Intelligent Automation for Fun and Profit!

E7d6e390a90513756419be75a43609ca?s=47 finid
June 29, 2019

Intelligent Automation for Fun and Profit!

The term “intelligent automation” has been used to describe everything from the application of AI/RPA to Python coding when discussing how to increase business process productivity; but what exactly is it? What’s needed is a common sense definition and an approach to understanding how to use it effectively.

This requires understanding the strengths of the various technology components (e.g., machine learning, deep learning, RPA, “smart RPA, computer vision, etc.) as well as how they can be effectively applied to achieve cost effective productivity growth within an organizations unique set of processes and tasks.

E7d6e390a90513756419be75a43609ca?s=128

finid

June 29, 2019
Tweet

Transcript

  1. None
  2. Big Data & AI Conference Dallas, Texas June 27 –

    29, 2019 www.BigDataAIconference.com
  3. 1 Intelligent Automation for Fun and Proft Big Data &

    AI Conference Dallas, TX 6/29/2019 Edward M.L. Peters, Ph.D. Founder and CEO Copyright © 2019 Data Discovery Sciences, LLC. All
  4. 2 Dr. Edward M.L. Peters Ph.D. Applied Economics, University of

    Amsterdam Thesis “Discovering Value Leaks and Service Imperfections in Business Processes” M.S. B.A. Industrial Engineering, Lehigh University Government/Russian studies, Lehigh University - Stanford University Graduate School of Business (EPGC) - MIT, Sloan School of Management (Corporate Strategy) - Carnegie Mellon University (Entrepreneurial Management) Copyright © 2019 Data Discovery Sciences, LLC. All LinkedIn.com/in/edward mlpeters Ernst & Young Entrepreneur of the Year - Regional Finalist Baltimore/Washington Region 2003, 2004 Dallas Region 2008 Maryland Tech Council Entrepreneur of the Year, 2004
  5. Lot’s of confusion: – RPA on steroids – AI –

    Next step in business process automation (BPM) DEFINE D What is Intelligent Automation? 5 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  6. • Intelligent Automation – Mimics activities carried out by humans,

    and then… …learns to do them even better – Rule-based automation (RPA) is augmented with decision-making capabilities (focus on process) – AI is about data and algorithm to augment human judgment. DEFINE D What is Intelligent Automation? 6 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  7. Intelligent Automation to do exactly what? Improve Quality Increase CapacityThe

    Zero- Point Reduce Cost Eliminate Human Labor 5 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  8. • High Interaction • Low Transaction Ex: Contract Awarding •

    High Transaction • High Interaction Tasks Ex: Customer Service • High Transaction • Low Interaction Tasks Ex: Invoice Processing • Low Transaction • Low Interaction Tasks Ex: Machine Maintenance Human Touch Transaction Volumes HIGH HIGH LOW Intelligent Automation Framework™ 8 Copyright © 2019 Data Discovery Sciences, LLC.
  9. Human Advisory Augmentation (AI/Predictive Analytics Process Automation Human Augmentation (Cognitive

    Augmentation) Autonomous Allocation (Hybrid Automation) Human decision making replaced by software robots following deterministic rules. RPA robots may be augmented with other technologies (e.g., computer vision) to increase efciency. Human interaction and decision making augmented by bots with cognitive enhancements (e.g., Chatbots, NLP, voice recognition and input) Humans advised by bots with decision- capable cognitive enhancements (e.g., ML, Deep Learning) Humans fully directed by software robots (e.g., IoT +Drones+ ML/Deep Learning+ optimization and scheduling) Human Touch HIGH Transaction Volumes HIGH LOW Intelligent Automation Framework™ 9 Copyright © 2019 Data Discovery Sciences, LLC.
  10. Purchase Planning Vendor Mgmt. Purchasing Receiving Invoicing Vendor Advances Vendors

    Reconciled • Opportunity Identification • Proposal Solicitation • Proposal Analysis • Contract Award • Goods Receipt & Verification • Invoice Digitization • Invoice Prioritization • Validate OCR Fields • Verify Invoice Details • Invoice Posting • Invoice Post GL • Post Pre-Approved Invoices • Investigate Vendor Mismatches • Contact Vendor for Resolution Procure to Pay Process PROCESS TASKS PROCESS ACTIVITIES 1 Copyright © 2019 Data Discovery Sciences, LLC.
  11. The Intelligent Process Automation Framework TM Human Touch Transaction Volumes

