ArticlePDF Available

Perspectives on the current and imagined role of artificial intelligence and technology in corporate governance practice and regulation

November 2020

Authors:

Natania Locke

Swinburne University of Technology

Helen Bird

Swinburne University of Technology

This article investigates how AI and related technology already augments global corporate governance practices and the potential for further augmentation as AI continues to expand its capacity to provide cognitive insights and cognitive engagement in corporate board rooms. Itis a scoping study to determine the current uses of these technologies. There is already a baseline of technology that boards have adopted to help manage their corporate governance responsibilities that assists in preparing and distributing reports, streamlining meeting preparation and scheduling meetings. Helpful as this is, it doesn’t address the bigger governance challenges that boards are facing, such as engaging in meaningful strategic, pro-active risk oversight and real-time insights into company operations. Artificial intelligence for strategic and operational decision-making enables boards to make better decisions and is likely to become an essential competitive advantage in its own right. The article imagines how these developments might play out, highlights the challenges involved and suggests their impact on the traditional norms and understandings of contemporary corporate governance.

Content uploaded by Natania Locke

Content may be subject to copyright.

Perspectives on the current and imagined role of artificial intelligence and technology in

corporate governance practice and regulation

Natania Locke and Helen Bird*

I INTRODUCTION

Corporate governance in Australia’s four major banking institutions is under the ray lamp

thanks to the 2018 Royal Commission into Misconduct in the Banking, Superannuation and

Financial Services Industry in Australia

, the 2018 Prudential Inquiry into the Commonwealth

Bank of Australia

and the 2019 commencement of civil penalty proceedings and an associated

prudential inquiry against Westpac Banking Corporation relating to a money laundering

scandal.

These inquiries revealed and continue to reveal all too clearly that company size does

complicate but not excuse board transparency and accountability for governance problems and

difficulties. While much attention has rightly been given to the judgment, shrewdness, acumen

and experience of banking executive and non-executive directors during the period of review,

the resulting analysis tends to underplay the realities of the modern banking boardroom beset

by huge volumes of data that directors must digest to perform their governance functions, data

that is only increasing as their industries are rapidly changing and responding to technological

disruption.

For another revolution, apart from regulatory review, is going on in boardrooms

across the globe. It is the incorporation of technology and artificial intelligence (AI), and the

creation and curation of data that could assist in the use of such AI systems, into the practice

of corporate governance and strategy. This is borne from the realisation that businesses, like

Australia’s major banks, are too complex for their boards and executives to make good

decisions without the aid of intelligent systems.

This article investigates how AI and related technology already augments global corporate

governance practices and the potential for further augmentation as AI continues to expand its

capacity to provide cognitive insights and cognitive engagement in corporate board rooms. It

* Dr Natania Locke, Senior lecturer, Swinburne Law School and Visiting Professor, University of Johannesburg.

Ms Helen Bird, Course Director of Master of Corporate Governance and Research Fellow, Swinburne Law

School. An earlier version of this article was presented at the Comparative Corporate Governance Conference,

hosted by Singapore Management University and the University of Adelaide, 24 – 25 January 2019, Singapore.

The Royal Commission into Misconduct in the Banking, Superannuation and Financial Services Industry was

established on 14 December 2017 by the Governor-General of the Commonwealth of Australia, His

Excellency General the Honourable Sir Peter Cosgrove AK MC (Retd). The Final Report of the Commission

was presented to the Governor-General on 1 February 2019, <https://treasury.gov.au/publication/p2019-fsrc-

final-report/>.

Australian Prudential Regulation Authority, Prudential Inquiry into the Commonwealth Bank of Australia –

Final Report (20 April 2018), https://www.apra.gov.au/sites/default/files/CBA-Prudential-Inquiry_Final-

Report_30042018.pdf.

Pleadings for AUSTRAC v Westpac Banking Corporation, issued in November 2019 are available at

https://www.austrac.gov.au/lists-enforcement-actions-taken. On 17 December 2019, the APRA announced an

investigation into possible breaches of the Banking Act and increased capital requirements for Westpac, see

APRA, ‘APRA Launches Westpac Investigation and Increases Capital Requirement Add-ons to $1 Billion’,

Media Release, 17 December 2019, at https://www.apra.gov.au/news-and-publications/apra-launches-

westpac-investigation-and-increases-capital-requirement-add-ons.

Barry Libert, Megan Beck and Mark Bonchek, ‘AI in the Boardroom: The Next Realm of Corporate

Governance’ (October 2017) MIT Sloan Management Review <https://sloanreview.mit.edu/article/ai-in-the-

boardroom-the-next-realm-of-corporate-governance/>; David Lancefield and Carlo Gagliardi, ‘Reimaging the

Boardroom for an Age of Virtual Reality and AI’ Harvard Business Review (April 2015)

<https://hbr.org/2015/04/reimagining-the-boardroom-for-an-age-of-virtual-reality-and-ai>.

is a scoping study to determine the current uses of these technologies

. There is already a

baseline of technology that boards have adopted to help manage their corporate governance

responsibilities that assists in preparing and distributing reports, streamlining meeting

preparation and scheduling meetings. Helpful as this is, it doesn’t address the bigger

governance challenges that boards are facing, such as engaging in meaningful strategic, pro-

active risk oversight and real-time insights into company operations. Artificial intelligence for

strategic and operational decision-making enables boards to make better decisions and is likely

to become an essential competitive advantage in its own right. The article imagines how these

developments might play out, highlights the challenges involved and suggests their impact on

the traditional norms and understandings of contemporary corporate governance.

All of the available technology set out below may be used in the boardroom to augment the

natural capability of directors to fulfil their duties. We do not argue here that AI will supplant

the role of directors in corporations any time soon but rather that the insight offered by AI will

inevitably improve analytical accuracy and resulting efficiency of the governance task. The

combined intellectual capability of a group of actors is referred to as ‘collaborative

intelligence’.

It was predicted as early as 1993 that so-called ‘intelligence amplification’ might offer a

different road into super-human intelligence than sole reliance on the development of artificial

intelligence alone.

The premise is that any human cognitive behaviour is already amplified or

augmented through the use of even basic technology,

but that the rapid development of

technology would elevate this amplification to an extent never previously possible. We assume

that the collaborative intelligence of the human actors on the board of directors and the AI

systems acting in support and augmentation of those actors far surpasses the intelligence of the

board in absence of those systems.

However, we acknowledge from the outset that the availability of this intelligence and

technology does not necessarily mean it will lead to better corporate governance. AI and

technology assists, but does not supplant, human actors performing governance functions,

aspects of governance decision-making, such as ethics and culture, remain outside AI’s sphere

Others describe it as a ‘a thought experiment’ on corporate management and AI: See Martin Petrin,

‘Corporate management in the Age of AI’UCL Working Paper Series (No. 3/2019), at

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3346722##

See in general H James Wilson and Paul R Daugherty, ‘Collaborative Intelligence: Humans and AI are Joining

Forces’, Harvard Business Review (online), July-August 2018, < https://hbr.org/2018/07/collaborative-

intelligence-humans-and-ai-are-joining-forces?referral=03758&cm_vc=rr_item_page.top_right>. The study

of the collaborative or collective intelligence of human actors forms a distinct field of study in social

psychology and falls outside the scope if the present paper. For further reading, see for instance Anita Williams

Woolley, Christopher F Chabris, Alex Pentland, Nada Hashmi and Thomas W Malone, ‘Evidence for a

Collective Intelligence Factor in the Performance of Human Groups’ (2010) 130 Science 686; Michael A

Woodley and Edward Bell, ‘Is Collective Intelligence (Mostly) the General Factor of Personality? A Comment

on Woolley, Chabris, Pentland, Hashmi and Malone (2010)’ (2011) 39 Intelligence 79; Anita Williams

Woolley, Ishani Aggarwal and Thomas W Malone, ‘Collective Intelligence and Group Performance’ (2015)

25 Current Directions in Psychological Science 420; Timothy C Bates and Shivani Gupta, ‘Smart Groups of

Smart People: Evidence of IQ as the Origin of Collective Intelligence in the Performance of Human Groups’

(2017) 60 Intelligence 46.

Vernor Vinge, ‘The Coming Technological Singularity: How to Survive in the Post-Human Era’ in NASA

Lewis Research Center, Vision-21: Interdisciplinary Science and Engineering in the Era of Cyberspace,

(NASA Conference Publication 10129, 1993) 11 17 available at

<https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19940022855.pdf>.

Vinge (n 5) 17 uses the example of a PhD student and his desktop computer, which is so much more

intellectually powerful than the PhD student alone. Link the desktop to the internet and the amplification is

immediately extended many times over.

of influence with the result that human actors remain accountable for governance outcomes in

the final instance.

The article is divided into VI parts, of which the introduction and conclusion form the 1st and

6th parts of the paper. Part II describes the AI framework used and outlines the key AI concepts

used. Part III investigates current applications of AI and supporting technology in the corporate

governance space with reference to the AI framework set out in Part II. Part IV speculates

about the future advances that AI might bring in the corporate governance sphere, some of

which are already being explored in current systems. Part V considers the challenges that

current and foreseen applications of AI in the boardroom hold for directors in the fulfilment of

their duties and corporate governance practices generally and explores how these challenges

might influence future corporate governance theory and practice.

II AI FRAMEWORK FOR ANALYSING CORPORATE GOVERNANCE

An examination of artificial intelligence (‘AI’) applications in corporate governance invites a

preliminary inquiry about the nature of AI and the systems and technologies that underpin it.

This is a task easier said than done. Definitions are problematic, characterised by the loose

application of terms associated with AI (machine learning,

neural networks,

deep learning,

natural language processing (NLP),

rule-based expert systems,

robotics

) as synonyms,

when their actual functionality is far more narrowly defined.

