565 DB2
Auditing (with) Artificial Intelligence
17 mei 2018 Van der Valk Hotel Utrecht
Mona de Boer, PwC
Who had a customer service interaction in the past month?
2
Who is 100% sure the customer service interaction was with a human?
3
4
Exponential Growth and Advanced Technologies
5
x
1450 2000 210019601790 2030
Book printing InternetPersonal computerSteam machine Rise of AI No one knows?!
“AI will become more powerful than a human
brain”*
Open source machine learning
Algorithms improve rapidly by collaboration and joint development
Affordable high performance computing High-performance computers today are thousand times more powerful than they were 15 years ago
Big data Since 2006, the worldwide data volume has increased tenfold
*Ray Kurzweil, Google
Robotic Process Automation – The Next Big Thing
6
Extension to classical RPA approach by using Artificial Intelligence (AI)
Classic rule based robots excel at AI excels at
Algorithmic processing
Repeatable tasks
Workflow interaction
Natural Language
Pattern identification
Locating knowledge
Machine learning
Eliminate biases
Business process automation platforms
Robotic Process Automation (RPA)
Natural Language Processing (NLP)
AI / Cognitive computing
Algorithmic business Human work
+
Almost half of all processes can be automated by
Classic RPA…
…but most of these automated processes have to stop at human interaction. They can be
unleashed by using Cognitive Automation
Robotic Process Automation – Developing the Robo Auditor
7
Algorithmic Business
▪ Industrialized use of complex mathematical algorithms to drive improved business decisions or process automation for competitive differentiation
Robotic Process Automation (RPA)
Alias: Robotic Desktop Automation (RDA)
▪ Automating labor-intensive, repetitive activities across multiple systems and interfaces by training and/or programming third-party software to replicate a user’s workflow
▪ Operates at the presentation layer without the need to change existing systems
▪ Users intervene to handle exceptions as they arise
Business Process Automation (BPA)
▪ Reengineering existing business processes by using software, integrating systems, and restructuring labor to optimize workflows and minimize costs
Intelligent Process Automation (IPA)
Aliases: Cognitive Computing, Smart Workflows
▪ Combining RPA with artificial intelligence technologies to identify patterns, learn over time, and optimize workflows
▪ Through “supervised” and “unsupervised” learning, algorithms make predictions and provide insights on recognized patterns
▪ With IPA, robots can replace manual clicks (RPA), interpret text-heavy communications (natural language processing), make rule-based decisions that don’t have to be pre- programmed (machine learning), and offer customers suggestions (cognitive agents)
Today Future
How do RPA and IPA differ?
RPA directly mimics human behavior
IPA learns how to become more efficient
ProgramInput Output
ProgramInput
Learning
Output
improve systems simple manual activities and learn complex activities automate automate
judgement automate
8
Journal entry testing today…
Soon to be…
Identifying risks at a level that humans can’t
9
The power of combining exploratory and confirmative analytics
10
Intelligence
Knowledge
Information
Big Data
Exploratory
Confirmative
11
Finding outliers in A/R listing today…
A/R listing anomaly detection through Machine Learning
12
Reviewing MD&A today…
MD&A review on steroids with Natural Language
Analysis & Classification
Visual/Image Recognition – a different view on stock count?
13
Digital Auditing Assistant: Auditing Support through Voice Recognition
14
Blurred lines: Who ‘owns’ the algorithms?
15
PwC
Who audits the algorithms?
16
• Every company will become an IT / analytics firm
• In the short term we will be utilizing algorithms in auditing
• In the long run we will be auditing algorithms
Opening AI’s black box
17
18
Bye bye black box…
Researchers teach AI to explain itself
19
20
Be open-minded about your future colleagues…
21
mona.de.boer@pwc.com