馃攳
ESG, followed by AMA - YouTube
Channel: unknown
[0]
hi everyone thank you very much for your
[3]
time today
[4]
my name is antoine amand and i'm the
[6]
technical director for financial
[8]
services
[9]
at databricks i'm joining from a
[13]
financial services background having
[15]
breached the gap between business and
[16]
technology
[18]
and today i want to show you the
[19]
solution accelerator program that we
[21]
built
[21]
at databricks to show you the technical
[24]
capability of the platform within a use
[26]
case you may be familiar with
[28]
and to give our clients and our
[29]
customers a head start on those
[31]
use cases in that context environmental
[34]
social and governance
[36]
is a top level priority for all
[38]
financial services customers
[40]
retail customers or any large
[42]
organization may have a esg strategy
[45]
and how to migrate from a marketing
[47]
concept of esg
[48]
to a data driven and actionable insights
[52]
through the series of notebooks i want
[54]
to quickly show you
[55]
how your data practitioners within your
[57]
organization
[59]
will be able to leverage ai advanced
[61]
analytics to extract
[63]
key information from corporate social
[65]
and governance
[66]
uh csr reports using
[70]
advanced analytics and how to correlate
[73]
those esg
[74]
initiatives that were disclosed with esg
[77]
sentiment
[78]
as the way your brand your organization
[80]
or your suppliers
[82]
may be perceived through the news
[83]
analytics and bridging the gap between
[86]
what was disclosed versus what was
[88]
perceived to bring a true
[90]
data-driven esg strategy
[94]
in this first example then i want to
[96]
start
[97]
using information coming from
[100]
uh from my financial services customers
[102]
or different
[104]
clients using all those pdf documents
[106]
that i could find
[107]
in the web a document itself is usually
[110]
released on a quarterly or yearly basis
[113]
and contains about hundreds pages long
[115]
is a hundred pages long document that
[117]
contains
[118]
few pages around hard metrics and a lot
[121]
a lot of different text
[122]
where we want to effectively start
[125]
extracting this text content
[127]
in an actionable way using advanced
[129]
analytics
[130]
scraping this content being able to
[132]
extract each and every single sentence
[135]
and using machine learning to really
[138]
programmatically learn those
[140]
themes those topics so naturally in the
[143]
context of esg
[144]
you will find some themes that when
[146]
machine learned around supporting
[148]
community
[148]
valuing employee reducing carbon
[151]
emission
[153]
investing in a more sustainable finance
[155]
helping employees
[156]
and risk management that is intertwined
[159]
with
[160]
esg that will help us as well
[163]
starting to categorize each of those key
[166]
statements in pdf documents
[169]
drastically summarizing a document to
[172]
its core esg initiatives as you can see
[176]
without getting into the nitty-gritty of
[178]
the map behind
[179]
we can summarize a document that
[181]
contains only 20
[183]
of a document may be actionable maybe
[186]
specific enough
[187]
towards those nine themes that we've
[190]
we've been able to machine learn
[191]
and how you can then start comparing
[194]
your different suppliers your different
[196]
investments your different
[198]
competitors against those nine key
[200]
metrics such as
[202]
valuing employees sustainable finance
[204]
code of conduct
[206]
how much more this company is valuing
[207]
employee compared to others
[209]
to bring that holistic view of your
[212]
different
[213]
csr reports and help you to drastically
[216]
and programmatically then
[218]
summarize a complex 170 pages long
[221]
document
[222]
into nine key initiatives those were the
[225]
key initiatives from
[226]
j.p morgan in this specific example
[230]
but you don't necessarily need to be a
[232]
python expert or a spark expert
[235]
to start interacting with your model
[238]
with our model
[239]
stored therefore on ml flow you can
[241]
start creating
[243]
this simple analytics
[246]
using a simple right click here i create
[248]
a chrome plugin that will
[250]
programmatically extract the key esg
[252]
initiative for a specific pdf document
[256]
scraping the the content and scoring
[258]
that specific csr
[260]
