![24
John
Kermit
Harry
Tom
Dobby
First Name
Configure automation to log the required
metrics – Dimension vs. Table Calculation
Table Calculation
Table calculations allow you to
perform ad-hoc calculations on
existing data. Examples
• String formatting
• Finding the year to date
total
Calculated in
Insights
Tied to a
specific
visualization
John Smith
Kermit Frog
Harry Potter
Tom Riddle
Dobby Elf
Full Name
Dimension
A column of data
• physical -> exist as columns
in the database
• Logical -> a calculation or
translation of actual data
Full Name
John Smith
Kermit Frog
Harry Potter
Tom Riddle
Dobby Elf
SUBSTRING([Full
Name], 1,
CHARINDEX(' ', [Full
Name]) - 1)
Q&A
Metrics &
Calculations
Log Metrics Build
Defined in
Studio
Logged in
Orchestrator](https://image.slidesharecdn.com/devdivesseptemberfinal-230929085625-3ce54530/85/Dev-Dives-Mastering-AI-powered-Document-Understanding-24-320.jpg)
Dev Dives: Mastering AI-powered Document Understanding
- 1. Dev Dives Go deeper, automate smarter
- 2. 2 Meet today’s team: Cristina Vidu Global Manager Marketing Community @UiPath Roxana Ivan Senior Product Marketing Manager @UiPath Sophia Zhylych Product Marketing Manager @UiPath
- 3. 3 Meet today’s speakers: Daniel Lerner AI/ML Solution Architect @UiPath Lahiru Fernando UiPath MVP Country Director / RPA Lead @Boundaryless Group
- 4. 4 About today’s meeting • Enjoy the next 50-60 min packed with best practices, lessons learned, use cases and live Q&As. • Join the poll. Explore and download your preferred Document Understanding Accelerators to automate faster. You’ll receive the recording and a full guide to Solution Accelerators, via your email. • Get answers to your questions and challenges. Please use the chat box for Qs during the presentation. Live Q&A session at the end. • You're encouraged to network and share your LinkedIn/Twitter in the chat. • Have fun! Feedback is welcome.
- 5. Mastering AI-powered Document Understanding UiPath Dev Dives Webinar Series
- 6. 6 Agenda 01 02 03 04 Intro to Document Understanding Measuring what Matters – Document Understanding & Insights Best Practices & Lessons Learned – Document Processing Case Study Product Roadmap & Solution Accelerators
- 7. 7 The enterprise is inundated with documents processed manually leading to business inefficiency Finance • Invoices • Purchase orders • Expense reports HR & People • Candidate applications • Onboarding documents Document processing challenges Limit business growth and scalability Labor-intensive document processing limits ability to scale efficiently and capture market opportunity. Result in poor customer experience Complex, unstructured data mandates human decisioning, slow onboarding, and servicing. Increase risk Higher chance of data input errors, missed information, and incorrect procedures. Sales • Contract agreements • Order amendments • Customer queries Customer queries • Customer emails • Customer tickets
- 8. 8 Intelligent document processing (IDP) with UiPath Real value. Real results. Accelerated productivity Higher accuracy Better customer experience Happier employees 600% increase in the volume of claims handled daily 70% time savings for 824,000 documents annually
- 9. 9 1 2 3 Understand → Act → Receive → Train & configure Get your documents processed intelligently Audit & analyze Monitor operational dashboards and get insights on performance • Initiate other automations • Input into system of record • Drive other actions UiPath Automation Validate the extracted information and handle exceptions Human in the loop Digitize AI-powered processing Digitize Classify Extract • Multiple languages • Various formats, tables • Handwriting & signatures • Skewed & low-quality scans • Checkboxes Documents Use the new data to improve model performance
- 10. Measuring what Matters Tracking value from your document processing automation
- 11. 11 Identify trends & weigh options Create business rule ML (re)training Open bug Document processing – path from start to finish can be more complex than you think External influence (4500 docs) Document and/or system of record have incorrect data System processing (500 docs) The system did not interpret document data correctly 100 invoices OCR issue 400 invoices Field not identified 400 invoices Wrong tax rate % 4000 invoices Address mismatch with system of record Fix system of record Avoid rework Automated resolution(s) via RPA increases overall business value 5,000 docs untouched 5,000 docs HITL 10,000 docs Document specialist corrects Dependency: collect & analyze metrics for all HITL events via Insights 100 invoices Duplicates found Don’t pay twice! System processing (500 docs) The system did not interpret document data correctly 100 invoices OCR issue 400 invoices Field not identified