Technology supported by AI tends to be analysed in silos, without any attempt to systemise

the underlying contribution made by AI in each case. While there are a variety of explanatory

frameworks that might be referenced for this purpose,

we adopt the systemisation articulated

Thomas H Davenport, The AI Advantage: How to Put the Artificial Intelligence Revolution to Work, (MIT

Press, 2018) ch 1 defines ‘statistical machine learning’ as ‘a technique for automatically fitting models to data

and to “learn” by training models with data’. See further Nick Bostrom, Superintelligence: Paths Dangers,

Strategies, (Oxford University Press, 2014) 8 – 18.

This is a form of machine learning that models that functioning of the human brain. It conceptualises problems

as a set of inputs and outputs and uses artificial ‘neurons’ to assign a weight to the inputs that would lead to

particular outputs. See Davenport (n 7) and Bostrom (n 7) 7 – 8.

Deep learning is a more complex form of machine learning using neural networks. It uses the backpropagation

algorithm to make it possible to train multi-layered neural networks. These layers lie in between the input and

output layers, which make it very difficult for human actors to interpret. This technique is used increasingly

to enable the system to predict or classify outputs. See Davenport (n 7) and Bostrom (n 7) 8.

This refers to the application of AI in speech recognition, text analysis, translation, the generation of text or

speech and other goals related to language (Davenport, n 7). Understanding language is a task so difficult that

many are of the opinion it would only be completely mastered by AI at the time when AI has general human

capability. See Bostrom (n 7) 14.

Bostrom (n 7) 7 defines rules-based expert systems as ‘rule-based programs that [makes] simple inferences

from a knowledge base of facts, which had been elicited from human domain experts and painstakingly hand-

coded in formal language’. Expert systems were the earliest form of AI. These systems were hard to maintain

and characterised as ‘brittle’ – even small errors in assumptions could leave it useless.

Patrick Lin, Keith Abney and George Bekey, ‘Robot Ethics: Mapping the Issues for a Mechanised World’

(2011) 175 Artificial Intelligence 943 define a robot as ‘an engineered machine that senses, thinks and acts’

and it must have the capability to do so autonomously.

For instance, Ajay Agrawal, Joshua Gans and Avi Goldfarb, Prediction Machines: The Simple Economics of

Artificial Intelligence, (Harvard Business Review Press, 2018) offer a systemisation from an economics

perspective where the prediction capability of AI is the driver of all other layers of influence. For their

purposes, ‘prediction’ is ‘the process of filling in missing information. Prediction takes information you have,

often called ‘data’, and uses it to generate information you do not have’ (ch 3).

by leading US business AI authors, Davenport and Ronanki, to describe the current and

potential roles that AI play and is likely to play in the boardroom in the near future.

Davenport and Ronanki describe AI according to its business capabilities rather than the

technology underlying it. They identify three specific classes of AI in a business context:

process automation; cognitive insight; and cognitive engagement. Each class reflects the

evolution of cognitive capacity in AI, from lowest to the highest in cognitive engagement.

However, AI enabled systems may show elements of all three classes in various degrees.

Process automation refers to those systems where AI is used to automate digital and physical

tasks, very often administrative or financial tasks.

Robotic process automation technology

can access various IT systems much like a human actor would do. This is the area where many

businesses have started incorporating AI technology, since it is relatively easy to do and often

leads to immediate bottom-line benefit in that the automated tasks are often those already

outsourced. Second stage AI, cognitive insight, is offered by those systems that gain predictive

ability through big data analysis. They detect patterns in data and interpret their meaning

through machine learning.

This is the second most used application of AI technology. Third

stage AI, cognitive engagement, refers to systems that use natural language processing to

interact directly with customers or employees of an organisation.

Examples include chatbots

answering basic call centre queries from customers or employees and cognitive assistants such

as Siri, Alex and Echo that anticipate instructions and provide answers to questions.

The

engagement may be further aided by augmented reality visualisations of business activities that

create the impression of being inside an activity in real-time. While all three classes of AI are

embedded in current corporate governance technologies, their impact is uneven. As will be

seen, process automation has had the most immediate impact on governance because it

facilitates the product and distribution of more reliable data for the purposes of internal auditing

and risk management and in turn governance oversight of those functions.

III CURRENT APPLICATIONS OF AI IN CORPORATE GOVERNANCE

A. Introduction

Current boardroom technologies focus on the production and distribution of information that

boards need to assist in their supervisory and strategic roles.

Many of these technologies do

not offer the advantages of AI systems themselves but instead generate the data which is the

vital lifeblood that AI needs. Information systems that generate and curate data and AI systems

must therefore be discussed simultaneously. Of these, board portals and risk reporting systems

are the most widely used, tied in with systems that assist in legal compliance and internal

auditing. Each of these technologies is at present an application of Davenport and Ronanki’s

class one AI, process automation. While there has been progress in using cognitive insights in

Thomas H Davenport and Rajeev Ronanki, ‘Artificial Intelligence for the Real World’, Harvard Business

Review (online), January-February 2018, https://hbr.org/2018/01/artificial-intelligence-for-the-real-world.

See further Davenport (n 7).

Davenport and Ronanki (n 14).

Ibid. This is the form that algorithmic trading takes.

Ibid.

Tom Davenport and Rajeev Ronanki, ‘The rise of cognitive agents’ (2018) <

https://www2.deloitte.com/insights/us/en/focus/cognitive-technologies/rise-of-cognitive-agents-artificial-

intelligence-applications.html>.

Financial Reporting Council (UK), Digital Future of Corporate Reporting (December 2017) 8

<https://www.frc.org.uk/getattachment/9279091c-a4e9-4389-bdd6-

d8dc5563b14a/DigFutureXBRLDec.pdf>.

the boardroom, they are not yet widely in use. However, the future potential of cognitive

insights and cognitive engagement technologies in the governance of corporations is immense.

B. Board portals

1. Baseline technology & AI extensions

Board portals are critical to all classes of AI. They are both the chief mode by which digital

information is distributed to boards and the platform by which boards are accessing the rapidly

expanding range of cognitive technologies.

Originally developed as a means of automating

the secure preparation and distribution to directors of the board pack or board book, the advent

of board portals kick-started the shift from paper-based board preparation to an all-digital

environment.

Along with video and web-based conferencing, portals also worked to remove

the constraints against remote participation in meetings by directors.

Algorithms help collect, collate, format, distribute and regularly update papers and reports from

officers such as the CEO, CFO, COO etc and various divisions of a corporation that form part

of the board pack periodically required across the board reporting cycle so that boards can

analyse the state of the business at regular intervals. This type of handling and management is

considered to be perfectly suited to AI optimisation because it can be configured to the needs

of the user/specific company and yet, leveraged individually by board members who can access

the server/cloud on which it is stored.

The data must be regularly updated of course, but the

underlying algorithms have generally low levels of variation, facilitating time and cost

efficiencies.

Critical to these transformations has been the ability of portal technology to

provide secure access and control of regularly updated, sensitive information to board

members. This extended to automating the archiving and digital shredding of documents.

More recent evolutions have extended portal functionality beyond that of a digital data

depository to also include AI class 2 or cognitive insight applications including note-taking

and data review functionality; interactivity between board members and immediacy of data

available to directors. The effect has been to turn portals into AI platforms for board

collaboration; data analysis and interactions, thereby enhancing board effectiveness.

Depending on the portal provider, tools may include scheduling, alerts, intra-board messaging,

surveys, e-signatures, voting capabilities, multi-language capability, reading rooms, document

libraries, search functionality and secure support for the use of portals in multiple environments

accessed through a range of devices including iPads and tablets, web-based tools and

smartphones.

Raphael Goldsworthy, Tech in the Boardroom: Beyond the Board Portal (9 May 2016) <

https://betterboards.net/governance/tech-boardroom-board-portal/.

Corporate Secretary, Trends in Board Portal Adoption – Special Edition (1 June 2017) 4 <http;//www.

Corporatesecretary.com/node/30644>.

John Cormican and Luke Phillips, ‘Board Portals – evolution in board communication’ Keeping Good

Companies (May 2011) 207.

David Gardner, ‘Is General Counsel Ready for AI Adoption?’ (11 July 2018) <

https://ct.wolterskluwer.com/resource-center/articles/is-general-counsel-ready-for-ai-adoption>.

Shared Services & Outsourcing Network and SSON Analytics, Global Intelligent Automation Market Report

(H1 2017) (2017) < https://www.sharedservicesweek.com/downloads/global-intelligent-automation-market-

report-h1-2017-part-1 10. The technology that underlies these systems is typically rules-based and training is

supervised.

Corporate Secretary (n 21).

Corporate Secretary (n 21); Steve Cocheo, ‘Should you “plug in” your board of directors?’ (2007) 99 American

Bankers Association Banking Journal 40; Cormican and Phillips (n 22) 212. See also the discussion below par

III.E. on how the Internet of Thing is assisting in this task.

Corporate Secretary (n 21) 4.

2. Factors affecting adoption and wide use

Despite the absence of a published academic study of the adoption of board portals in various

jurisdictions, privately commissioned surveys provide some insight into the adoption of portal

technology. First, a global survey of 400 governance professionals undertaken by Forrester

Consulting and commissioned by board portal provider, Diligent, revealed that portal adoption

is high across the world, with the highest rate of adoptions in the Asia Pacific,

followed by

Europe and surprisingly, the lowest rate of adoption in North America.

Secondly, a similar

study by Azoth Analytics

determined that the global leaders in board portal production

include Diligent Corporation,

Nasdaq Boardvantage,

ComputerShare,

Directorpoint,

BoardPaq;

Convene

and BoardEffect.

Of these, Diligent is the market leader, with its

website claiming to provide board portals to more than 145,000 corporate officers globally.

