to get there
let's talk about them so the three
examples today will be
first identifying target accounts
most marketing sales organizations
have a huge universe of potential target
accounts that they can go after
what if you could apply ai to identify
the accounts most likely to buy from you
so it gets to use ai to disqualify
tens of thousands of accounts you should
never call because they have a low
conversion rate
so how would you do that the second
example
is engaging with prospects at scale
so if you have lots of leads coming in
and you only have let's say
20 or 30 humans to pick up the phone and
follow up
you may not be able to follow up in a
timely manner so can you use bots
just like sundar used a bot to make that
haircut appointment can use a bot
to engage with a prospect qualify
disqualify and do a lot of the work that
normally a human being would do
and the real advantage here is speed uh
of time uh very often with leads they
don't come in an even way but there
could be a spike of leads
let's say you run a big webinar and you
get 10 000 leads one day
and you have a modest sized team they
can't work through the leads in a timely
manner
and we know that the decay rate is quite
fast and steep
for leads if you don't respond right
away and a third area is improving
conversion rates
um one area of marketing that's been
particularly impressive
is believe it or not direct mail so
everyone sort of leaned over to digital
communications but we're seeing that if
you mail something to a prospect
you have a much higher uh chance of
having a meeting with them
that they feel somehow beholden to you
because you sent them a gift and so i
want to talk about ai gef gift giving is
there a way to scale that
rather than have all these sales reps
trying to come up with a gift
that's either personalized or you just
send everybody the same gift which would
have a low conversion rate how do you
personalize a gift at scale
let's talk about these the first
question is which accounts are most
likely to buy
from a given company and so this is the
target account question
and so uh what we've done is
we've looked at um we've done a back
test we've looked at two years of
account history
to identify accounts uh where we've won
and lost
and then we've appended a whole bunch of
additional attributes to those accounts
to build out a very
robust data set and so
basically it's a multiple regression we
run a model to identify the most
predictive attributes
and then we acquire accounts that
have those attributes so for example
if we find here that the tenure of the
cio
one thing we found in this example is
that during the first six months of a
new
cio you'll have much higher chance of
having a meeting with them
and a much higher conversion rate once a
cio has been in their job for more than
six months
they're sort of locked and loaded and
they're much less open to talking to new
vendors
so we would append that information to
the data set
and we would look for other accounts
like that so in this particular case
we found out that there was an ideal
company size companies that were smaller
or larger had a low conversion rate
certain industries
the existence of certain technology was
predictive so you may ask the question
well how could i possibly know
what technologies companies are using
there's lots of good ways to do that one
is
scraping through their job descriptions
so if you scrape through a lot of the
job descriptions you'll see that
what technology they're using in this
particular case we found that multiple
locations
was much more predictive than a company
that had one location global presence
and some other on-premise systems we
identified 20 almost 22
000 accounts we assigned accounts to
rsms regional sales managers
then we got contacts at those 22
000 accounts and we started running ads
across all of those accounts and across
all of those contacts
what's important is it's only 22 000
accounts
in this particular case they started
with an initial list of over two hundred
thousand accounts
so they paired it way back less than ten
percent of the accounts
that allows them to have much greater
presence and more marketing there
the bdrs uh a higher rate of calling
target accounts
it began when this program began only 59
and then it went up to 84
and this are these are maps of where the
target accounts are located
one of the great benefits of
understanding where your
most likely accounts are located is you
can
put local events in those regions for
example if i got a call from a rep in
nashville that would say
brian can you run a dinner or a special
program here in nashville
i would look at these data and i'd say i
don't have enough target accounts in
nashville
why don't we do it in another city
let's do it in houston or let's do it in
new orleans or let's do it in miami
that is a much greater density of target
accounts
and so what were the results of using
artificial intelligence to identify
target accounts
well the first area is cycle time
the non-target accounts were closing in
205 days the target accounts in 140 days
a dramatic difference in uh in cycle
time
the deal sizes were also quite a bit
larger
and the win rate from opportunity to
closed one
was up quite a bit and so these are very
dramatic results
and the whole idea was to focus on fewer
accounts that we knew based on the back
test of looking at historical data
were much more likely to close and so
the win rate we could have expected
the deal size is something that was a
wonderful side benefit
and the the sales cycle time was was
fairly predictable
so that was uh target accounts using
predictive analytics
let's talk about the second application
of ai and marketing engaging with
prospects
at scale
sometimes marketers and sales people
have uh
the curse of abundance and what that
means is they just have too many darn
leads
and how do you decide which leads to
follow up with
the example that i gave earlier is very
relevant
let's say you go to a trade show and you
get several thousand leads
we know that responding to a lead
properly
really matters to conversion rates and
very often human beings can't respond
quite as quickly
as as artificial means
so our bdr team the business development
rep this team
is responsible for calling the leads
following up on the leads
the first challenge is lead surge so if
you have a big spike a surge of leads
how do they follow up with very promptly
the consistency
so the bdrs are human beings and we
wanted them to have a cadence of seven
touches two calls
and five emails but we had no way to
guarantee that
and we wanted fast follow-up
so here's some data that talks about um
uh dials and uh
and response and you can see that if you
respond within the first five minutes
of someone indicating interest you have
a 10 times greater
chances of engaging than after
10 minutes and you can see even after 30
minutes it falls
and so how do you do this very quickly
how do you respond to leads very quickly
human beings can't do that necessarily
and so can you automate this in some way
and so we decided to use a bot and
her name is natalie we called her
natalie because
our best bdrs uh at
at our company are women and so we
wanted to use a female name
and so natalie uh really never needs
time off
natalie doesn't take vacations no sick