against my 50 financial service
[262]
institution
[263]
i've trained a model for this example
[266]
will show me that
[268]
barclays will be in the top 40 percent
[270]
for everything related to supporting
[272]
community
[273]
this is a simple example how to use that
[276]
model
[277]
programmatically or to the use of sql
[279]
for instance
[281]
the second aspect is consolidating and
[284]
correlating that with news analytics
[286]
data
[287]
now that we understand what was
[289]
disclosed let's look at
[291]
what was perceived
[296]
we use news analytics data to scrape the
[299]
content for the last 18 months of
[301]
history
[301]
and as you can see the data is massive
[304]
and alternative data is publicly
[306]
available
[307]
but no easy and massive but available
[309]
every 15 minutes
[310]
if we can drive some insights out of
[312]
that 15 minutes time window
[314]
then we'll create a real-time view of an
[317]
esg rating
[318]
no longer waiting for a yearly
[320]
disclosure but can operate on a
[322]
real-time
[323]
minutes window so we'll be extracting
[326]
the
[326]
environmental social and governance
[328]
related themes that we can extract
[330]
together with the sentiment analysis
[332]
to give this data-driven view to give
[334]
that sentiment analysis for each and
[336]
every single business mentioned in the
[338]
news
[339]
not necessarily related to the four the
[342]
few handful companies we have a
[344]
csr report for but really to each and
[347]
every single
[348]
business medium or small companies
[350]
financial services or retail customers
[352]
to build that data-driven view and to
[355]
capture the influence
[356]
any business may have to your esg rating
[359]
if you are directly or indirectly
[361]
related to a badly rated esg company
[364]
the framework being data driven will
[366]
quantify will capture will propagate and
[369]
will
[369]
show you that positive or negative
[371]
influence to your brand
[373]
to your suppliers to your investments
[376]
how do you act then on this esg
[380]
rating is really endless opportunities
[383]
so you want to look at this from a
[384]
supply chain perspective you want to
[386]
look at this from
[388]
a market risk perspective in this
[390]
example i create a simple synthetic
[392]
portfolio made of 40 or 15 equities
[396]
i simply download historical data of
[398]
their stock returns
[400]
and i overlay with my data-driven esg
[402]
score
[404]
interestingly the best and the worst
[406]
core in my framework correspond to the
[408]
best and the worst performance in my
[410]
portfolio
[411]
i apply this with more data to
[414]
understand
[415]
more than just two data points and yes
[417]
esg
[418]
is directly correlated with market
[420]
volatility badly esg related
[423]
have in my example a risk that is two
[425]
times higher
[426]
leading paving the way towards a more
[428]
agile view of
[429]
risk management such as what is your
[432]
risk exposure to the e to the
[433]
s to the g what action should you take
[436]
to reduce that risk
[438]
how do you effectively operate in a more
[441]
operational resiliency by applying esg
[444]
framework this data-driven view of just
[445]
your suppliers
[447]
across your investments and how do you
[449]
package all that information
[451]
in a way that can be consumed by your
[453]
lines of business
[454]
through the use of dashboards to the use
[456]
of sql
[457]
to make that information actionable what
[460]
is the key as the strategy
[462]
why is the positive the negative impact
[464]
how much of your score was reduced or
[467]
improved based on a data driven versus a
[469]
disclosure
[471]
and what action what article what event
[474]
may have affected your brand or your
[476]
competitors
[477]
or your supply chain in a positive or
[479]
negative one
[480]
if you have any question about migrating
[483]
from a marketing concept of csr
[485]
to a data-driven esg and acting upon
[488]
those esg rating to a data-driven way
[491]
me and my colleague junta will be more
[492]
than happy to sit down with you
[494]
and your practitioner to enable that
[496]
transformation
[497]
knowing that all those notebooks are
[499]
publicly available
[501]
and can be used today on a data break
[503]
sometimes
[505]
thank you very much
Most Recent Videos:
You can go back to the homepage right here: Homepage