Intelligent Process Automation in Audit Chanyuan Abigail Zhang

Intelligent Process Automation in Audit Chanyuan (Abigail) Zhang Continuous Auditing & Reporting Lab Rutgers Business School American Accounting Association Strategic and Emerging Technologies Workshop 08. 04. 2018 1

Motivation and Objective Robotic Process Automation (RPA) Definabl e Rulebased Scope of RPA Cannot process unstructured data Cannot make complex decisions that require higher level of cognitive judgements Cannot perform further analysis of exceptions Cannot adapt to changes Standar dized Structur ed Cannot learn through time A call for a process automation in audit that is more Intelligent Process Automation scalable, flexible, and intelligent than RPA (IPA) Objective: Establish a framework of implementing IPA in audit to achieve streamlined and end-to-end process automation 2

Major Contributions Establishes a framework of implementing IPA in audit Constructs the “Auditor-in-the-loop” IPA ecosystem Proposes the audit production line driven by IPA 3

Intelligent Process Automation (IPA) IPA is a constellation of technologies that expands RPA by augmenting it with AI, cognitive technologies, and other technologies to make the process automation scalable, flexible, and intelligent. Intelligent Process Automation RPA Artificial Intelligence Machine Learning Computer Vision Virtual Agent NLP/ NLG Others Cognitive Computing Other technologies Data Analytics Drones Smart contracts Io. T Others 4

The “Sense-Think-Act” loop of IPA Think AI and cognitive technologies Computer vision, natural language processing, drones, Io. T, etc Human is in the loop RPA to execute workflows, APIs, computer command scripts, mechanical robotics Sense Act Adapted from Ui. Path (2017) 5

Types of Audit Tasks Type Examples Automatable Non-automatable a. Go to the Morningstar website, search for a particular fund name, copy paste the fund returns to Excel, compare with the return provided by the client Structured • • Repetitive Standardized Rule-based Deal with structured data • No cognitive judgements Semi-structured Unstructured b. Go to the. Semi-structured third-party website, input. Unstructured user name and password, then download • Repetitive • Ad-hoc • Repetitive the confirmation report • Cannot be • Standardized after standardizedin substantive testing cognitiverecalculations, judgements c. Reconciliations, confirmations are made • Cognitive judgements • Complex and fuzzy decision-making procedures documents and choose “yes” or “no” on audit planning a. Read the supporting • deal with semiwork paper May structured or unstructured data b. Go through the pension contract and extract the key term such as eligibility, • Cognitive judgements and then execute the corresponding testing queries. RPA √ a. Fair value estimation × Unstructured IPA × b. Internal control risk assessment √ c. Judgements on certain √ exceptions identified √ × × 6

Framework of Implementing IPA in Audit 1. Is the task structured, semistructured or unstructured? 6. Prototyping and implementing 2. Is the audit task automatable? 3. If it is automatable, can or will it be automated? 5. Is there any need to redesign the process for more efficient automation? 4. If the task will be automated, what potential automation tools that can be used for this task? 7

The “Auditor-in-the-Loop” IPA Ecosystem 8

IPA in Major Phases of Audit. Planning Phase Audit Example Tasks Phase Type of Example Tasks Type of Test. IPA of Controls RPA Tasks Planning the. Structured internal Information Test controls per pre-defined rules Establish planning materiality based on control testing based collection Structured pre-defined rules Process the notable items andon risk send the Semiauditors the most risky items to Test of Fill in the planning worksheet per prestructured Initial analysis Semiinvestigate (if full population are tested) Controls defined criteria based on structured or structured Planning unstructured information Assess likelihood of misstatements Audit plan based on pre-defined rules or based Conducting the tests on Unstructured Read and summarize internalreasoning and formation cognitive Unstructure external information about client’s d business and industry. Type of Audit Phase Example Tasks RPA Reporting IPAof Substantive Testing Type Concluding and Tasks Audit Phase Example Tasks RPA Tasks Standardize the testing data if needed Structured Reconciliation. Complete disclosure checklist on Test key items per pre-defined rules Structured Concluding formatted templates Process and the notable items and send Semi. Substantive the Confirmation auditors the most risky items Conclude andtoissue appropriate type Semi. Reporting structured investigate (if full population arereport tested) of auditor's structured Testing √ √ × Assess likelihood of misstatements Analytical based on pre-defined rules or based procedures on cognitive reasoning √ √ × √ √ IPA √ √ × √ Reporting Unstructured IPA × √

Audit Production Line Driven by IPA 10

Discussion and Considerations Facilitate continuous audit Impacts on audit quality Reliability of IPA solution Cost benefit considerations Stakeholders’ concerns 11

Future Research Empirical research on how IPA can affect audit quality and audit fees Prototypes of the “auditor-in-theloop” IPA ecosystem and the audit production line driven by IPA can be built Empirical research on how auditors will work in the “auditor-in-theloop” IPA ecosystem 12

“Automation is constantly changing and does not stop at RPA. ” -Ui. Path, 2017 Let’s embrace changes! Thank you! 13