This document discusses using artificial intelligence to enhance digital asset management operations. It makes the following key points:
1. AI can help identify and extract metadata from digital assets to improve searchability and reuse of content across systems. This reduces duplication and helps coordinate work among creative teams.
2. There are two types of AI: generic models that provide common services like classification and enrichment, and business-specific custom models that deliver more relevant insights tailored to a company's specific domains and use cases.
3. Business-specific models can be continuously trained on a company's own content and data to provide predictive outputs. This increases the business value and accuracy of metadata compared to generic models.
3. Access&Utilize
MediaAssets
• Know what you have
• Avoid expensive and inefficient
duplication
• Coordinate multiple creative suppliers
• Centralize managed collections of
content and metadata
4. Digital Assets Across The Enterprise
4
Create Make Sell
Materials Design PhotoStudio
CampaignDesign ProductKnowledge
ContentHub
PackagingDesign
5. Accelerating
IdeastoMarket
By connecting content assets to
data throughout the product
lifecycle from ideation to support.
For a more efficient digitalsupply
chain, and an increase in the value
of the assets.
6. Organization’s systems are not fully connected to each other
Time spent daily looking for information
See the potential
of AI to automate
mundane tasks
*Results of a November 2019 survey of UK Financial Services companies
Believe their
organization lacks the
skills to capitalize on AI
Average
number
of solutions
60% 60%
Theoldwayof
managing
content & data
isn’treally
working
anymore.
7. 7
Master Metadata Model
Metadata A
Metadata B
Metadata C
Metadata D
Metadata E
Metadata F
System1 System2 System3 System4 Systemn
Metadata A
Metadata B
Metadata C
Metadata D
Metadata E
Metadata A
Metadata B
Metadata C
Metadata D
Metadata E
Metadata A
Metadata B
Metadata C
Metadata D
Metadata E
Metadata A
Metadata B
Metadata C
Metadata D
Metadata E
Metadata A
Metadata B
Metadata C
Metadata D
Metadata E
EnrichingContent
AcrosstheEnterprise
8. 8
1 Recognize content types
4 Predictively deliver information
7 Identify outlying data points
2 Extract data from content
5 Analyze usage and importance
3 Enrich content metadata
6 Recognize patterns and connections
ThePromiseof
Artificial
Intelligence
Understandcontentand
dataaswellasa
knowledgeablehuman,
butatscale.
9. Generic
Connect to a broad set of public services for common use cases (general
classification, enrichment, OCR, speech-to-text, etc.).
Commodity models provide generic services.
Business-Specific
Deliver highly relevant insights and enrichments to enable specific
business use cases / domains.
Custom models deliver more meaningful outcomes for business.
There are
2 types of AI
Enrich.
Decide.
Optimize.
10. Generic AI
Configurable
Configure content structure
and processes
Normalized
Use different providers for
more power
Scalable
Enrich bilions of documents
and digital assets
Consolidated
Enrich documents
and searches
Image& Video
Recognition
Transcribe
Text Extract
Comprehend
Translation
10
11. Land Vehicle Vehicle Car Motor Vehicle
Transport Parking Mode of Transport Van
Automotive Tire Bumper MinivanTire
Sport Utility Vehicle Family Car
Vehicle Registration Plate
Chevrolet Tahoe
Asphalt Gas Minibus
GenericAI=GenericData
Note: Depicts actual data generated from Ford image by Google Cloud Vision API service
Demos ability of AI/ML to
apply data to assets
Generally useful in increasing
findability and reuse
Relevancy and accuracy of
the model is inherently limited
1
2
3
Model returns a set of labels
lacking business-critical data
elements
4
12. Business-Specific
AI/MLServices
ContentBots
Secure, customer models trained
on your own content.
Multiple Inputs/Outputs
Train with both content
(documents, media) & data.
Predict multiple values at once.
ContinuousTraining
Content bots continuously evolve
in an automated training cycle.
Active Learning
Gamified “human-in-the-loop”
system for over-the-shoulder
validation & learning.
ContentCreation
Track all-machine generated
content (metadata) and
corrections.
Performance Monitoring
Real-time performance reporting.
Quickly identify corruptions &
degradation in models.
Fail Proof
All content actions are reversable.
Models are versioned & can be
reverted to previous versions.
Auditable
Review training and evaluation
datasets for each Content Bot.
Enterprise-ready ML.
13. Business-SpecificAI=Information
Brand: Chevrolet
Model: Tahoe
License St.: Illinois
License No.: K24 1771
Custom models produce much
more relevant data values
1
True entity extraction enables
workflow automation
2
Business value increases with
specificity of the data
3
Veh. Color: Gold Mist Metallic
Brand: Dodge
Model: Ram Van
License St.: Illinois
License No.: XXX 8242
Veh. Color: Bright White
Operator: John Smith