- 12. 12 Utilization Errors Success Rate Duration Process Outcomes Time Saved Money Saved ROI Operational Datapoints Business Datapoints Business Value Reporting RPA Operations Orchestrator & Robot Data Automation Hub Document Understanding
- 13. 13 "How can I be sure my automations bring my internal stakeholders value?" "How do I track my team's operations? " What is usually asked after deploying a solution? Business Value Reporting RPA Operations
- 14. 14 Out of the Box Templates (easy starting point) Building dashboards in Insights – Dashboard Options Processes Queues Business ROI Business ROI Queues Robots and Machines Attended Reporting Attended-focused. Process metrics on most/rarely used and most Faulted. Also, has most and least active users, including Executions from Studio. Dollar and Time savings analyzed across Processes. Dollar and Time savings analyzed across Queues. Basic Process info with details on Errors. Includes most run Processes and Fault counts. Basic Queues info with details on Exceptions. Includes breakdowns per Queue for Status, Duration, and Date. Metrics focused on Robot and Machine assignments and Utilization Document Understanding Overview (preview) Document Understanding Processes (preview) High-level overview on Time Spendings, Processed Pages, Operations per Page, and Activity Usage. Comprehensive process- specific DU metrics focused on Time Spendings, Human Validation, Documents, Classification, and Extraction. Custom
- 15. 15 • How am I defining project success? How does solution performance impact business value? • What do I need to know about my operations as time goes on? • What info am I going to need to troubleshoot or investigate an issue? Identify questions you need answered • Business Process focused metrics (ex: Total Hours Saved per week) • Solution Performance focused metrics (ex: Number of System Exceptions per week) Determine metrics and calculations that answer questions • Has to be logged via automation workflow • Could be calculated in Insights but might be easier to calculate and log via automation workflow (ex: durations) • Easier to calculate in Insights (aggregations using multiple transactions) Configure automation to log the required metrics • Business Value Driven vs. Solution Performance • Single Use Case vs. Multiple Use Cases • Single Transaction vs. Multiple Transactions • Keep in mind personas tied to each dashboard Build dashboards Measuring what matters – framework
- 16. 16 Reporting/monitoring Identify questions you need answered Time savings/ increased throughput • What are my key value drivers? • How do I deem this project successful? • What % of my documents are not approved? • What are the top reasons they’re being rejected? • How long is it taking to validate an exception? • How many exceptions are due to external influence vs. system processing? • What fields are being extracted? • What doc types are being classified? Model output Extracted field predictions Doc type prediction Automated validation Business rule validations Confidence thresholds System of record lookups Human validation Average handling time Doc type/field validations Exception reason Data export Output accuracy Model accuracy Submission exceptions Cost/penalty avoidance Faster customer response • What business rule exceptions could be fired? • Do business rule exceptions have different priorities? Q&A Metrics & Calculations Log Metrics Build
- 17. 17 Determine metrics and calculations that answer questions Q&A Metrics & Calculations Log Metrics Build What are the most frequent reasons a document is rejected? Most Frequent Rejection Reasons Reason Frequency Rejection Reason (Extraction) Rejection Reason (Classification)
- 18. 18 Configure automation to log the required metrics Q&A Metrics & Calculations Log Metrics Build What are the most frequent reasons a document is rejected? Most Frequent Rejection Reasons Reason Frequency Rejection Reason (Extraction) Rejection Reason (Classification)
- 19. 19 Time Savings Total Time Spent (Automated Process) Number of Documents Manual Intervention Rate Number of Documents Processed Filename Number of Exceptions Rejection Reason (Extraction) Rejection Reason (Classification) Average Handling Time Number of Documents Validation Duration Validation Duration (Extraction) Validation Duration (Classification) Total Time Spent (As-Is Process) Number of Documents Manual Intervention Rate Average Handling Time Determine metrics and calculations that answer questions “How much time is my department saving each week?” Q&A Metrics & Calculations Log Metrics Build
