Achieving Automation through Cognitive Technologies - Deloitte
2 likes331 views
Timothy Ho from Deloitte discusses seven technologies driving change in finance: cloud, robotics, cognitive computing, in-memory computing, blockchain, advanced analytics, and core modernization. The roles of finance are changing from operator and steward to also include strategist and catalyst. Automation in finance can range from basic monitoring to fully automated processes. Robotic process automation can perform repetitive tasks across systems. Risk analytics uses predictive modeling to identify risks. Cognitive technologies like natural language generation can automate decision support. Getting started requires establishing a digital leadership team and piloting solutions quickly.
Achieving Automation through Cognitive Technologies - Deloitte
- 1. Achieving Automation through Cognitive Technologies Timothy Ho Deloitte
- 2. Technology is the key driver of change in finance
- 3. Cloud Process Visualization Robotics Cognitive Computing In-Memory Computing BlockchainAdvanced Analytics Seven technologies have growing relevance for how the work of finance gets done. Core Modernization Exponentials
- 4. Finance is expected to deliver in four key roles, however the balance of responsibilities is changing Operator StrategistCatalyst Steward How much time (in %) is your finance organization spending in top half vs lower half?
- 5. Definition : Automation Spectrum
- 6. The future of operational Finance is the establishment of a Finance factory; a fully automated control center with process visualizations tools Monitoring of processes on a continuous basis to provide insight Process Automation to drive greater productivity and quality
- 7. Robotic Process Automation (RPA) Opening email and attachments Logging into web/ enterprise applications Moving files and folders Copying and pasting Filling in forms Reading and writing to databases Making calculations Scraping data from the web Connecting to system APIs Extracting structured data from documents Collecting social media statistics Following “if/then” decisions/rules RPA is… RPA is not… Computer-coded software Programs that replace humans performing repetitive rules-based tasks Cross-functional and cross- application macros Walking, talking auto-bots Physically existing machines processing paper Artificial intelligence or voice recognition and reply software What itcan do
- 8. RPA Demo Video Process automation - generate billing in the system Process steps: 1. Receive email with attachment (in pdf – standard format) 2. Convert data into an excel file – Copy and paste from pdf 3. Log in to system 4. Go to generate billing screen and input billing request – Copy and paste from excel file 5. Send email to requestor to confirm process is completed Video Link: https://youtu.be/FV8lM9SIFQ8
- 9. RPA Benefits
- 10. Risk Analytics – Deloitte Analytics Platform Use more data, quickly • Unleash enterprise data, including internal and 3rd Party sources • Quickly cleanse and map data for analysis • Enhance data quality and completeness Produce better insights • Leverage proven risk models across business processes Procure-to-Pay, Payroll, Travel & Expense, and P-Cards • Identify emerging patterns driving anomalous behaviors through trending and benchmarking • Aggregate risk and use historical results to reduce false positives Understand business impact • Easy-to-use dashboards enable users to investigate and report results • Review historical results of similar risks identified to predict potential outcomes Analytics Platform Understand the risk universe • Deliver predictive risk modeling and risk sensing capabilities Develop the audit plan • Perform dynamic risk assessments • Generate dynamic audit plans Design the audit programs • Continuously audit / monitor risk & controls • Enable remote auditing • Conduct exception-based audits Deliver results & insights • Automated advanced dashboards
- 12. “Weare drowning in information but starvedfor knowledge”- JohnNaisbitt1
- 13. What is analytics? Predictive modeling/ optimization Scenario/tradeoff simulation Pushed exceptions and alerts Company and individual performance visibility Operational reporting Hindsight Enterprise data management Foresight Insight Industrialized Dashboards and KPIs Bespoke Traditional Analytics Cognitive Computing and Machine Learning at Scale
- 14. Analytics is evolving Drivers & disruptors Evolution of analytics IN-MEMORY PROCESSING INTERNET OF THINGS INTELLIGENT AGENTS TEXT ANALYTICS MACHINE LEARNING CROWD-SOURCING & COMPETITIONS ADVANCED HUMAN COMPUTER INTERFACE EXPONENTIALS CYBER SECURITY ARTIFICIAL INTELLIGENCE & COGNITIVE ANALYTICS CLOUD VISUALIZATIONS DATA LAKES BIG DATA & PREDICTIVE ANALYTICS TABLE STAKES Data proliferation New data sources Technology disruption Innovative Science DATA WAREHOUSING & DATA BUSINESS MODELING INTELLIGENCE LEVELOFENTERPRISE-WIDEADOPTION
- 15. Data Extraction - Argus Argus