Thirdly, based on a recent survey by Corporate Secretary of 300 US executives with

responsibility for preparing board materials for their organisations, board portals are being

adopted by organisations of all sizes, from both the public and private sector and profit and

non-profit enterprises with approximately 32% of such enterprises receiving digital only board

packs.

Implicit in each of these findings is the assumption that there were no barriers to full

digital adoption in the organisations surveyed such that board buy-in to the adoption of

technology was the same across the survey cohorts.

The foregoing discussion suggests that portals are a baseline technology that has been widely

adopted to help manage corporate governance responsibilities. However, despite their

potential, there is a long way to go before boards view portals as more than digital board

meeting packs.

This is a function of two issues. First, the slow integration of AI into corporate

governance due to significant variations in the technology features of portal products. For

example, not all portals presently offer cognitive insight or engagement technology such as

chat functionality or virtual dealing rooms for board interactions and cognitive engagement.

Secondly, even where such functionality exists, boards are cautious about using them, being

more comfortable with nascent technology practices such as their own personal email in

preference to board portal enabled closed-loop messaging for directors.

This must change,

the question is when. An issue likely to further galvanise digital governance is cyber security,

Note that 7% of respondent companies of the total survey were based in Australia, but Australian companies

constituted only slightly over 20% of the Asian Pacific contingent. See Forrester Consulting, ‘Directors’

Digital Divide: Practices Boardroom Aren’t Keeping Pace with Technology’ (October 2018), <

https://diligent.com/wp-content/uploads/sites/5/2018/11/UK-Diligent-Global-Report-Forrester-Directors-

Digital-Divide-Boardroom-Practices.pdf > 10.

Forrester Consulting (n 28) 7.

Azoth Analytics, Global Board Portal Market: Insights, Analysis and Forecast to 2023 - By Value, By

Number of Users, By Model, Penetration, End-User Industry, Company Analysis (2018)

www.researchandmarkets.com/reports/4585755/global-board-portal-market-2018-edition#pos>.

https://diligent.com/au/.

https://nq.nasdaq.com/Boardvantage-Reimagining-Governance?channel=PPC&source=Google&ppc-

campaignid=1533189567&gclid=EAIaIQobChMIpO7qq4rn3gIVlxOPCh2g7w0REAAYASAAEgKnavD_B

https://www.computershare.com/au/business/corporate-governance/stay-connected-to-the-boardroom.

http://directorpoint.com/.

https://www.boardpac.co/?gclid=EAIaIQobChMIgOy-h4vn3gIVS4aPCh30zw-tEAAYAiAAEgLjsfD_BwE.

https://www.azeusconvene.com/.

https://boardeffect.com.

https://diligent.com/au/.

Corporate Secretary (n 21) 2 and 6-7.

Forrester Consulting (n 28) 5.

Ibid 2. Also see the discussion in part III.E below.

Forrester Consulting (n 28) 2.

specifically the vulnerability of boards and their organisations to cyber-attack and the

repercussions of sensitive materials or matters being exposed to the public.

AI applications

will gain greater traction when both their cognitive and security benefits beyond the boardroom

are better understood.

C Risk reporting and Internal Auditing Systems

1. Outline

AI technologies that model and stress test risk management practices and internal audit

processes and controls within organisations are an emergent field gaining traction in the

financial services sector, particularly in banking organisations.

Part of an AI wave that first

focussed on customer-facing ‘journeys’ and the ‘front of house’ operations that support them,

the transformation of ‘back of house’ operations such as risk and internal audit, has been more

cautious and uneven. Tempered by concerns as to the untried and untested nature of AI in

helping firms to satisfy their legal obligations,

despite the obvious benefits of streamlining

the automation of monotonous tasks and the reallocation of budgeted times involved in those

tasks to more strategic concerns.

However, the rapid growth of digital customer-facing

activities is forcing the issue, with industry consultants arguing that risk and audit functions

within financial institutions must be digitalised or create bottlenecks for their front of house

operations.

In this section, we examine these trends then discuss how AI is helping address

governance challenges that boards are facing in the risk and audit responsibility fields.

2. Risk Management and Risk Oversight

AI already provides several building blocks that make possible the digitalization of a firm’s

risk function. These include: data management; process and workflow automation; advanced

analytics and decision automation; and smart visualizations.

Ever expanding commercial

versions of these technologies are available in multiple forms online from the big four

accounting firms;

the big three management consultants;

legal compliance and risk

management consultants;

and governance and entity management consultants.

These

products still predominantly depend or draw from the first class of AI, namely process

automation and best assist the function of risk management rather than risk governance. AI

enabled advanced analytics arguably illustrate the expansion of digital risk technology into

Davenport and Ronanki’s second AI class, cognitive insight because, as will be seen, it makes

possible more precise and in-depth analytics than were previously possible.

Decision

Ibid 3.

See for e.g. Deloitte EMEA Centre for Regulatory Strategy, AI and Risk Management – Innovating with

Confidence (April 2018) 1-2 <https://www2.deloitte.com/au/en/pages/financial-services/articles/ai-risk-

management.html; Jeanne Boillet, EY, “Why AI is both a risk and a way to manage risk” (1 April 2018) <

https://www.ey.com/en_gl/assurance/why-ai-is-both-a-risk-and-a-way-to-manage-risk>; McKinsey &

Company, “The future of risk management in the digital era” (October 2017)

<https://www.mckinsey.com/business-functions/risk/our-insights/the-future-of-risk-management-in-the-

digital-era>.

E.g. online marketing, customer onboarding and servicing. See McKinsey & Company (n 44) 8.

Deloitte EMEA Centre for Regulatory Strategy (n 44) 1.

Pascal Bizarro and Margaret Dorian, ‘Artificial Intelligence: The Future of Auditing’ (October 2017) Internal

Auditing 21.

McKinsey & Company (n 44) 17-21; Deloitte EMEA Centre for Regulatory Strategy (n 44) 1-2.

McKinsey & Company (n 44) 12-14.

Deloittes, EY, KPMG and Price Waterhouse.

McKinseys, Boston Consulting and Bain & Co.

E.g. SAI Global and Lexis Nexus.

E.g. Diligent and Nasdaq.

Davenport and Ronanki (n 14).

automation, aided by smart visualisations and the use of augmented reality in the performance

of a firm’s risk function, are still emergent technologies but would seem to fall within

Davenport and Ronanki’s third AI class of cognitive engagement.

3. Benefits of AI in risk & audit space

AI enables the faster capture and use of vast amounts of structured and unstructured data

(emails, text, social media posts, clickstreams, chat transcripts). AI algorithms permit linkage

analysis, pattern recognition and NLP,

which in turn makes possible more timely and precise

profiling of customer attributes and risks. This is supported by a wider variety and greater

quality of data available so that a firm’s risk function need no longer rely on traditional risk

data nor, in the case of financial service firms, burden customers with requests to supply data

that can be captured by other means. An example of a growing data source or real time data

on banking customers is the Internet of Things,

which constructs data networks using built-

in sensors on digital machines and appliances (fixed, mobile, wearable) to gather and combine

infinite sets of individual data from those machines and facilitates big data analysis.

Process and workflow automation facilitate streamlining, standardisation and the efficient

execution of routine tasks such as data entry and collection. Using robotic process automation

(RPA), NLP and optical character recognition, tasks can in turn be combined into integrated

sequences or smart workflows. The use of RPA and NLP is still in its very early stages but is

rapidly catching on in the financial services sector.

Blockchain is also expected to add further

potential for process automation in a few years.

It is claimed that automated workflows offer

the potential for increased precision in risk modelling, improved regulatory compliance and

significant cost reductions depending on a firm’s risk types.

Advanced analytics based on AI algorithms,

are able to identify more complex patterns than

is possible by human intervention, particularly in the context of identifying fraud and money

laundering in the financial services context.

Semantic analysis, a branch of NLP, can be

applied to news to extract market sentiment data on business entities. Further, AI analytics are

rapidly advancing with new applications arriving constantly. Decision automation is also

possible, through the use of models and algorithms. For example, advanced analytics are able

to make risk predictions, suggest optimal actions and extract relevant insights. It has been

suggested that these decisions can increase accuracy of risk management in many cases.

least in the financial services context, these decisions are not, as yet, being used to automate

actual risk decisions, but rather to improve the accuracy and precision of existing risk models.

AI smart visualisation technology allows users to access AI generated insights and data in a

more intuitive and customer friendly way. Key technologies include dashboards and portals,

Deepak Amirtha Raj, ‘Spotlight on the Remarkable Potential of AI in KYC’ (13 June 2017)

<https://medium.com/all-technology-feeds/spotlight-on-the-remarkable-potential-of-ai-in-kyc-

7441bf7eec38>.

‘Internet of Things’ refers to a global network of machines and devices that are enabled to gather information

from their environment or users and can then communicate and interact with each other. See In Lee and

Kyoochun Lee, ‘The Internet of Things (IoT): Applications, Investments, and Challenges for Enterprises’

(2015) 58 Business Horizons 431.

See, for e.g., Siemen’s MindSphere,

https://www.siemens.com/global/en/home/products/software/mindsphere.html.

McKinsey & Company (n 44) 46.

Ibid 46.

McKinsey & Company (n 44) 12.

Ibid 47.

Raj (n 55).

Rajdeep Dash, Andreas Kremer, Luis Natio and Derek Walton, ‘Risk analytics enters its prime’ (June 2017)

< https://www.mckinsey.com/business-functions/risk/our-insights/risk-analytics-enters-its-prime.

augmented reality and cognitive agents, such as Siri and Alexa.

Presently these technologies

are customer focussed, helping them in the banking context to better understand their spending

habits and financial capacity. However, the same technologies can also be used to improve,

streamline and fasten decision-making. Critical to this capacity is the ability of AI to perform

ongoing multi-dimensional analysis on consolidated data.