    HIGH HIGH LOW • Award & close contracts • Invoice Audit & Review • Investigate vendor mismatches • Contact vendor for resolution • Invoice digitization • Validation of OCR fields • Verification of Invoice details • Invoice posting • Input to GL • Post pre-approved invoices 1 Copyright © 2019 Data Discovery Sciences, LLC.
  12. Human Touch Transaction Volumes HIGH HIGH LOW Invoice digitization KPI:

    Reduce reject rate to <5% Desired outcome: Lower cost, reduction in manual interventions, and customer calls Vendor Inquiry KPI: >90% 1st call resolution Desired outcome: Improved customer experience measured on surveys, lower cost, higher volume of calls/worker Contract Audit & Review KPI: >95% within audit guidelines Desired outcome: Fewer adjustment costs, improved fnancial reporting, future contract language, credit granting, and vendor contract renewals The Intelligent Process Automation Framework TM 1 Copyright © 2019 Data Discovery Sciences, LLC.
  13. 1 Copyright © 2019 Data Discovery Sciences, LLC. Data and

    data- types required to perform the task Decisions can be either deterministic (rules-based) or stochastic (probability based) The desired result of the task execution as measured by process metrics or KPI’s The relevant next step or state change after decision made The Zero Point MethodTM Process Task Primitives
  14. 1 2 Copyright © 2019 Data Discovery Sciences, LLC. All

    Rights Reserved. Levels of Applied Intelligence RP A Imitate human actions Rule- based decisions Structured Data Enhanced RPA Augment “Bot” Probabilistic decisions Structured Data Semi-Structures Text, Image Cognitive Augmentation Augment “Human” Probabilistic decisions Structured Data Semi-Structures Text, Image, voice Predictive Augmentation Augment “Human” Probabilistic decisions Analytical Outcomes Structured Data Semi-Structures Text, Image, voice
  15. Learning and Predictive Requirements Copyright © 2019 Data Discovery Sciences,

    LLC. • Deep Learning Models • Predictive • Prescriptive • Structured data • Semi-structured data • Imaged • Text • Voice • Other • ML • NLP • Computer Vision • Structured data • Semi-structured data • Image • voice • Deterministic decision rules • Structured data • Semi-structured data • Computer Vision/OCR AI Capability and Data Type Requirements 13 The Intelligent Automation Framework™
  16. “RPA is about taking the robot out of the person”

    - McKinsey The Intelligent Automation Framework™ 1 6 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  17. Now the fun begins…. 1 7 Copyright © 2019 Data

    Discovery Sciences, LLC. All Rights Reserved.
  18. Intelligent Automation for Profit 1 8 Copyright © 2019 Data

    Discovery Sciences, LLC. All Rights Reserved.
  19. Initi al RPA only RPA + IDP RPA + IDP

    + RPA + IDP + Chatbots + Other Chatbots AI FTE's 82 29 19 17 4 Robots 0 22 22 23 27 Operational Cost $3,075,000 $1,637,500 $1,387,500 $1,337,500 $1,3 50,0 00 $3,500,0 00 $3,000,0 00 $2,500,0 00 $2,000,0 00 $1,500,0 00 $1,000,0 00 $500,00 0 $0 9 0 8 0 7 0 6 0 5 0 4 0 3 0 2 0 1 0 0 WORKFORCE COMPOSITION Invoice Processing Learning Level 0 FTE vs. Robots Invoice processing Estimated Cost Reduction 47% decrease via RPA Reject rate 20% 1 9 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  20. • Thousands of pend types  Pends vary by claim

    type (dental, medical, etc.)  Pends vary by claim processing system  Pends vary based on coverage / treatment / geography • Claims processing changes  New regulation brings new requirements  New products (e.g. HSA) bring new issues • Complex, partially document adjudication process  Real process for resolving pends resides in staff 35% 12% 8% 7% 5% 3% 4% 1 1 % 3% 21% Insurance Claims Adjudication Reasons for Pended Claims Duplicate Claims Submitted Lack of Necessary Information No Coverage Based on Date of Service Non-covered / Non-network Benefit or Service Coordination of Benefits Coverage Determination Utilization Review Authorizatio n Pre-existing condition review Invalid Codes submitted
  21. Ed it Cod e Hou rs Man Days Man Months