- 20. 20 Time Savings Total Time Spent (Automated Process) Number of Documents Manual Intervention Rate Number of Documents Processed Filename Number of Exceptions Rejection Reason (Extraction) Rejection Reason (Classification) Average Handling Time Number of Documents Validation Duration Validation Duration (Extraction) Validation Duration (Classification) Total Time Spent (As-Is Process) Number of Documents Manual Intervention Rate Average Handling Time “How much time is my department saving each week?” Configure automation to log the required metrics Q&A Metrics & Calculations Log Metrics Build
- 21. 21 Time Savings Total Time Spent (Automated Process) Number of Documents Manual Intervention Rate Number of Documents Processed Filename Number of Exceptions Rejection Reason (Extraction) Rejection Reason (Classification) Average Handling Time Number of Documents Total Time Validating Documents Validation Duration (Extraction) Validation Duration (Classification) Total Time Spent (As-Is Process) Number of Documents Manual Intervention Rate Average Handling Time “How much time is my department saving each week?” Configure automation to log the required metrics Q&A Metrics & Calculations Log Metrics Build 1. Has to be logged via automation workflow (leaves of data model) 2. Could be calculated in Insights but may be easier to calculate and log via workflow (ex: durations) 3. Easier to calculate in Insights (ex: aggregations using multiple transactions) Deciding what data to log?
- 22. 22 Configure automation to log the required metrics Metrics Dimension Table Calculation Measure Dimensions, Table Calculations, and Measures can not only be used to track metrics, but can be metrics themselves Q&A Metrics & Calculations Log Metrics Build
- 23. 23 Configure automation to log the required metrics – Dimension vs. Measure Dimension Measure A column of data • physical -> exist as columns in the database • Logical -> a calculation or translation of actual data Aggregations and calculations across one or many rows • Ex: sum, minimum, maximum, average, median, or count Defined in Studio Logged in Orchestrator Defined and Calculated in Insights $150.00 $45.00 $15.00 $85.00 $65.00 Invoice Amount 123456 395827 793583 239582 129582 Invoice Number 2023-08-18 2023-08-18 2023-08-17 2023-08-17 2023-08-16 Date Processed $195.00 $100.00 $65.00 Invoice Amount 2023-08-18 2023-08-17 2023-08-16 Date Processed Q&A Metrics & Calculations Log Metrics Build
- 24. 24 John Kermit Harry Tom Dobby First Name Configure automation to log the required metrics – Dimension vs. Table Calculation Table Calculation Table calculations allow you to perform ad-hoc calculations on existing data. Examples • String formatting • Finding the year to date total Calculated in Insights Tied to a specific visualization John Smith Kermit Frog Harry Potter Tom Riddle Dobby Elf Full Name Dimension A column of data • physical -> exist as columns in the database • Logical -> a calculation or translation of actual data Full Name John Smith Kermit Frog Harry Potter Tom Riddle Dobby Elf SUBSTRING([Full Name], 1, CHARINDEX(' ', [Full Name]) - 1) Q&A Metrics & Calculations Log Metrics Build Defined in Studio Logged in Orchestrator
- 25. 25 Configure automation to log the required metrics Define datatype for target custom fields Build Visual in Dashboard Insights Orchestrator Store Log Data Robot Run Automation Add log statements in automation workflow Publish Automation Studio Custom Logging Queue Activities Q&A Metrics & Calculations Log Metrics Build
- 26. 26 Build dashboards Business Process (Solution) Document Lifecycle Specific Doc Processing Results Automation Workflows Custom Logging Aggregate Doc Processing Results Insights Studio Add Log Fields Queues Queue Activities Audit Trail Specific Doc Troubleshooting Business Value/ROI Overall Solution Performance Model Performance Q&A Metrics & Calculations Log Metrics Build
- 27. 27 Business Process Dashboard – Understanding the overall performance of your document processing use case Business Process SLA Violation Rate Submission & Approval Rate Most Frequent Rejection Reasons False Positive Rate Average Validation Time (Extraction) Num Duplicates Detected Average Field Accuracy Average Cycle Time Num Approved Invoices over Time FTE Throughput What are the most frequent reasons a document is rejected? Most Frequent Rejection Reasons Reason Frequency Rejection Reason (Extraction) Rejection Reason (Classification) **Not exhaustive Q&A Metrics & Calculations Log Metrics Build