4. Impact on C-Suite and Boardroom

AI automated risk management has the potential to improve the functionality of risk and risk

related executives, boards and board risk committees in two important ways.

First, for risk

management and the C-Suite, AI automates away routine risk functions and reduces the number

of manually handled exceptions. This leaves the C-Suite and the risk executives in particular

to focus on strategic and high value decisions using AI driven advanced analytics, as now

described, to generate insights such as more complex correlations and trend analyses than were

previously available and in turn to optimise risk and other management decisions. When added

to improved connectivity, executives will be able to detect emergent risks immediately and set

cross-risk mitigation strategies and limits more dynamically than before.

Secondly, risk reports and strategic advice on risk-oriented business decisions will be generated

automatically and/or be sent on demand to the C-Suite and/or board of directors, for risk

oversight and review. As noted earlier, strategic advice can take the form of risk predictions

and suggested optimal actions, which the board or board risk committee can use as a decision-

making point of reference, a valuable cognitive insight. Risk reports and advice need no longer

be static, nor necessarily require a risk register, although boards still need to have an

appreciation of their risk appetite and an understanding of the firm’s digital risk frameworks.

Boards will increasingly be presented with intuitive visuals that provide summary information

with appropriate levels of detail (as to market, portfolios and products) but which enable the

reader to drill down into the headline figures as required and run further analysis virtually using

real time data. Instead of focussing on the oversight of risks ex-post facto it will increasingly

be possible to oversee risk more strategically, by profiling emerging risks ex ante using AI’s

precise, preventative control mechanisms and reporting features.

The resulting governance

outcomes depend on the persuasive but as yet untested assertion that improved governance

capacity resulting from improved analytics will lead to improved governance per sè, better

performance and improved business value.

These arguments may yet been shown to be

particularly strong in board governance in a crisis, when governance would clearly benefit from

access to real-time and precise reporting.

5. Internal Audit Systems

As technology continues to change the working landscape of the business world, AI impacts

on the nature of the workloads and deliverables requiring core assurance. Adopting digital

risk-based technology, for example, raises security, data quality and methodology, regulatory

and supervisory challenges, both for a firm’s audit teams and for audit governance. Keeping a

check on those challenges across an organisation as a whole is the responsibility of internal

audit.

It involves engagement in ongoing audits of a firm’s systems and operations across all

McKinsey & Company (n 44) 50-52.

Ibid 52. Applications such as Tableau, Alteryx and Qlikview are mentioned in this regard.

Ibid 27-29.

McKinsey & Company (n 44) 27-29.

Ibid 28.

Forrester (n 28) 5-8.

The Institute of Internal Auditors Australia, ‘What is an Internal Audit?” <https//www.iia.org.au/about-iia-

australia/WhatIsInternalAudit.aspx>.

areas of the firm to identify how well risks are managed, whether the right processes, controls

and authorisation procedures are in place and in turn, whether those processes and controls are

being implemented and/or followed.

Audit results are translated into reports providing advice

and assurance regarding the organisation’s risk management, governance and controls to

boards, board committees and senior management and facilitating co-ordination between the

first and second lines of defence involved in the company’s risk management framework.

The use of computer assisted audit software in conducting internal audits is already an

established practice that has brought cost and time savings to a traditionally manually intensive

exercise hindered by timely available data.

However, the adoption of AI to define and

automatically pilot an internal audit end to end is less well-established.

The human factor,

encapsulating knowledge, creativity and professional judgment, is still regarded as an

important control/authorisation procedure for ensuring audit outcomes are reliable.

Implementation of AI tools to transform audit service delivery is the focus of the Big Four

accounting firms.

Critical audit tasks that lend themselves to automation via AI include

analytical review procedures, classification, internal controls evaluation, risk assessment and

going concern decisions.

AI already outperforms humans when streamlining and automating

data acquisition for audits and converting that data into report formats.

Human error is

reduced and time can be allocated to higher order tasks including the leveraging of digital audit

data to identify irregularities and improve weaknesses.

These aspects fall under Davenport

and Ronanki’s process automation category.

The advanced analytics within AI technologies can also compare, contrast and summarise huge

data collections and provide greater support for audit findings through reliance on accurate and

real time data rather than audit sampling.

Furthermore, the machine learning capability of the

The Institute of Internal Auditors Global, Global Perspectives and Insights – The IIA’s Artificial Intelligence

Auditing Framework (September 2017) 2, <https://na.theiia.org/periodicals/Pages/Global-Perspectives-and-

Insights.aspx>.

Paul Holland, Shamus Rae and Paul Taylor, ‘Why AI must be included in audits’ (June 2018)

<https://assets.kpmg.com/content/dam/kpmg/uk/pdf/2018/06/why-ai-must-be-included-in-audits.PDF>.

He Li, Jun Dai, Tatiana Gershberg and Miklos Vasarhelyi, ‘Understanding Usage and Value of Audit

Analytics for Internal Auditors: An Organisational Approach’ (2018) 28 International Journal of Accounting

Information Systems 59; Nurmazilah Mazhan and Andy Lymer, ‘Examining the Adoption of Computer-

assisted Audit Tools and Techniques’ (2014) 29 Managerial Auditing Journal 327; 28 Managerial Auditing

Journal 88; Muhammad Razi and Haider Madani, ‘An Analysis of Attributes that Impact Adoption of Audit

Software’ (2013) 21 International Journal of Accounting & Information Management 170; Aidi Ahmi and

Simon Kent, ‘The Utilisation of Generalized Audit Software (GAS) by External Audits (2012) 28 Managerial

Auditing Journal 88; Fatima Alali and Fang Pan ‘Use of Audit Software: Review and Survey’ (2011) 26

Internal Auditing 29.

Holland, Rae and Taylor (n 72).

Bizarro and Dorian (n 47) 21.

Ibid 24-35. Deloitte is considered the current market leader. The firm leverages a range of AI technologies

such as Argus, which learns from human interaction and machine techniques to extract key accounting data

from electronic documentation including sales; leasing; derivative contracts; employment agreements;

invoices; client meeting minutes; legal letters and financial statements. See Deloitte, ‘Advancing audit quality

through smarter audits’ (2018) <https://www2.deloitte.com/us/en/pages/audit/articles/smarter-audits.html>.

Syed Moudud-UI-Huq, ‘The Role of Artificial Intelligence in the Development of Account Systems: A

Review” (2014) 13 Journal of Accounting Research & Audit Practices 7.

Ibid.

Bill Brennan, Mike Baccala, Mike Flynn and 10Rule, ‘Artificial Intelligence Comes to Financial Statement

Audits’, CFO Magazine (February 2017) <http://ww2.cfo.com/auditing/2017/02/artificial-intelligence-

audits/>.

Michael P Cangemi and Patrick Taylor, ‘Harnessing Artificial Intelligence to Deliver Real-Time Intelligence

and Business Process Improvements’ EDPACS (online), 27 April 2018, 3

<https://www.tandfonline.com/doi/abs/10.1080/07366981.2018.1444007>.

systems enable them to benchmark across different organisations to, for instance, pick up on

the characteristics of extraordinary transactions. This capability is enhanced when they are run

on the cloud.

These capabilities extend the AI functionality into cognitive insight. The major

advantages still, however, remain cost and efficiency. The risks lie in ensuring that AI has been

effectively tested to ensure that results are accurate and reflect the objectives of internal audits.

Fully AI automated internal audits are the foreseeable future but are not yet here. They have

the potential to improve the functionality of the audit process, risk management at all levels of

an organisation and risk governance by the board. As with risk governance, AI driven audit

reports make audit governance by the board or board committee more efficient and create new

opportunities for strategic oversight of business activities in place of monotonous tasks.

However, as with all technologies, what remains unclear is the extent of tailoring, integrating,

testing and monitoring required to ensure the benefits outweigh the resources devoted to its

use.

D Legal compliance systems

Integral to risk and audit governance is a firm’s commitment to operating their businesses in a

legally and ethically responsible manner, in the long term interests of their shareholders.

Legal compliance issues are increasingly sophisticated, requiring not only that companies meet

the requirements of relevant laws but also to identify and self-report possible breaches of the

laws to the appointed regulator in a time-conscious way.

The sheer volume of business

transactions, regulatory requirements and breach reporting obligations have inevitably led to

big data, AI (also known as RegTech

) and blockchain-based solutions being developed. It

has caused regulators globally to create initiatives such as regulatory sandboxes to engage with

and respond to the emerging solutions.

The focus here is however confined to the issue of

how AI facilitates and reports on legal compliance to company boards and in turn how boards

are able to make better business decisions because of that AI functionality.

AI legal compliance techniques have been used by financial service firms in areas such as credit

card fraud and analysing data sets to detect Anti-Money Laundering (AML). AI adoption in

these fields has focused on automating costly manual work, leaving key decisions to human

beings or rules-based systems. AI is proving particularly effective because fraud and money

laundering detection typically involve large numbers of documents and repetitive processes

that suit process automation. Whilst automation in some firms is as much as 90% of their

processes, it is not the only digital answer anymore and others, such as blockchain, may

increasingly become more financially justified.

Ibid.

Royal Commission into Misconduct in the Banking, Superannuation and Financial Services Industry (Interim

Report, September 2018) 191. See further Royal Commission into Misconduct in the Banking, Superannuation

and Financial Services Industry (Final Report, February 2019) vol 1, 12: ‘The conduct identified and criticised

in the Commission’s Interim Report and in this Report has been of a nature and extent that shows that the law

has not been obeyed, and has not been enforced effectively. It also points to deficiencies of culture, governance

and risk management within entities. Too often, entities have paid too little attention to issues of regulatory,

compliance and conduct risks.’

See for example, the breach reporting duty of Australian Financial Service Licence holders in Corporations

Act 2001 (Cth) s 912D (‘Corporations Act’).

