은행 데이터를 신뢰할 수 있는 AI 인사이트로 전환하기

00:13 Banks have customer data trapped across
core banking, CRM, risk systems, and so
00:18 many other sources.
Instead of waiting weeks for IT
00:21 pipelines,
the Denodo platform can deliver customer
00:25 insights instantly, simply because there
is no copying or moving data from these
00:29 source systems.
This is called zero copy delivery.
00:34 Here is the Denodo platform data
marketplace, which is an intuitive shop
00:37 for data experience.
Any business user here can just search
00:41 as though you were using an internet
search engine to find a curated data
00:44 product.
Let's do a quick search
00:48 for customer metrics.
As you can see, a data product has
00:52 appeared with a description.
This is a primary view for customer
00:56 activity analysis, and we can use it to
answer questions about active customers,
00:60 >> [music]
>> transaction volumes, visits, and
01:02 engagement.
Let's go inside and have a look at this
01:05 data product.
You can see that the semantics are
01:08 already being built into the product.
For example, it is tagged as personally
01:13 identifiable information, which is very
important.
01:17 We've already seen the summary when we
saw the preview of the product.
01:21 And looking at the lineage of the
product will give us a view as to the
01:23 source systems from which we derive the
actual insights.
01:27 Through universal connectivity, Denodo
federates real-time mobile logs from
01:31 Apache Kafka, checking accounts from
Postgres, ledger history from
01:35 Databricks, and profiles from Snowflake.
All this information is used to create
01:40 what we call an active context for AI.
This means that the data estate of this
01:44 bank is simultaneously connected,
governed, live, and semantically
01:48 coherent so artificial intelligence can
retrieve the right data, live data, and
01:52 with the appropriate guardrails.
Let us see an example of how artificial
01:57 intelligence can give [music] us
business insights from this data.
02:00 This is our AI-powered assistant, which
can be used for natural language
02:04 queries, and no knowledge of SQL is
required.
02:07 Let's give it a simple query.
List the name, email, and transaction
02:12 count for our top five customers as a
bulleted list.
02:15 Now, you may notice that here I am
logged in as an administrator with full
02:19 privileges.
So, when I query Denodo assistant, it is
02:23 scrolling through what we described as
our active context layer, which is
02:26 represented by the metadata you see
here.
02:29 And it is collecting the insights that
we asked for. And here we are.
02:33 These are the top five customers.
>> [music]
02:35 >> The names and emails are clearly visible
and the number of transactions.
02:40 Now, of course, not everyone should have
full visibility, and many banking
02:43 professionals should not have access to
customer information, such as email
02:46 addresses, for example.
If you remember, we had tagged this data
02:51 product personally identifiable
information.
02:54 So, what if I am a data analyst?
In this view, I am logged in as an
02:58 analyst, and I do not have the same
privileges as the administrator. [music]
03:03 Let us ask the Denodo assistant the same
question, but this time as a data
03:06 analyst, list the name, email, and
transaction count for our top five
03:10 customers as a bulleted list.
The process being followed is the same.
03:15 Denodo is querying the metadata, not the
actual data, to retrieve the information
03:19 and insights that we need.
And voila, here we go.
03:24 As you can see, the key difference is
that the names and the email address are
03:27 hashed, while the transaction counts
remain in plain text.
03:31 The Denodo platform's governed access
ensures security travels directly with
03:35 the data product, no matter the source.
Compliance is therefore automated at the
03:40 point of consumption.
This is Denodo as your AI data layer.
03:45 We turn scattered data into semantic
trust, giving your users and your AI one
03:49 secure view of reality.
03:52 >> [music]
>> Mhm.