馃攳
Telco case study - Accelerating AI and advanced analytics with MOSTLY AI's synthetic data - YouTube
Channel: unknown
[0]
[Music]
[6]
one of the biggest benefits of ai
[8]
generated synthetic data
[10]
is that it's so accurate super realistic
[13]
and statistically representative
[15]
not only for static data but also for
[17]
behavioral data
[18]
and for time series data and this is
[21]
exactly the reason
[22]
why it's the only anonymization
[24]
technique out there
[25]
that really produces useful data for ai
[28]
training
[29]
so another use case my colleague andreas
[31]
and i want to talk more about
[33]
is synthetic data for ai training and
[35]
advanced analytics
[36]
one of our clients one of the largest
[39]
global operating telco providers
[41]
had the issue that of course as always
[44]
customers demand personalized services
[47]
and
[48]
excellent customer experience which of
[50]
course requires them
[51]
to utilize data and analyze this data
[54]
but due to gdpr most of the customers
[57]
didn't give content for this kind of
[59]
analytics to happen
[60]
so they only had access to 15 to 20
[63]
percent of the customer database
[65]
which didn't give them the full view on
[66]
the customer data and of course
[69]
hold them back in regards to developing
[71]
exciting new products and services
[73]
that really serve the majority of their
[75]
clients andreas can you tell me a little
[77]
bit more about the client
[78]
and what the problems were there sure
[81]
yeah so
[83]
the challenge each large corporation has
[85]
is of course
[86]
access to highly realistic and
[88]
customer-centric data it's very
[90]
sensitive
[91]
it's protected by gdpr and it's good
[93]
that it's the case
[94]
so speaking from my perspective when i'm
[98]
asked by apps or websites i
[100]
also tend not to give consent to share
[102]
my data and that's exactly the problem
[104]
they face
[105]
clients expect them to have super
[106]
customer-centric products
[108]
but on the other hand they are quite
[110]
reluctant when it comes to sharing
[112]
their data for analysis and therefore
[115]
this makes the life of these
[116]
big corporates quite hard to really
[119]
provide cutting-edge products
[120]
absolutely and i would say that's the
[122]
beauty of synthetic data because
[123]
on the one hand it gives the
[125]
organizations access to
[126]
not the personal data but the insights
[129]
hidden in this data
[130]
and at the other hand it's completely
[132]
privacy friendly and it protects the
[134]
privacy of each and every
[135]
customer and therefore i think it
[137]
combines the best of both worlds
[139]
what exactly did they do with synthetic
[141]
data what
[142]
use did they use it for in the analytics
[144]
space yeah so so they
[146]
of course are looking out for startups
[148]
because big corporates want to have some
[149]
new innovative
[150]
companies to collaborate with to speed
[153]
up the innovation process to bring new
[154]
ideas
[155]
and one startup for example focused on
[157]
them on customer churn
[159]
and since customer data is highly
[162]
sensitive not easy to access
[163]
the startup of course had a challenge
[165]
then to utilize the data
[167]
this big corporation had because it was
[169]
locked away
[170]
so then this big corporate said okay
[172]
there must be a solution how we can
[174]
ensure
[175]
full customer privacy but at the same
[177]
time
[178]
utilize the great data assets we have
[181]
and then the topic synthetic data came
[183]
up and
[184]
also the discussions with us absolutely
[186]
and if i remember correctly
[187]
before they used mostly generate sharing
[189]
data with these external companies with
[192]
these startups
[193]
took them something in between six to
[194]
eight months which is ages
[196]
but with synthetic data they could
[198]
really reduce this time to share super
[200]
realistic granular yet
[201]
privacy safe data to i think a few days
[204]
which is
[205]
amazing yeah it's really amazing and
[207]
it's also then super productive because
[209]
then
[209]
they can run the tests much quicker in a
[211]
higher frequency they can test scenarios
[213]
impossible before
[214]
and also the collaboration with the
[216]
startups in general gets easier it's
[217]
one startup but imagine how many else
[220]
they have in their portfolio and if they
[222]
can then share highly realistic fully
[223]
private
[224]
data for collaboration this is just an
[226]
amazing step forward and
[228]
instead of six months of preparation
[230]
they can get going within
[232]
a few weeks after the paperwork is done
[234]
absolutely so i think it really helped
[235]
them to get innovation into the company
[237]
but of course it was also a benefit for
[239]
the smart people within the organization
[241]
because
[241]
also the internal data requests are
[243]
something that takes quite some time at
[245]
a large corporate
[246]
and also these people now had access to
[248]
realistic data to develop some models
[250]
in-house
[251]
so super nice use case but they didn't
[253]
stop at advanced analytics and ai
[255]
training
[255]
what else did this telco provided yeah
[257]
so so one interesting use case was
[259]
around mobility data
[261]
so these telco companies of course have
[263]
a lot of clients and
[265]
it can be interesting for other
[266]
companies like transportation services
[268]
to
[269]
also analyze the movement of people and
[271]
so they want to ensure full privacy but
[273]
at the same time also support
[275]
other companies to help people to
[277]
commute in a more efficient way to work
[278]
for example
[279]
or to help them to analyze when for
[281]
example if someone wants to open a shop
[283]
somewhere what is the frequency there on
[284]
this main square
[286]
and so on so there are plenty of further
[288]
use cases to be
[289]
explored um yeah absolutely i would say
[291]
location date is really one of our
[293]
most valuable assets when we look in the
[295]
telco industry and it's not really about
[297]
the movement of individuals i think
[299]
what's really interesting are the
[300]
patterns of larger groups of people so
[302]
also if you look into the public sector
[304]
their people are interested in building
[306]
smart cities optimizing
[308]
traffic routes and so on and so forth
[310]
and i think for this telco provider it
[311]
will be moving forward a quite nice
[313]
additional revenue stream
[315]
but i thought they also had another use
[316]
case when it came to data retention
[318]
where they looked into synthetic data
[320]
what was the consideration there and why
[322]
were they interested in the technology
[324]
yeah so so one challenge they had
[325]
previously was when they
[327]
had this lengthy approval process over
[330]
many months and they finally get access
[332]
to this data and can
[333]
work with it that based on some
[335]
regulations they have to delete this
[337]
data again after a few months and then
[339]
the fun starts all over again right so
[341]
in this case here by having this
[343]
this data now retains they can build up
[345]
a long history of this data and then
[347]
create as well test and then run
[349]
scenarios
[349]
not possible before and this also helps
[352]
them in their
[353]
analytics department to better
[354]
understand patterns over time
[356]
absolutely so i think data retention is
[358]
is another of the big use cases of
[360]
synthetic data
[361]
because you can definitely preserve all
[363]
the insights for the years to come
[365]
without infringing or risking that that
[368]
the privacy of your customers is leaked
[370]
so also i think another client where we
[372]
really see that synthetic data started
[374]
at one place
[374]
and then is moving to other departments
[376]
and hopefully much more
[377]
moving forward but we will see that
[393]
[Music]
Most Recent Videos:
You can go back to the homepage right here: Homepage