Chartis Research, Demystifying AI for Risk and Compliance (2018) 2 <https://www.ibm.com/blogs/insights-

on-business/banking/demystifying-ai-risk-compliance/>.

Global Financial Innovation Network (GFIN), Consultation document (August 2018) 4

<https://www.fca.org.uk/publication/consultation/gfin-consultation-document.pdf>.

Ibid 6.

AI takes process automation protocols and facilitates another layer of cognitive insight through

the use of text analytics and insights to identifying relevant content, negative news, case notes

from structured and unstructured data like emails, posts and chat room conversations. It also

makes possible network analytics, of the kind described in relation to risk management

earlier,

that facilitate the identification of connections between individuals in order to

evaluate risky parties and networks.

AI can also be used to facilitate legal compliance. NLP and RPA are the most commonly used

AI process for compliance. An example is the automatic generation of written reports or

notifications to a regulator regarding cash deposits required by money laundering laws.

Scenario comparisons and stress testing can be tailored to banking regulation requirements in

different jurisdictions. Finally, in something of a ‘holy grail’ for AI,

AI can establish and

automate a system of internal notifications of legal changes throughout an organisation, a

significant issue for financial institutions who have to integrate content from thousands of

regulatory publications each month.

Data integrity remains the biggest challenge for AI-based legal compliance solutions.

Advanced analytics through AI depend on data quality and reliability. We return to this issue

below for further consideration.

AI requires proper data governance procedures and the

validation or assurance processes previously discussed regarding internal audits.

This is

particularly the case when converting documents (essentially unstructured) into structured

formats, while preserving the information they contain. The goal is to obtain segmented,

cleaned and structured data sets that could then feed into companies’ compliance and control

activities.

Boards receive reports on these issues as part of their digital pack that is the primary

functionality of board portals. However, in the case of large corporations, board criticism may

not be so much concerned with accuracy and precision of compliance data but rather, its

relevance. In the Prudential Inquiry Report into the Commonwealth Bank of Australia (CBA),

one of the complaints concerned the failure of CBA board to be alarmed by and follow up on

breaches of money laundering laws despite warnings in internal audit reports to the relevant

board sub-committee. The failure to take up that concern was considered to be a cultural rather

than a data accuracy question with the implication being that no AI procedure was sufficient

to improve the proper governance of this issue.

This illustrates that the fundamentals of good

governance practice remain important regardless of the benefits that more reliable information

brings.

E Cognitive insight and the Internet of Things

The formal use of cognitive insight in the boardroom was first reported in 2014 when Deep

Knowledge Ventures, a Hong Kong-based venture capitalist firm specialising in biotechnology

investments in aged care, announced the ‘appointment’ of VITAL

to its board.

On close

See Part III.C.2 above.

Chartis, (n 84) 7.

See par V.

See par III.C. 3.

APRA (n 2).

The acronym refers to ‘Validating Investment Tool for Advancing Life Sciences’.

Nicky Burridge, ‘Artificial Intelligence Gest a Seat in the Boardroom’, Nikkei Asian Review (online), (10 May

2017), <https://asia.nikkei.com/Business/Artificial-intelligence-gets-a-seat-in-the-boardroom>; Rob Wile, ‘A

Venture Capital Firm Just Named an Algorithm to its Board of Directors – Here’s what it Actually Does’,

Business Insider Australia (online), (14 May 2014), <https://www.businessinsider.com.au/vital-named-to-

board-2014-5?r=US&IR=T>.

inspection, however, it becomes clear that these first-comers are generally more aptly described

as robo-advisors and have been narrowly employed in business areas where decisions are

focussed on investments. In the case of Deep Knowledge Ventures, the high failure rate of

investments led the company to invest in data analytics of criteria that indicate success or

failure of investments in biotechnology. AI allowed for the identification of patterns that is

claimed not to have been recognisable by human actors before. VITAL is, for instance, fed

with data about prospective investments’ financing, clinical trials, intellectual property and

previous funding rounds.

In 2017, the CEO of Deep Knowledge Ventures credited the use of VITAL for saving the firm

from financial collapse.

VITAL showed that the focus of the firm’s investments ought to be

on longevity or prevention of diseases rather than on-treatment. The limitations of the system

were, however, acknowledged. While being hailed as the first AI member of a board, VITAL

was actually a tool used by the board to augment their own decision-making capability. It did

not have an independent vote at meetings and it certainly did not participate in meetings in the

manner one would expect from a human actor. Instead, the board agreed that it would not take

positive investment decisions without the corroboration of VITAL.

IBM is driving a new move in artificial intelligence in the boardroom, devoting an entire stream

of research into the potential of Watson to augmenting corporate governance.

IBM Watson

first came to the attention of the general public when it won the American game show

Jeopardy! in 2011.

Developed specifically with beating this game in mind, it showcased the

ability to interpret natural language even when faced with apparent ambiguity. The technology

that underlies Watson takes information from a variety of sources to compare possible answers

to the question posed. It evaluates which of these answers is supported by the best evidence

and then offers an answer.

100

Importantly, the sources of information are unstructured, meaning

that they are not neatly organised in datasets but come from natural language material. IBM

Watson can interpret these sources to aggregate knowledge sets and can continually update its

knowledge sets in this fashion.

After success on Jeopardy! the researchers working on Watson immediately announced the

extension of their research into other fields of application, the first being its potential use in

medical diagnostics.

101

Two main differences between this application field and that employed

for Jeopardy! were identified.

102

The first is that the system would not offer only a best answer

but would rather offer a range of hypothesis that would assist a health care practitioner in

arriving at the best diagnosis and treatment plan.

103

The final decision would remain with the

Wile (n 93).

Burridge (n 93).

See Tom Simonite, ‘A Room Where Executives Go to Get Help from IBM’s Watson’, MIT Technology Review

(online), 4 August 2014, <https://www.technologyreview.com/s/529606/a-room-where-executives-go-to-get-

help-from-ibms-watson/>. IBM Watson first came to the attention of the general public when it won the

American game show Jeopardy! in 2011, while competing against Ken Jennings and Brad Rutter, two of the

most successful human contestants in history.

See Adam Gabbatt, ‘IBM Computer Watson Wins Jeopardy Clash’, The Guardian (online), 17 February 2011,

<https://www.theguardian.com/technology/2011/feb/17/ibm-computer-watson-wins-jeopardy>; John

Markoff, ‘Computer Wins on ‘Jeopardy!’: Trivial It’s Not’, The New York Times (online), 16 February 2011,

<https://www.nytimes.com/2011/02/17/science/17jeopardy-watson.html>. Watson competed against Ken

Jennings and Brad Rutter, two of the most successful human contestants in history.

100

David Ferrucci et al, ‘Watson: Beyond Jeopardy!’ (2013) 199-200 Artificial Intelligence 93 at 94—95.

101

Markoff (n 97); David Ferrucci et al (n 98) 93.

102

David Ferrucci et al (n 98) 95.

103

David Ferrucci et al (n 98) 95—96, 104. The system is described as a ‘massive abduction machine’, meaning

that it would come up with an initial answer based on the information (symptoms) provided, followed by

eliminating some answers less supported by evidence.

doctor. Secondly, the system would actively encourage the completion of missing information

needed to improve its diagnostic result. In other words, it would encourage the doctor to ask

certain relevant questions or to run additional diagnostic tests. It immediately becomes

apparent that a similar boardroom capability could be of tremendous use – offer a range of

possible decisions or actions to a defined problem or encourage further investigation before

suggestions are offered by the intelligent system.

The major benefit that Watson’s cognitive computing solutions offered when compared to

other pre-existing systems was that Watson would be more user-friendly. Input of data could

happen in real time through the use of its natural language capability – it simply listens in on

the conversation between the patient and the health care practitioner.

104

The latter technology

is labelled ‘Voice-Enabled Cognitive Rooms’, which means that the physical space is

connected by speakers to the IBM Watson Internet of Things Platform.

105

Persons interact with

the system by verbal demands to change the physical space such as room temperature or

lighting, or to request information from the system. The same technology underlies the

governance solution of Ricoh discussed below.

Other explored applications of IBM Watson include the identification of potential research

areas in medical research,

106

the repurposing of existing medication,

107

improving image-based

diagnosis,

108

enhancement of legal research,

109

due diligence in mergers and acquisitions,

110

the hospitality industry

111

and many more. It is predicted that all information-intensive domains

will in future benefit from some form of AI application.

112

The IBM Watson governance research has resulted in a first iteration of product towards the

end of 2017 with the release of the Ricoh Cognitive Whiteboard.

113

This technology combines

the process automation and cognitive insight capability of AI with the Internet of Things, which

enables data collection and curation. Apart from performing such tasks usually performed by

digital whiteboards, the Ricoh Whiteboard is essentially a scribe, which records what is said at

meetings in real time. Additionally, it can translate the discussion in the participants’ language

104

David Ferrucci et al (n 98) 97—98.

105

Chris Preimesberger, ‘If These Walls Could Speak: IBM, Harman Powering Cognitive Rooms’, eWeek

(online), 19 April 2017, <http://www.eweek.com/innovation/if-these-walls-could-speak-ibm-harman-

powering-cognitive-rooms>.

106

Ying Chen, Elenee Argentinis and Griff Weber, ‘IBM Watson: How Cognitive Computing Can be Applied to

Big Data Challenges in Life Sciences Research’ (2016) 38(4) Clinical Therapeutics 688, 697.

107

Chen, Argentinis and Weber (n 104) 697—698.

108

Davenport (n 7) ch 2.