days uh natalie also can respond at two
in the morning
or uh eleven o'clock at night and uh
natalie responds in a very consistent
way
and so um the bdrs were telling us they
can't always respond as fast as they
like to a prospect
but a bot can and we've seen that bots
uh
the intelligence you can program into
the kinds of conversations you have
um really are as good as human
interactions in email
now natalie cannot do phone calls just
yet
only humans do that we're working on
that but natalie is extremely effective
at sending out
very effective emails personalized
emails and even
using things like calendly to schedule
appointments with a rep
so very high conversion rates and great
engagement rates
from natalie the bot
the third area i want to talk about is
gift giving
which we've seen to be quite effective
at converting
prospects so trying to get a meeting
with a c-level person is extremely hard
these days
we have all sorts of uh you know
blockers on our
on our phones we have ways to
um put into junk mail email from people
we've never heard from before
so email has extremely low conversion
rates cold calling is extremely
difficult
one area that we've seen work pretty
well is direct mail we've seen
all of us have probably received a
bottle of wine or some sort of generic
gift with a note
what we've seen is that if the gift is
highly personalized
the conversion rate and the receptivity
of the recipient to engage with the
sales rep is very high
so when i learned that personalized
gifts
when we sent them out were very
effective i asked our bbr team
to start going into facebook to start um
looking at um prospects profiles to try
to come up with a personalized gift
now the downside was i remember a week
after i talked to the bdr team about
doing this i walked over to where the
bdrs were sitting
and they're all on facebook all day like
writing down things they saw in pictures
and profiles and this person went to
notre dame sent them a notre dame
you know hat and a sweatshirt so they're
trying to personalize gifts
so how do you do this at scale and there
are tools to do this
i'm going to talk about one
[Music]
this is one where if you
give this tool the tool's name is alice
alyce
if you give alice a um a business email
address
it will match that email address with
the social profiles of that person so if
you give it my email address there's me
brian cardin at envision app
it'll find my facebook instagram twitter
linkedin accounts
it will then crawl through my profiles
and using its logic its ai it will make
specific personalized gift
recommendations
and they will actually send out the gift
or they'll send out a card
that says a gift has been chosen for you
click here
with a perl a personalized url
so here's what really happens is the bot
goes out and this is an actual picture
on my facebook account
and it sees that i have a new dog
and it sees it rated this picture very
high because the number of likes that it
had
and also that i'm with my two children
my sons and so the bot found images that
would suggest that maybe there would be
a gift related to a new dog
and also found this photo i'm there with
my son and my wife
and we're actually at an opera house and
the software was smart enough to know
that it was an opera house not a movie
theater based
on those balconies based on the
architecture so it knew that this was an
opera house in prague in the czech
republic
and so it recommended two things that
were personalized one is a bark box
which is a monthly uh box of uh of
little treats
and and little toys for your dog
and the other was a coffee table book on
the most beautiful opera houses of the
world
so highly personalized gifts and it was
just wonderful and as
i said alice this software actually will
send out the gift
automatically or at least a card that
says the gift has been chosen for you
and the conversion rates were quite high
um
so in our world nbm means new business
meetings and that's a stage in our
process
and we did a proof of concept where we
had good conversion rates
and then we tweaked along the way a
personalized note and some other things
we had amazing results so um
we sent about uh 2 000 gifts in this
test
and uh we closed eight deals
and uh the average deal size the asp was
250 000
a year so about two million dollars in
new bookings and the cost of this whole
program was only twenty five thousand
dollars
so personalized gifts at scale if you
try to send out two thousand two
thousand personalized gifts
and you let human beings pick out the
gifts it could take thousands of hours
for people to find all those gifts and
personalize it so this was extremely
effective
so today we talked about three
applications of ai and marketing the
first
is target accounts using predictive
analytics the second
is following up leads with a bot natalie
and a third is choosing gifts at scale
for a direct mail using ai so what's
been the cumulative result
i've masked a little bit of this data in
the vertical axis
but basically the sales rep productivity
gains over
this period that we applied all three
they've more than doubled
so the mrr was his monthly recurring
revenue goal
was eight hundred thousand dollars per
rep uh
the quote actually went up to 1.6
million it doubled
based on the value of all of these tools
i want to sort of conclude with some
thoughts about ai and marketing and
what we're seeing across a range of
organizations
and i think one of the conclusions uh
as william gibson said is that if you
apply it the right way you can have
extremely beneficial results but also
for a lot of organizations
has been a time sync so you have to do
it in a way where you know precisely
what you're trying to get done
but i think ai is becoming something
that we've seen through these decades
i'll give you a few examples
if you are old enough to remember the
1990s every company in the world
wanted to be a quote.com company and of
course every company is now
and so the word dot com sort of goes
away and every company is a digital
company in some way
even companies making physical things
think about ge making big engines
they have sensors on their engines
they have ways to track performance
you know companies making drugs they
become digital companies as well as
people order prescriptions
uh in the 2000s every company wanted to
be a cloud company
and so people were bragging that they
were cloud first they're a cloud company
and now that's pretty much gone away
because every software company is a
cloud company
the third area and i remember this very
clearly was the era of big data so
obviously
with social platforms and email and
and sort of back-end tools we're seeing
the era of big data
billions of bits of data are now coming
out and
every company was going to harness the
power of big data but big data is no
longer something special just like
com cloud and big data it's become in
the background and it's assumed
i think the same is true of artificial
intelligence that just like every
company became
a dot-com company and every company is
putting their technology in the cloud
and
every company is using big data every
company in some way is an
ai company as well
so i want to uh thank you all for
being part of this presentation and uh
and i hope to meet you soon in person
once kovit's gone
thank you very much