- 16. CognitiveSpend retains human intelligence to get smarter as more data is processed How the Tool Learns Feedback Portal Dynamic in-tool pivot table to capture human input Initial Training Data Initial Classification Additional Feedback Previous feedback from other projects totaling ~$50B of spend across 100M line items 250 most meaningful line items (identified using data science) classified by user to jumpstart classification Additional amendments by user are retained for learning purposes + + How the Tool Classifies CognitiveSpend leverages previous classifications to predict classifications for data it has never seen before Example Classification Line Item Description: “SEM” Rules-Based Unknown Cognitive Search Engine Marketing 95%+=Accuracy through Automation
- 18. Cognitive Insights ➢ Extract concepts and relationships from various data streams, text sources and social media and generate new insights including market threats and opportunities through cognitive analysis and correlation of data ➢ Deliver applications that intelligently “understand” the context and provide the right insight at the right time explaining why it is relevant and important
- 19. Statistical Modelling $0 $50 $100 $250 $200 $150 $300 $350 Revenue(millions) 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year $0 $50 $100 $150 $200 $250 $300 $350 Revenue(millions) 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year Traditional Approach: Linear Projection Predictive Approach: Multiple Linear Regression Forecast Forecast Year 08 09 10 11 12 13 14 15 16 Revenue $150 $155 $202 $193 $210 $280 $293 $311 $336 Year 08 09 10 11 12 13 14 15 16 Revenue $150 $155 $202 $193 $210 $280 $293 $266 $254 Var A* 2 3 5 4 6 7 8 2 3 Var B* 7 6 5 6 4 4 3 4 5 Var C* 1 1 4 5 3 7 8 5 7 R2 = .95 Standard error = 16.42 R2 = .99 Standard error = 3.52 Outcome for two different forecasting approaches •1 More accurate model, illustrated by the higher R2 and lower standard error •2 Can highlight problems based on nuanced changes before they have a negative impact on the company •3 Allows executives to see which drivers have the greatest impact on revenue forecasts •4 Enables better and quicker decision-making in order to take action and enhance future performance
- 20. Visualization Visualization • Multi Dimensional reporting, and ability to drill down to details • Real time information • Dashboard for management
- 21. Natural Language Generation Data Audience ANALYZE INFER GENERATE INSIGHTS Natural Language Generation Produces an analysis and text automatically Generates standardized text from the same domain knowledge base Tailors the text to the user’s expertise level and context Drives actionable insights, increases productivity, and operational efficiency
- 22. Natural Language Generation LinktoYoutube
- 23. Integrating NLG with Robotics can significantly automate Decision Support activities, enabling a greater focus on analysis Report Submission Data Collection Data Manipulation Insights Generation Commentary Generation Driver Analysis Report Submission Data Collection Data Manipulation Insights Generation Commentary Generation Driver Analysis Legend: Manual Process RPA Tool NLG Tool Current State Desired State Illustrative Business Partnering support process
- 24. The CFO is the logical position to oversee predictive analytics capabilities and to partner with the business
- 25. Getting Started THINK BIG Immerse Yourself in Innovation Join an immersive experience (e.g., Digital Finance Lab) to explore the “art of the possible” and determine a future state vision, goals, and benefits Build Your Ecosystem Evolve your Finance organisation by collaborating with other business functions,, and vendors ESTABLISH A DIGITAL LEADERSHIP TEAM Identify a visionary program leader and assemble a team to accelerate your digital goals. Determine a governance model and policies that might need to be adapted to execute successful change management and ensure the solution is absorbed into the business. START SMALL Scaling the Edges Disconnect from the core business and set up a digital leadership team to assess disruptive opportunities within the organisation Pick One or Two Plays Prioritise your desired tactics and pick just one or two to get started in order to establish proof of concept ACT FAST Prove it Works (Quickly) Use an agile, iterative piloting approach to move from strategy to prototyping as quickly as possible – “fail fast” and achieve rapid results Market Your Own Success Seek opportunities to share digital experiences with other functions – knowledge share
- 26. Crunch Time Series Crunch Time: Finance in a Digital World Crunch Time, too. CFOs talk off the record about Finance in a digital world Crunch Time IV: Blockchain for Finance – What CFOs Need to Know Crunch Time III: The CFO’s guide to cognitive technology Crunch Time V: Finance 2025 Link toreports: https://www2.deloitte.com/sg/en/pages/finance-transformation/articles/finance-digital- transformation-for-cfos.html
- 27. Contact : kwho@deloitte.com Contact Details