109

Anthony Sills, ‘ROSS and Watson Tackle the Law’, IBM Blogs (online), 14 January 2016,

<https://www.ibm.com/blogs/watson/2016/01/ross-and-watson-tackle-the-law/>; Jen Clarke, ‘Spotlight on

“LawTech”: How Machine Learning is Disrupting the Legal Sector’, IBM Blogs (online), 11 December 2017,

< https://www.ibm.com/blogs/internet-of-things/iot-spotlight-on-lawtech/>. ROSS Intelligence uses a

technique called ‘technology aided review’ to extract relevant data from unstructured data sets like legal

documents. In other words, it can aid legal research by quickly going through mountains of legal precedent to

come up with relevant authority. According to Clarke it can process over a billion text documents in a second.

See further Paul Lippe, Daniel Martin Katz and Dan Jackson, ‘Legal by Design: A New Paradigm for Handling

Complexity in Banking Regulation and Elsewhere in Law’ (2015) 93 Oregon Law Review 833, 849 – 850.

110

See for instance Rage-AI, < http://www.rageframeworks.com>; eBrevia, < https://ebrevia.com/diligence-

accelerator>; Seal, < https://www.seal-software.com/role-head-ma>; Kira, < https://kirasystems.com>.

111

Preimesberger (n 103).

112

Davenport and Ronanki (n 14).

113

D Craig MacCormack, ‘Alexa and Watson are Taking Over Your Conference Room’, Commercial Integrator

(online), 30 June 2017, <https://www.commercialintegrator.com/blogs/alexa-watson-control-conference-

room/>.

of choice,

114

create action items and perform other tasks previously left to a human assistant.

It can send a full transcript of the meeting to participants afterwards.

Others have warned that IBM’s claimed successes are overplayed and that the natural language

capability of the system often presents real difficulty in transposing it into different business

sectors.

115

For instance, investment analysts at a major investment bank wanted to interpret

their organisation’s combined investment reports in order to make recommendations to clients.

However, each analyst uses vastly different language and structuring in his or her reports. This

could not be solved by current technology, despite some of the claims made in media by IBM.

Instead, the firm had to appoint an outsourced service to systemise the reports so that they may

be mined for information.

116

As in the case of AI-driven legal compliance systems and other

applications of AI, the quality of the data available for analytics remains important.

IV POTENTIAL FOR FUTURE DEVELOPMENT

It is clear that the potential of AI business systems has not yet reached maturity and that many

benefits will only follow with technological advancement.

117

Not wanting to limit the

discussion to current AI, in this section, we attempt to answer the hypothetical: what is the

potential future development of the above technologies and how may they advance governance.

We have limited this discussion to capabilities that are already mentioned in the context of

existing technology but which have not yet developed.

Cognitive computing systems

118

may in future become a super-reporter of information. If the

system is integrated with other information sources in the organisation it has the potential to

become a deep pool of information that may be accessed by the board to assist in decision-

making.

119

Cognitive computing systems not only have the capacity to access large amounts of

data but also to analyse it and to come up with answers or potential answers. It is therefore

possible that a future board could rely on an intelligent system for information about the affairs

of the business rather than on the report of the executive that it is currently reliant on - data

presented on the surface using smart visualisations but which enables directors to zone in on

and drill down into more detailed information if required. This might not sit well with the

executive, which has always maintained a firewall between the day-to-today business of a

corporation and the information it shares with the board.

120

Certainly, recent Australian director

sentiments on board overreach into management in Australia have been very critical,

suggesting that AI overreach in the same field is likely to meet the same frosty reception.

121

Moreover, such systems could be enabled to pool information about governance behaviour

across many users (companies) so that patterns in governance behaviour might be traced, or

feedback could be given about benchmarking of certain governance issues across industries.

For instance, the system could be asked to report on the trends in executive remuneration across

114

At the time of release, it could translate into nine languages.

115

See, for instance, Davenport (n 7) ch 2.

116

Ibid.

117

Davenport (n 7) ch 3 and ch 4. The author lists some issues in ch 4 that work against the wide-scale adoption

of AI in business models at present, to which must be added as an overriding factor the cost of implementation.

118

‘Cognitive computer systems’ is used here in reference to technology that evidences all or any of the traits

according to Davenport and Ronanki’s classification set out above in par II.

119

See above n 107 for references on technology aided review.

120

Lancefield and Gagliardi (n 3): ‘[The executive] may worry that more information in board members’ hands

could lead to poorer decisions if it’s not being examined with the right set of lenses.’

121

Simon Evans, Patrick Durikin & Joanne Mather, ‘It's so soul-destroying': Business backs David Murray’,

Australian Financial Review (12 August 2018) https://www.afr.com/business/banking-and-finance/its-so-

souldestroying-business-backs-david-murray-20180801-h13ejn>.

a sector or the number of women in senior executive roles in comparable corporations. It could

then cross-reference this with the contributions of the various actors and their general

performance as directors. The concern this raises is the escalating degrees of transparency of

board processes, not all of which might be welcomed by boards. For example, it is foreseeable

that a data platform based on the Internet of Things as previously discussed could develop

benchmarking data on whether directors open and read their board reports and if so, how often

and for what periods of time.

The system could potentially log the contributions of the various participants and flag those

that have not yet delivered input, thereby countering group think and unconscious bias, a

primary concern in improving board performance.

122

It could in future address the board to ask

for the input of a particular director who has not yet contributed but who, according to data

already captured by the system, has the expertise to provide valuable input. It could ask the

chair to provide an opportunity for minorities represented on the board to address the board

when their opinion has not been heard for a while. It could also serve as a fact checker to flag

mistaken assertions. All of these capabilities have been mentioned as prospective abilities of

the Ricoh Cognitive Whiteboard.

123

Davenport recommends that companies design new systems from the outset with proper

consideration of the extent to which AI will be involved in decision-making compared to

human actors, as well as the strengths and weaknesses of each that the other seeks to

augment.

124

In the governance context the final responsibility for decision-making must always

remain with the members of the board.

125

It is therefore their capability that must be augmented

by the use of cognitive systems. For instance, there are current examples of consulting firms

developing, with the use of machine learning, tools to determine the strategic orientation of

companies and to make appropriate recommendations from those determinations.

126

The same

can, and is being, done in terms of business models. Similar tools could in future play a key

role in augmenting the ability of boards to plan strategically. Strategic planning needs to

include express consideration of the adoption of AI systems in the corporation, both in terms

of governance practices as outlines above as well as operational and customer-facing

interfaces.

Strategic planning and the monitoring of corporate culture goes hand in hand. The board’s

influence over the culture of an organisation may be improved through the use of organisation

network analysis enhanced by machine learning capability.

127

Instead of relying on staff

surveys to determine attitudes, this type of analysis mines the internal emails or telephone calls

of staff to determine who is communicating with each other.

128

It then picks up on how words

are written or spoken, rather than on the content of the messages, to determine the trust and

likeability factor between the actors. The outcome is that it becomes possible to tell who the

influencers are in organisations – those key persons who have the ability inside the organisation

122

Simonite (n 96).

123

See above n 111.

124

Davenport (n 7) ch 3.

125

This is in terms of the general law duty not to fetter discretions as well as the general law and statutory duty

(Corporations Act, s 180(1)) to act with care and diligence. See in general Robert P Austin and Ian M Ramsay,

Ford, Austin and Ramsay’s Principles of Corporations Law (LexisNexis Butterworths, 2015) [8.290] –

[8.305]. It is widely acknowledged that AI systems at present do not have ‘common sense’. For instance, in

diagnostic medicine they tend to be biased in favour of diagnosis of rare diseases, when it is much more likely

to be a more commonly found malady. See Davenport (n 4) ch 4.

126

Davenport (n 7) ch 4 describes the work of Boston Consulting Group in this regard.

127

Harry Toukalas, Tim Boyle and Ian Laughlin, ‘Turn Culture into Competitive Advantage with AI’, (2017) 70

Governance Directions 72.

128

Ibid, 74.

to change the attitudes of those around them.

129

The focus of the implementation of new ideas

or values can then be directed to these individuals, rather than to the organisation as a whole.

This can be done continuously, which enhances reliability, since influencers change over time.

This tool also has predictive capability since it may predict when staff will fail to comply with

certain directives.

130

V CHALLENGES AND IMPLICATIONS FOR GOVERNANCE THEORY AND

PRACTICE

Technological advances in AI stand to bring real benefits to corporate governance but

governance will not be business as usual. Some of the challenges that AI systems will present

in the boardroom are similar to that presented by technology in general. Other challenges arise

more specifically from the nature of the technology and its current limitations.

A Opaqueness of decision process

AI systems using machine learning use the presence or absence of a pre-acquired set of

reference points to decide whether an outcome is probable or not. While the reference points

will be known to human actors if the system was built via supervised learning, these reference

points are not known when the systems train through unsupervised learning. In supervised

machine learning a set of training data with a labelled output is used to train the system. The

instances or examples that would lead to a particular outcome is well-known. The analysis is

similar to statistical regression analysis incorporating scoring, which is the likelihood of the

outcome given the presence of a particular set of variables.

131

The training data is known from

the outset in the programming process. In unsupervised machine learning, the data analysed is

not labelled and the result is not known. Instead the system seeks to find patterns in the data

over millions of iterations to predict outcomes. This process may be bolstered by a process

called ‘reinforcement learning’, which means that the system has a defined goal and each time

it moves closer to that goal acknowledgement is fed back into the system to indicate that

progress has been made.

132

Essentially, the system trains itself through trial and error until it

has established the necessary patterns and will continue doing so. It becomes smarter over time.

In this model the data is not labelled, which means that the conclusion of the AI system may

not be traceable to particular reference points.

In both cases there is a measure of opaqueness in the steps that the system took to reach a

recommendation or prediction. In the case of supervised learning, one would know that specific

reference points were used to reach the conclusion. However, the relative weight assigned to

each of the reference points in reaching the conclusion may not be clear. This makes the testing

of the reliability of the system very difficult.

133

In an unsupervised learning system, the system

aggregates its reference points by virtue of learning from previously analysed data. This implies

that human actors may not be able to ultimately determine which reference points were used

by the system in reaching its conclusion or decision.