- 28. 28 Document Lifecycle Dashboard – Understanding what happened with each transaction in my business process How long did it take to process a specific document? False Positive Document Type (Pre-Validation) Document Type (Post-Validation) Rejection Reason Reviewer (Classification) Field Values (Pre- Validation) Field Values (Post- Validation) Reviewer (Extraction) Validation Time Submission Time Number of Errors Document Lifecycle Total Processing Time **Not exhaustive Q&A Metrics & Calculations Log Metrics Build Total Processing Time Digitization Duration Classification Duration Extraction Duration Validation Duration Validation Duration (Extraction) Validation Duration (Classification) Data Export Duration
- 29. 29 Build dashboards – Scaling up Operations Document Lifecyle 1 Business Process 1 Document SMEs Managers Department Leads Directors Operations HQ CoE Document Lifecyle 2 Business Process 2 Business Users Team Managers Department Heads Directors Doc Processing Use Case 1 Doc Processing Use Case 2 Q&A Metrics & Calculations Log Metrics Build
- 30. 30 Demo
- 31. Document Understanding Best Practices & Lessons Learned By Lahiru Fernando
- 32. 32 Geography: Department: Industry: Products: EMEA Financial reconciliation Customer use case – foster family applications Insurance Document Understanding Implementation partner: Challenge Solution 80% Over 80 different document types in each foster care licensing application. Submitting applications for foster care licensing can be very time consuming because of the overwhelming amount of documentation required. The documents can be structured or unstructured, merged into one containing different document types with different layouts. Each case worker spend at least 5 to 6 hours processing each file. We built a UiPath automation solution that takes over the manual work, reducing the processing time from hours to minutes. Using state-of-the-art functionalities available in UiPath Document Understanding, we created classification and extraction methods to process all documents Time savings – from 6 hours to 10 minutes Results The new foster care families are able to welcome children to their families sooner – owing to the automation helping with the paperwork and the administrative backlog of applicants.
- 33. 33 Identified scenarios Implemented approach Delivered impact Process & document standardization • Field workers follow different methods to create the files resulting in structural issues in the file • Changes in document quality • No planned process improvements • Standardization steps for document creation to ensure all input files follow a specified standard • Introduced document pre-processing steps in the Document Understanding flow • Introduced a Continuous Improvement methodology • Improved time spent in document creation and meeting submission deadlines • Improved accuracy levels in classification and data extraction & improved STP rates • Improved and streamlined processing steps Addressing complex scenarios • Ambiguity in classifying documents manually • Complex business rule checks and activities • Conducted detailed document discovery sessions to identify standard document types • Cross-verify classification and extracted data through API/ Database integrations • More accurate and standardized classification approach across the business unit • Better accuracy levels and completeness of data before submitting into state applications
- 34. 34 High-level architecture Document collection Process documents Upload to state systems Start Check new documents Check for duplicates Add documents to queue End Start Digitize document Classify documents Validate classification Create split files Document data extraction Prepare output data Update master and child queues End Start Application login Case assignment Upload documents End Queue trigger Time-based trigger Time-based trigger
- 35. 35 Document Understanding flow Start Digitize Document UiPath Document OCR Omni Page OCR Classify Document ML Classifier Intelligent Keyword Apply Validations Valid Manual Verification Split and Merge Documents Generate Classification Training Data Extract Key Information Custom ML Model Form Extractor (IFE) Validate Extracted Data End Valid Connect to Internal Apps via API/ DB Post to Queues for Downstream Apps Low Confidence Valid Classification Invalid/ Incomplete Accurate/ Complete