    Monthly Cost Yearly Cost FG 72.30 10.63 233.92 $950,416.14 $11,404,993.68 BN 23.57 3.47 76.26 $309,864.19 $3,718,370.32 MK9 21.60 3.18 69.87 $283,882.86 $3,406,594.36 LK 20.61 3.03 66.67 $270,880.31 $3,250,563.78 PO 17.98 2.64 58.18 $236,397.77 $2,836,773.24 W3E 12.66 1.86 40.97 $166,455.49 $1,997,465.91 BC 11.76 1.73 38.04 $154,568.02 $1,854,816.29 KL9 11.12 1.63 35.97 $146,131.12 $1,753,573.45 W32 9.89 1.45 32.00 $130,017.74 $1,560,212.91 ZIU 9.41 1.38 30.44 $123,671.80 $1,484,061.61 LI9 6.18 0.91 20.01 $81,285.79 $975,429.46 NNB 6.02 0.89 19.47 $79,118.65 $949,423.78 DRG 3.57 0.52 11.54 $46,892.07 $562,704.89 MM2 3.41 0.50 11.04 $44,845.07 $538,140.88 Top 15 Edit Codes Direct Labor Cost by Code and Time 2 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  22. PROCESS TASKS PROCESS ACTIVITIES Sample Receipt Sample Preparation Testin g

    Analysis Reporting and Customer Service • Receive and log sample • Prepare Samples • Batch into trays • Calibrate LCMS • Update LCMS Toxicology Sample Processes • Tokenize HIPPA Data • Batch for Prep • Batch for Analysis • Run test sample • Run tray for analysis • Review LCMS output • Check FDA schedule tests • Review normal results • Adjust ‘peak’ issues • Re-run test if required • Send report to customer • Customer service • Customer billing • (Customer onboarding) 2 Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved.
  23. 2 Copyright © 2019 Data Discovery Sciences, LLC. Human Touch

    Transaction Volumes HIGH HIGH LOW The Intelligent Process Automation Framework TM • Review LCMS Output • Check FDA Schedule Test • Review Normal Results • Adjust “Peak” issues • Re-run Tests as Required • Customer Inquiry • Customer On-Boarding • Receive Sample • Tokenize HIPPA Data • Batch for Prep • Prepare Samples • Batch into Trays • Prepare for Analysis • Calibrate LCMS Run Test Sample • Run Tray for Analysis • Update LCMS • Send report to Customer
  24. $3,500,0 00 $3,000,0 00 $2,500,0 00 $2,000,0 00 $1,500,0 00

    $1,000,0 00 $500,00 0 $0 45 40 35 30 25 20 15 10 5 0 Initi al RPA only RPA + IDP RPA + IDP + Chatbots RPA + IDP + Chatbots + Cognitiv e Toxicology Testing and Data Review Operating Cost Reduction with RPA & Cognitive Technologies FTE' s Robot s Operational Cost Toxicology Results Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved. 2 4
  25. Now the fun begins…. Copyright © 2019 Data Discovery Sciences,

    LLC. All Rights Reserved. 2 5
  26. Now the fun begins…. • Theresa M. Welbourne • USC

    Marshall School of Business – Studies “Corporate Death”! • What separates organizations who survive from those that don’t: Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved. 2 6
  27. • Welbourne… – When employees understand the inner workings of

    the organization, and they begin to engage in behaviors that support the company overall (instead of just their own job), this type of activity is not easily replicated by another organization, and these behaviors bring high value. Now the fun begins…. Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved. 2 7
  28. • Welbourne – …if a company values only the core

    job role, and employees engage in behaviors exclusively associated with the core job, these jobs are very easy for competitors to copy. It is a simple matter to replicate this type of company, hire employees Now the fun begins…. Copyright © 2019 Data Discovery Sciences, LLC. All Rights Reserved. 2 8
  29. Dr. Ed’s Three foundational principles of Intelligent Automation: 2 Copyright

    © 2019 Data Discovery Sciences, LLC. All Rights Reserved. Conclusion 1. Software robots and cognitive computing will do to services what industrial robots did to manufacturing 2. All work that can be performed by software robots and cognitive computing will be 3. As software robots and intelligent automation become ubiquitous, an engaged workforce will become the most important resource. epeters@datadiscoveryscience s.com