129

Ibid, 75 – 76.

130

Ibid, 75.

131

See in general, Davenport (n 7) ch 1. Supervised learning is still most commonly used.

132

Ibid.

133

See in this regard Joshua A Kroll, Joanna Huey, Solon Barocas, Edward W Felten, Joel R Reidenberg, David

G Robinson and Harlan Yu, ‘Accountable Algorithms’ (2018) 165 University of Pennsylvania Law Review

633. The authors consider different technological mechanisms to address concerns about reliability of systems

but ultimately conclude that none of the current mechanisms individually present a complete solution to this

problem.

If testing is difficult, the human actors that need to rely on the system cannot through their own

diligence tell whether they may safely rely on the recommendation or prediction. The result

may be that there is either an under-reliance on the system or an over-reliance. This may be

best illustrated through the use of an example.

Suppose that a corporation’s board resolves that it would not make decisions about acquisitions

without the corroboration of an AI system. At some point, there is a proposal before the board

to acquire a particular company and the proposal is supported both by management and all of

the members of the board. If the AI system does not corroborate their decision, will the board

go ahead with the acquisition despite the dissent and thereby under-rely or will they supplant

their own opinion with that of the AI system? Current governance theory holds that each

director must exercise independent judgment and cannot blindly rely on experts in the exercise

of his or her duties as a director.

134

This would extend to reliance on expert systems built on

AI technology. Weary of these obligations, directors may be reluctant to rely on AI predictions

or conclusions, leading to under-reliance with the attendant loss in opportunity.

By the same token, if the board of directors in our example used the system for a period of time

and it proves to be reliable, how much time will it take before the board’s reliance on the AI

system switches from corroboration to initiation? To put it differently, how long will it take

before the board relies on AI to suggest the target for acquisition, to be corroborated by the

board? And when this point is reached, how much dissent from the board will it take to override

the decision or opinion of the AI system? There is a very real risk that directors may reach a

point where they over-rely on the system because it is impossible to show that the system is

incorrect in its finding. This risk will be increased once the intellectual capability of the AI

system equals and then surpasses that of the human actors, a capability that is predicted to

eventuate and evolve anywhere from 2029 onwards.

135

B Over-dependency

Apart from the risk of over or under-reliance, a further risk is that dependency on AI may lead

to an overall decrease in skill and expertise in governance as more of the functions are left to

AI to perform.

136

Over-dependency is a risk attached to the use of all forms of technology, not

just AI, but it is exacerbated in this case in that AI could eventually make decisions on behalf

of human actors.

137

This may in turn lead to the temptation to defer to the decision of AI as a

rule. But what happens when the technology fails or becomes unavailable? Moreover, without

deep knowledge and skill of proper governance, how will human actors remain capable of

judging the merit and reliability of decisions or recommendations emanating from AI?

134

ASIC v Healy (2011) 83 ACSR 484. See further Austin and Ramsay, (n 123) [8.305] and the discussion in par

V below.

135

Ray Kurzweil, the current Director of Engineering at Google, has maintained for some time that AI will

achieve human intelligence by 2029. Recently, Google has publicly predicted that the ability to surpass human

intelligence should eventuate between 2029 and 2045. See Sean Martin, AI Warning: Robots will be Smarter

than Humans by 2045, Google Boss Says’, 17 October 2017, Express (online),

https://www.express.co.uk/news/science/867565/google-artificial-intelligence-ray-kurzweil-AI-singularity.

A general distinction is drawn between artificial general intelligence and technological singularity. Artificial

general intelligence refers to the ability of machines to match humans in that they ‘[possess] common sense

and an effective ability to learn, reason, and plan to meet complex information-processing challenges across a

wide range of natural and abstract domains’ (Bostrom, n 7, 3). ‘Technological singularity’ refers to the point

when machines exceed human cognitive ability. The term was first coined by Vernor Vinge (n 5) 11.

136

Lin, Abney and Bekey (n 12) 947.

137

We only consider technology outside of the human body here and not robotic augmentation that forms part of

the human body. See in this regard, Lin, Abney and Bekey (n 12) 946; Vinge (n 5) 12.

Dependency on AI must therefore be carefully monitored, as must the continued development

of sound governance knowledge and practices on the part of the human actors who use it.

C The business judgment rule

The advances in technology set out above has the most immediate effect on the duty of directors

to act with care and diligence.

138

The availability of enhanced information about the

corporation and the enhanced reliability of checks in the form of internal audit imply that the

duty of care and diligence will be elevated. In essence, the collaboration of these information

sources and the skills that the directors bring to the boardroom should lead to an increased

standard of care. At the same time, especially when AI capability comes into its own, there

may be the temptation to rely on the support of the technology to the exclusion of independent

judgment.

139

This brings the applicability of the business judgment rule into play.

A director’s business judgment will comply with her duty of care and diligence if she informed

herself of the subject matter of the judgment to the extent that she reasonably believes to be

appropriate

140

and if she rationally believes that the judgment is in the best interests of the

corporation.

141

A director’s reasonable reliance on the information or advice provided by others

will also need to be reassessed.

142

A director may rely on the information and advice provided

by others if the reliance was made in good faith and after making an independent assessment

of the information or advice, having regard to the director's knowledge of the corporation and

the complexity of the structure and operations of the corporation.

143

It is hoped that technology

will reduce the difficulty that complexity in structure and operations may pose, but the

independent assessment of the information or advice offered by the system remains in place.

This might prove difficult, especially if the system is one that derives its functionality through

unsupervised machine learning.

144

Moreover, the reasonable reliance provision is currently

couched in terms that only foresees the information or advice emanating from specified human

actors.

145

Increasingly the creation of information and advice will become autonomous from

human actors, which might mean that the content of the reasonable reliance provision could

become unnecessarily restrictive and in need of amendment.

Overcoming the challenge of applying independent judgment in a decision, or independent

assessment of information or advice more generally, when the reasoning of an AI system

leading to the specific recommendation or prediction is not open for determination or

comprehension is much harder. At the same time, it raises the prospect that directors may find

themselves between a rock and hard place – if they refuse to use predictive or augmenting AI

capability in the boardroom on the basis that they cannot exercise independent judgment over

138

Corporations Act (n 123) s 180.

139

See par V above.

140

Corporations Act (n 123) s 180(2)(c).

141

Corporations Act (n 123) s 180(2)(d).

142

Corporations Act (n 123) s 189.

143

Corporations Act (n 123) s 189(b).

144

See the discussion par V above.

145

Corporations Act (n 123) s 189(a): ‘a director relies on information, or professional or expert advice, given or

prepared by: i) an employee of the corporation whom the director believes on reasonable grounds to be reliable

and competent in relation to the matters concerned; or (ii) a professional adviser or expert in relation to matters

that the director believes on reasonable grounds to be within the person's professional or expert competence;

or (iii) another director or officer in relation to matters within the director's or officer's authority; or (iv) a

committee of directors on which the director did not serve in relation to matters within the committee's

authority [emphasis added]’.

its functions they may act in conflict with the duty to act in the best interests of the

corporation.

146

D The board/management divide

Increased use of automated processes and the potential improvements it may bring in terms of

more reliable internal auditing, better risk flagging and speedier and direct access to

information about organisational conduct brings with it a challenge to the board/management

relationship. The classical and well-entrenched informational divide between management and

the board may become muddled as directors and board committees become able to directly

access information that they may want about certain managerial actions and operations. This

introduces the risk, and fear from management, that the board may usurp functions typically

left to management instead on focussing on the strategic and supervisory role left to the board.

Ultimately, it begs the question about the proper role of the board. We believe that the time

and remuneration constraints attached to board appointments will keep interference in

management in check. Fears about meddling in management will probably not pan out.

E Data considerations

Some of the benefits of benchmarking may be hampered by a lack of generalisability of systems

across many corporations – the unique business attributes (read data) of corporations might

imply that most solutions will not be off the shelf but rather developed within a specific

organisation. Bespoke systems are costlier. Moreover, the usefulness of data gathered, for

instance, via the Internet of Things through AI systems across many organisations could prove

limited if the variance between these organisations is too great.

A key obstacle to the realisation of AI advances is the fact that data in organisations has not

kept pace with the development of AI, machine learning or even automation. Most data are

unstructured and to systemise it would take extensive resources. Furthermore, with the

emphasis in the recent past having been on optimising storage of data to increase computing

power, often much of the valuable data that could be used to inform the development of AI

capability in organisations have been intentionally destroyed.

147

Cyber security concerns may

further damper organisations’ enthusiasm to keep data for longer periods, which may also

weaken the pool of available data.

Tied in with the previous point is the push for greater digitisation of information inside

corporations.

148

AI systems rely on reliable data in order to function and so it stands to reason

that the creation and curation of data to enable the functioning of AI systems must become

institutionalised for the full benefits of AI to eventuate. We predict that this digitisation effort

will increasingly extend to the discussions at board and board committee meetings. The days

of the redacted board minutes may be numbered.

149

The technology already exists to create and

146

Corporations Act (n 123) s 181(1)(a).

147

SSON (n 24) 18.

148

See Lippe, Katz and Jackson (n 107) 844 where they make a similar point with regards to operations (contracts)

inside organisations.