- 37. Product Roadmap & Solution Accelerators
- 38. 38 Get more done with a digital workforce that seamlessly collaborates with your people and automates work via UI and API, powered with native integrated AI Automate Document Understanding Extract info from documents, images and more Expand unstructured document intelligence • Generative AI for querying documents with natural language • Enhanced Communications Mining + Document Understanding integration Accelerate time to value • AI-assisted active learning-based training (tag the minimum, real-time retraining) • Annotating documents with Gen AI • Enabling document classification with Generative AI User experience • Improved UX to help users build better models faster • Enhanced capability discovery, guided labeling experience and improved model insights Deployment insights & operations • Dashboards & document audit, containing metrics for STP rate, time saved, and more
- 39. • Reduce discovery time • Based on proven high-value automation opportunities • Build confidence in UiPath solutions Choose use cases with Accelerators • Reduce design time • Based on & incorporate best practices • Modularized design enables easy application swaps Reference solution designs based on UiPath best practices • Accelerate development • Orchestration included • Pre-built starting points for AI Customize and build The UiPath ™ word mark, logos, and robots are registered trademarks owned by UiPath, Inc. and its affiliates. ©2023 UiPath. All rights reserved. Discovery Build an automation pipeline Design Create solution design Development Orchestrate & build automations Automate Faster with Solution Accelerators 50% FASTER TOTAL AUTOMATION TIME Pre-built modular automation frameworks that accelerate time to value
- 40. 40 Features Package content ➢ Overview ➢ Step-by-step deployment guide ➢ High level & detailed solution designs ➢ Document Understanding, dispatcher, and performer workflows ➢ Orchestration assets ➢ Incorporate the Document Understanding Process template into the REFramework ➢ Equipped with Specialized AI models tailored to document types ➢ The models may be further trained to meet a customer’s specialized needs. UiPath Document Understanding UiPath Action Center UiPath Document Understanding Accelerators → Document Understanding Accelerators at UiPath Marketplace
- 41. 41 Billing 1. Purchase Orders 2. Receipts 3. Remittance Advices 4. Utility Bills - Generic Insurance 1. Healthcare Insurance Claim Form 1500 2. Ins. Commercial Application - Acord 125 3. Ins. General Liability - Acord 126 4. Ins. Liability Coverage - Acord 25 5. Ins. Property Section - Acord 140 6. Ins. Umbrella or Excess Section - Acord 131 Invoice 1. Invoices or Credit Notes - China 2. Invoices or Credit Notes - Generic 3. Invoices or Credit Notes - India 4. Invoices or Credit Notes - Japan Manufacturing 1. Children Product Certificate 2. EU Declaration of Conformity UiPath Document Understanding ML models in Accelerators Personal ID 1. ID Document Cards & Driver Licenses 2. Official Travel Document - Passports Tax Transportation Treasury 1. Employment Eligibility Verification Form i9 2. Personal Pay Slips 3. Tax Form 1040 4. Tax Return Form 4506T and 4506C 5. Taxpayer ID Number & Certification Form W9 6. Wage and Tax Statement Form W2 1. Bills of Lading 2. Certificate of Incorporation - Good Standing 3. Certificate of Origin 4. Invoices - Shipping 5. Packing Lists - Shipping Document 6. Vehicle Titles 1. Bank Statements 2. Checks - Bank Pay Order 3. Financial Statements Form 10-K
- 42. 42 Have you registered for this webinar? • Jeffrey Martin, Solution Architect at Encova Insurance • Steve Tegeler, Senior Director Solution Engineering at UiPath • Todd Pratt, AI/ML Solution Engineer at UiPath You can still register to watch it!
- 43. 43 Date/Time Topic Status October 26 9:00 AM EDT / 2:00 PM BST Explore UiPath Solutions Management for seamless packaging, deployment and maintenance Register AMER Register EMEA & APAC November 21 9:30 AM EST / 3:30 PM BST Accelerate development with Generative AI and automation – Wingman in action Coming soon Next steps Explore the collection of UiPath Document Understanding Accelerators > https://bit.ly/3RBg1xG. Don't miss the next Dev Dives sessions. Save your seat > https://bit.ly/Dev-Dives_2023
- 44. Thank you for your participation! Join us: community.uipath.com