Learn, Discover & Build- Generative AI powered Intelligent Document Processing (PDF Slide Compendium).pdf

AWS

1. GENERATIVE AI POWERED CONTACT CENTER TRANSFORMATION
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Generative AI powered Intelligent
Document Processing
AWS Partner Network
Generative AI Use Cases Enablement Series
January 23
rd
, 8-9 AM PT
Amit Singh
G l o b a l U s e C a s e L e a d e r ,
G e n e r a t i v e A I P a r t n e r
A c c e l e r a t i o n
Anjan Biswas
S e n i o r S p e c i a l i s t
S o l u t i o n A r c h i t e c t -
I D P , W W S O A I / M L
Shibin Michaelraj
S e n i o r P r o d u c t
M a n a g e r -T e c h ,
A I S e r v i c e s

2. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Agenda
2
•Generative AI on AWS
•Intelligent Document Processing –challenges, key use cases,
business value drivers
•Generative AI powered IDP: How to get started?
•Demos
•Next steps & resources

3. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Generative AI on AWS
3

4. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Generative AI Stack
GPUs Inferentia Trainium SageMaker
EC2 Capacity Blocks NeuronUltraClusters EFA Nitro
Amazon Bedrock
GuardrailsAgentsCustomization Capabilities
Amazon Q Amazon Q in
Amazon QuickSight
Amazon Q in
Amazon Connect
Amazon
CodeWhisperer

5. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved. 5
Choice of leading FMs via single API
Model customization
Retrieval Augmented Generation (RAG)
Agents that execute multistep tasks
Security, privacy, and safety
Amazon Bedrock
The easiest way to build and scale
generative AI applications with
foundation models

6. © 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Bedrock supports leading foundation models
Amazon Titan
Summarization, image and text
generation and search, Q&A
Command & Embed
Text generation, search, classification
Jurassic-2
Contextual answers, summarization,
paraphrasing
Claude 2.1 & Claude Instant
Summarization, complex reasoning,
writing, coding
Stable Diffusion XL 1.0
High-quality images and art
Llama 2
Dialogue use cases and language tasks
6

7. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Intelligent Document Processing
7

8. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 8
Legacy document processes
do not meet today’s needs
Legacy OCR and manual processes
are time-consuming, error-prone,
and expensive
Manual processes do not scale
easily with document volume
Difficult to find useful information
needed for business decisions
Faster document processing
shortens decision cycles;
this enables you to serve
more customers and have
people do higher value tasks

9. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 9
Use-cases
Improving loan underwriting process productivity
Process invoices at scale for accounts receivable
Scaling claims processing through automation
Health insurance companies spend millions of dollars to extract sensitive information from claims forms
and accompanying attachments to perform their business operations. To reduce manual labor, the
company wanted to automate the process.
Incomplete loan packages, missing data, and the discovery of additional documentation requirements
during the underwriting process often creates more work and slows the process down, resulting in fewer
loans reviewed by underwriters and added lender costs.
Efficiently handle large volumes of invoices in the accounts receivable department, utilizing automation
and AI to streamline and speed up the invoice processing workflow.

10. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 10
Amazon Intelligent
Document
Processing is used
across industries
Insurance Medical
Accounting Education
Real estate Legal
Finance Business operations
Tax management And many more…

11. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 11
AWS IDP Business Value
AWS IDP has proven to drive compelling ROI with rapid payback (Source: Forrester Total Economic Impact Study of
AWS IDP)
Profit drivers
Launch new services
facilitated by IDP
Grow revenue from existing
services by increasing
speed with fewer errors
Reduce costs Improve customer
and employee
experience
Cost drivers
Business or people
process change
System integration
of IDP
Pay-per-use cost
model of AWS IDP
73%
ROI*
$4.6M
NPV*
< 6
Months
Payback*
*Based on assessment of organizations that processed 10M docs monthly

12. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Generative AI powered IDP
12

13. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 13
IDP with Foundation Models
13
Textract and Bedrock
✓Accurate and scalable OCR, forms,
tables, layouts, signatures, invoices
✓Easy to implement Q&A provides
exact answers from the document
✓Confidence scores for exact match
vs. original content on the page
✓Find key business entities based on examples
you provide
Foundation Models
Normalize outputs based on
your desired data format
Q&A can summarize or use
logic to provide answers
Checks for reasonable answers
(a SSN with 8 digits is an error)
Classify or find business entities
using built in or prompted categories
(examples not required)
Summarize documents with
option to use redaction
E2E managed service for document processing
using FMs -coming soon !!!

14. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
IDP with purpose built ML
Forms
Loan application
Optical character
recognition (OCR –
Printed, Handwriting)
Tables
Layout
Form 1003
Domain-specific APIs
(Expense, ID, Lending)
Signatures
Queries &
Custom Queries
?
NEW
NEW

15. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Queries/ Custom Queries
15
Vision-Language model to extract data using natural-language Q&A
Example questions for a pay stub:
1.What is YTD gross pay?
2.What is gross pay this period?
3.What is employee SSN?
4.How many sick hours did the employee earn?
Questions
Document
1.What is YTD gross pay?
$23,526.80
2.What is gross pay this
period? $452.43
3.What is employee SSN? 987-
65-4321
4.How many sick hours did the
employee earn? 16.00
Output
Document Types –Checks, paystubs, invoices, birth/death certificates, etc.

16. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
IDP workflow with generative AI
Categorized
documents
Route based on
business rules/type
of info needed
Vector DB,
knowledge
base
Categorize, tag documents
with Classification
Classification
OCR
Amazon
Textract
Amazon
Bedrock
Derive key insights with
Extraction & Enrichment
Q&A, forms,
tables, and more
Summarization,
normalization,
chatbots, and more
Amazon
Textract
Amazon
Bedrock
Human
in the loop
Post processing
completeness
check,
rule-based
validation &
review
Route low
confidence
predictions
Route high
confidence
predictions
Verification & human review
Send data to
downstream
databases/
apps
Spell check,
corrections
Amazon
Bedrock
Documents
Data
capture
Amazon S3
Bucket

17. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Demos
17

18. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Next Steps
18

19. GENERATIVE AI POWERED CONTACT CENTER TRANSFORMATION
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Explore, learn, build and co-sell
19
Explore Generative
AI services from AWS
& Intelligent
Document
Processing use cases.
12
Learn with
resources available
on AWS Skill
Builder & AWS
Partner Central
(Generative AI CoE)
3
Develop new
solutions/offerings
around use cases or
migrate existing
solutions to AWS
4
Use the guidance
shared in this
session to co-sell
with AWS

20. TheGenerativeAICenterofExcellenceforAWSPartnersprovidesvital learningresources forpartners embarking ontheir
AI journey.
Resourcesinclude industry-focused training,thought leadership insights ontheuseofthetechnology,anduse-case-specific
bestpractices.
Access& Resource
Guide
RecommendedResources
Accessand Engage with the CoE
Most resources require access to Partner Central. Sign up here.
The
Generative
AI Partner
Resource
Page
Partner
Central
Resources > Guides >
Generative AI Center
of Excellence
Engage with Us
Build a 2024 CoEEngagement Plan
Email Us
Give Feedback & Request Resources
Complete Survey
Nominate Content to the CoE
Contact PDM or email Jacob Newton-Gladstein
Access the CoE
Start With the Partner Playbook. Review
by Interest: Build Market, Sell, and Grow
Dive Deep into Amazon Bedrock Asset Pack
Understand Deployment and
Economics of Generative AI
Dive Deep into Anthropic Asset Pack
Get Industry-Specific Resources for
Healthcare and Life Sciences
Learn Market Trends and Opportunities
in a Q&A with
AWS and McKinsey & Company

21. Resources
Blogs
Textract + Bedrock: https://aws.amazon.com/blogs/machine-learning/amazon-textracts-new-layout-feature-
introduces-efficiencies-in-general-purpose-and-generative-ai-document-processing-tasks/
Custom Queries: https://aws.amazon.com/blogs/machine-learning/customize-amazon-textract-with-
business-specific-documents-using-custom-queries/
GitHub
https://github.com/aws-samples/aws-ai-intelligent-document-processing
Workshop
https://catalog.us-east-1.prod.workshops.aws/workshops/c2af04b2-54ab-4b3d-be73-c7dd39074b20/en-US

22. GENERATIVE AI POWERED CONTACT CENTER TRANSFORMATION
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 22
Reach out or opt-in for the
Roadmap session –GenAIpowered IDP capability
(NDA Required)
Questions & Feedback
[email protected]

23. GENERATIVE AI POWERED CONTACT CENTER TRANSFORMATION
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 23
Please complete the survey at
the end of the training & share
your feedback with us.

24. UPDATE THIS PRESENTATION HEADER IN SLIDE MASTER
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. © 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Thank you!
24