565 DB2
Artificial Intelligence in Auditing
H-1
©McGraw-Hill Education
1
learning objectives
Define machine learning and artificial intelligence
Introduce the common use of artificial intelligence
Illustrate Robotic Process Automation
Demonstrate artificial intelligence in auditing
2
PwC | Statistical Programming I
2
Data analysis cycle
3
Acquire data
Analyze data
Present findings
Ask a question
1
2
Transform data
3
4
5
PwC | Statistical Programming I
3
What is data science?
4
Adapted from http://drewconway.com/zia/2013/3/26/the-data-science-venn-diagram
Computer Science
Statistics
Domain Knowledge
Data Science
Danger Zone!
Traditional Research
Machine Learning
PwC | Statistical Programming I
4
Discussion
What have you heard about…
Predictive analytics
Machine learning
Artificial intelligence
What are some examples that use these techniques?
Consumer products
Business solutions
5
PwC | Statistical Programming I
5
Common contributions to machine learning:
Updating e-mail spam filter rules.
Telling newsfeeds (e.g. Flipboard) your preferred content.
Using Google maps and Google translate.
& recommendations based on previous picks.
Alexa, Google Home, Bixby = AI assistants
Data and analytics in industry
7
Trade surveillance
Predictive maintenance
Performance management
Marketing
Spam filtering
Sentiment analysis
Document classification
Facial recognition
Education outcomes
Video search
Recommendation engines
Customer retention
Cybersecurity
Text auto-completion
Regulatory compliance
Customer service
Fraud detection
Translation
Self-driving cars
PwC | Statistical Programming I
7
What is predictive analytics?
Predictive analytics is the systematic computational analysis of historical data to predict unknown values or states of the world
A predictive model is a description of the relationship between one or more variables (X) and another variable (y) that enables us to:
Quantify and determine the significance of effects of X on y
Predict new values of y given new values of X
Assess the quality of the model and resulting predictions
Predictive models fall into two general categories:
Regression models predict values
Classification models predict states
8
| y | X |
| Dependent | Independent |
| Predicted | Predictor |
| Response | Control |
| Explained | Explanatory |
PwC | Statistical Programming I
8
What is machine learning?
Machine learning is a field of computer science relating to algorithms that can recognize patterns and make predictions on data
Unsupervised learning
There is no single dependent variable
The algorithm is used to identify patterns and anomalies according to similarities and differences among observations
Supervised learning
supervised learning ≈ predictive analytics
There is a dependent variable and at least one independent variable
The algorithm is trained on historical observations with known values or states and used to make predictions about new observations
9
PwC | Statistical Programming I
9
Machine learning and AI
In traditional programming, the computer consumes data and generates an output according to a given set of instructions
A machine learning algorithm takes input data and known outputs to learn the program (a statistical model) that should be applied to new inputs
Artificial intelligence (AI) is a field of computer science relating to computer systems that can perform tasks typically requiring human interaction
Machine learning algorithms and very large training data sets often provide the predictive power behind an AI system
10
Traditional Programming:
Machine Learning:
Input + Program Output
Computer
Input + Output Program
Computer
PwC | Statistical Programming I
10
How AI terms are related:
A subset of AI that includes abstruse statistical techniques that enable machines to improve at tasks with experience. The category includes deep learning.
The subset of machine learning composed of algorithms that permit software to train itself to perform tasks, like speech and image recognition, by exposing multilayered neural networks to vast amounts of data.
Any technique that enables computers to mimic human intelligence, using logic, if-then rules, decision trees, and machine learning (including deep learning)
Machine Learning
Artificial Intelligence
Deep Learning
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
Why use AI?
Potential advantages of AI:
Rapid response times
Analysis of large quantities of data
Reduced labor costs
Reserve use of human judgement for hard problems
Lack of bias?
Potential challenges of AI
AI does not have “common sense”
May not respond optimally to new and unusual events
High up front costs
Legal and ethical questions
Inherent bias?
12
PwC | Statistical Programming I
12
Artificial Intelligence is a productivity tool
Accountants have used tools to support their roles for centuries.
Artificial Intelligence = software automation like an Excel macro.
You provide instructions and the software follows through.
Leverage AI as you would a calculator.
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
When will AI affect accountants?
Planned: Bigger, public companies with many transactions have the financial resources and the best case for ROI in regards to automating accounting tasks.
There needs to be good documentation and standardization to base workflows that can translate well to automation.
Unplanned: You may be pushed into using AI in the form of risk management.
Example: AI anti-virus will be used to defend against
AI driven cyber attacks.
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
Introduction to RPA
Robotic Process Automation (RPA): the use of a software robot or “bot” that replicates the actions of a human to execute tasks across multiple computer systems.
A minute of work for a robot is equal to about 15 minutes of work for a human. - (Deloitte)
Robotics is predicted to automate or eliminate up to 40 percent of transactional accounting work by 2020.
- (2015 Accenture report)
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
IMAGE: https://www.alten.com/alten-designs-a-unique-approach-for-effectively-developing-rpa-projects/
15
What can be automated with RPA?
Document Capture Data Entry
3-Way Matching
GL Coding
Vendor Interactions
Payments
Document Review & Approvals
Process Management & Controls
Exception Processing
And more…
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
https://www.processmaker.com/blog/how-do-banks-benefit-from-robotic-process-automation-rpa/
What can Robotic Assistants do?
Control processes
Enforce rules
Automate communications
Provide reminders
Manage resources and escalates
Perform data entry
Collect and present data and documentation
Ask for your expert input for review or approval
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
What can Robotic Assistants do?
Control processes
Enforce rules
Automate communications
Provide reminders
Manage resources and escalates
Perform data entry
Collect and present data and documentation
Ask for your expert input for review or approval
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
18
Automated Data Entry
Invoices
Purchase Requests
Purchase Orders
WEB Docs/E-mail
Eliminate Double Entry
Key-Process Docs
Orders
Special Transactions
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
You can start RPA anywhere…
Purchasing
Accounts Payable
Accounts Receivables
Order Entry
Customer
Service
Operations
Compliance
Human Resources
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
Benefits of Accounting Automation
More Efficient Process Less Labor Fewer Mistakes
Ensure Accountability &
Compliance
Rapid
Implementation
Satisfied Users Do More
With Less
BENEFITS
COST TO PROCESS
$22.75
TIME TO PROCESS
16.3DAYS
EARLY PAYMENT DISCOUNTS
16%
78%
REDUCTION
63%
REDUCTION
298%
INCREASE IN
PAYMENTS
before.
RESULTS
COST TO PROCESS
$5.03
TIME TO PROCESS
6.1DAYS
EARLY PAYMENT DISCOUNTS
47%
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
21
Why prepare for AI?
The benefits of AI are expected to outweigh the cons. AI will continue to grow in nearly all aspects of our lives.
AI is a designed to be a problem solving technology.
The world economy is expected to gain billions in GDP through utilizing AI.
Your career will be impacted by this coming change in AI and other related technologies.
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
How to prepare for AI?
Learn the difference between what AI can do easily and what will be difficult for AI to do.
Develop new skills that will benefit from AI driven data.
Be vigilant to keep your data accurate.
Expect routine, rule-based manual tasks to be the first activities to be automated.
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
Examples of AI related to Accounting & Finance
Risk management
Audits
Compliance and reporting
Forecasting and analytics
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
Example using AI
Today’s Automation with an ERP System
Automation through rules-based work flows by user-defined fields
Automatic sub-ledger reconciliations , like AP Trade
Automatic job-costing
Automatic reporting
FIFO inventory by purchase order for automatic proper COGS posting
Application Programming Interface (API) ready for RPA integration
API ready to connect to other databases to push and pull information
Dave Sackett, Get Ready For Artificial Intelligence, CFO University
https://www.noitechnologies.com/artificial-intelligence-and-erp-a-match-made-with-codes/
25