149

Corporations Act (n 123) s 251A requires proceedings and resolutions of directors’ meetings to be included in

minutes. Catherine Livingstone recently claimed in proceedings before the Financial Services Royal

Commission that she challenged the board over its downplay of the severity of AUSTRAC’s investigation in

allegations of the Commonwealth Bank of Australia’s breaches of anti-money laundering regulation. She also

acknowledged that this challenge was not recorded in the minutes. The issue of the AUSTRAC investigation

was listed in the agenda, but the minutes only recorded that the matter was discussed. See Sarah Danckert and

Clancy Yeates, ‘CBA Accused of Board Minutes Criminal Breach’ The Sydney Morning Herald (21

November 2018) (online), <https://www.smh.com.au/business/banking-and-finance/cba-accused-of-board-

minutes-breach-20181121-p50hcp.html?promote_channel=edmail&mbnr=MTYwMzg3Nw&eid=email:nnn-

13omn659-ret_newsl-membereng:nnn-04%2F11%2F2013-business_news_pm-dom-business-nnn-smh-

store full transcripts of proceedings with minimal interruption or effort.

150

It will become

increasingly difficult for directors to justify the absence of such data when it may be useful for

the improvement of the governance of the corporation. The creation of the data for purposes of

AI analysis and governance improvement stand separately from the rights or interests of third

parties to gain access to the information.

151

The creation of the data does not necessarily mean

that access should be granted to a larger group of persons. The latter consideration stands

separate from the former.

The digitisation of board proceedings does, however, hold confidentiality implications.

152

This

will be especially true if the data is stored on the cloud or if the AI analysis is done via off-the

shelf products with the customisation done with the assistance of proprietors who also serve

competitors or who have conflicting interests with that of the corporation. These concerns have

already received some attention in other areas.

153

Moreover, dependency on the cloud may

have consequences across whole business sectors if it should fail, which holds some risk of its

own.

There will inevitably be some push-back from directors against the transcripts of meetings in

full on the basis that they must be able to speak their mind openly and without restriction at

such meetings. Furthermore, there will be the unspoken fear that their contributions, or lack

thereof, at meetings may be held against them in future when they are taken to task for breach

of duty.

154

It must further be remembered that much of the work of boards happen outside of

the boardroom in private discussions and ground work that precedes that actual board meetings.

These discussions will typically not be captured, meaning that the data captured at the actual

board meeting may not be as rich in information as hoped for.

155

This might be alleviated by

an awareness amongst directors of the importance of recording outside discussions formally

inside the board meeting. More generally, directors need to become aware of the importance

of data generation and reliability inside their organisations.

u&campaign_code=13IBU021&et_bid=29152781&list_name=2032_smh_busnews_pm&instance=2018-11-

21--07-29--UTC>.

150

The Ricoh Cognitive Whiteboard (above, par III. D) provides this function, amongst others.

151

Members do not generally have the right to inspect the minutes of board meetings, but may be granted limited

access if the member can show that she the inspection is necessary for a specific dispute in which the member

has a special interest. See generally, Austin and Ramsay (n 123) [11.390].

152

A full discussion of this area falls outside the scope of this paper. For more on data governance, see John

Ladley, Data Governance: How to Design, Deploy, and Sustain and Effective Data Governance Program

(Morgan Kaufmann, 2012); David Plotkin, Data Stewardship: An Actionable Guide to Effective Data

Management and Data Governance (Elsevier, 2014).

153

See for instance Maziar Peihani, ‘Financial Regulation and Disruptive Technologies: The Case of Cloud

Computing in Singapore’ (2017) Singapore Journal of Legal Studies 77, considering the regulation of the use

of cloud computing and other substantial outsourced services for the financial services industry in Singapore.

See further Basel Committee on Banking Supervision, Sound Practices: Implications of Fintech Developments

for Banks and Bank Supervisors (2017) 36 – 38, where recommendations are made for best practice where

outsourced services are extensively relied upon by financial institutions.

154

See Ellie Chapple and Elisabeth Sinnewe, ‘So What’s a Secretary to Do? Banking Royal Commission Raises

Questions About What’s in Minutes’ The Conversation (29 November 2018) (online)

<http://theconversation.com/so-whats-a-secretary-to-do-banking-royal-commission-raises-questions-about-

whats-in-minutes-107509>: ‘Company secretaries are already acutely aware that every set of minutes of every

board meeting might one day end up as evidence.’ See further Australian Securities and Investments

Commission v Hellicar (2012) 286 ALR 501 at [138] where the resolution adopting a misleading ASX

announcement in the minutes of a board meeting was held to have been a ‘contemporaneous record of

proceedings at the meeting’. See further Austin and Ramsay (n 123) [7.570].

155

The study by Forrester Consulting (n 28) 2 found that half of governance professionals (board members and

C-level executives) surveyed use their personal emails for sensitive internal board communications despite the

availability of board portals that offer secure chat rooms.

VI CONCLUSION

Corporate boardrooms globally are taking nascent steps towards their own digital

transformation. Adopting Davenport and Ronanki’s framework for AI systems, our

investigation shows that most of the current AI applications that aid governance fall in the

process automation classification (board portals, risk and auditing systems, legal compliance),

with some inroads having been made in cognitive insight (risk management, internal audit,

legal compliance). Systems that exercise cognitive engagement have not yet come into their

own but may yet add value in the boardroom as the technology matures (smart visualisations

with recommended actions, robo-advisors, capabilities enabled by the Internet of Things). All

of these technologies increase the collaborative intelligence in the boardroom as the directors

have quicker and more reliable access to information they need for efficient decision-making.

It frees up time that may be better spent on strategic planning and oversight.

Our central conclusion is therefore that AI, and the technology that supports it by way of data

collection and management, is poised to lend real benefit to governance practice. The

consequential gaps in the current statutory formulation of the business judgment rule could be

easily remedied by legislative amendment or even by means of expansive judicial

interpretation.

However, recent events brought to light by the Royal Commission show that despite the

availability of improved information, as facilitated by the technology discussed, governance

failures still abound. For instance, the benefits of greater risk reporting and internal audit

facilitated by the automated systems discussed above could have led to swifter and more

decisive action at CBA in the case of its breaches of anti-money laundering legislation.

156

However, despite these issues being flagged to the board the focus was mostly on financial risk

rather than on non-financial risk arising from non-compliance with laws.

The message seems clear: technology can only bring us so far. The rest of the way is determined

by the exercise of sound governance principles by company boards and the consistent nurturing

of a culture of compliance and ethical behaviour in their firms.

156

See Royal Commission (‘Final Report’ n 82) 396.

Unlocking the potential of augmented intelligence: a discussion on its role in boardroom decision-making

Article

Full-text available

Oct 2023

Recent developments in artificial intelligence (AI) have made AI applications in corporate governance an area of increasing interest to researchers and practitioners. Augmented intelligence, also known as advisory intelligence, is a form of collaboration between humans and AI, where the goal is to enhance human capabilities. It represents the second stage of AI development in corporate governance. In the realm of boards of directors, augmented intelligence offers a unique opportunity to enable machines to effectively collaborate with board members. This collaboration allows boards to harness the expertise of AI without supplanting the essential human element. In contrast, there exist more autonomous AI systems designed to operate independently, which could potentially lead to the replacement of humans on the board. This study seeks to assess the suitability of augmented intelligence for corporate boards, particularly in comparison to these more autonomous AI systems. Furthermore, it addresses the application of augmented intelligence in boardroom decision-making, its role in improving directors' decision-making abilities, and the associated risks and challenges.

State-of-the-art and possible fields of application for the integrated support of merger and acquisition processes by means of artificial intelligence

Article

Jan 2022

State-of-the-art and possible fields of application for the integrated support of merger and acquisition processes by means of artificial intelligence

Article

Jan 2022

A Smart Updater IT Governance Platform Based on Artificial Intelligence

Article

Sep 2020

This Information technology (IT) has a crucial role to improve business processes in companies. Getting the best technologies rapidly becomes as significant as understanding and developing the business plan of organizations. Thus, different IT best practices and norms are used by companies to help their services and IT business. These standards are set of best practices based on the experience and knowledge of numerous organizations; each of them focuses on specific governance issues such as ISO 27001 and ISO 27002 for IT security management, PMBOK and PRINCE2 for project management, ITIL for IT service management and COBIT for overall governance of an organization. As part of a collective research project that focuses on the IT governance axis, we have developed a smart global IT-GRC platform that allows to IT manager to design his own repository, considering the powers of each best practice, the organization's context and the IT strategic needs expressed by their stakeholders. To ensure the durability and the continuity of this project, we must consider the evolution of the IT GRC market, the problem posed is how to integrate recent versions of IT GRC frameworks, and how to ensure a periodic update of the knowledge base of the global IT-GRC solution. It is the subject of this paper and the second part of an IT-GRC research project.

تطبيقات الذكاء الاصطناعي ودورها في إرساء قواعد حوكمة الشركات

Article

Full-text available

May 2024

تمثل الهدف من هذا البحث في التعرف على تطبيقات للذكاء الاصطناعي ودورها في إرساء القواعد التي تقوم عليها حوكمة الشركات، من خلال تحليل ومعالجة البيانات والمساعدة في اتخاذ القرارات وتعزيز الشفافية والمساءلة واستغلال الفرص ومواجهة المخاطر في الشركات، كما يتم من خلال هذا البحث إلقاء نظرة على بعض الدول التي تبنت إستراتيجية التوجه نحو اعتماد تطبيقات الذكاء الاصطناعي بشأن إرساء قواعد حوكمة الشركات. تم التوصل إلى أنه يمكن لتطبيقات الذكاء الاصطناعي معالجة مجموعات كبيرة من البيانات والتعرف على الأنماط والاتجاهات وتقديم رؤى قيمة لإرشاد عملية صنع القرار في الوقت المناسب من قبل مجلس الإدارة والمديرين التنفيذيين. كما تعمل هذه التطبيقات على أتمتة المهام المتكررة مثل إدارة المستندات وإعداد التقارير وتقليل الأخطاء البشرية، وتحسين ممارسات إدارة المخاطر وهذا يعزز الشفافية والثقة بين أصحاب المصالح، مما يحمي سمعة الشركة ونزاهتها. ومع ذلك فإن تطبيق الذكاء الاصطناعي على حوكمة الشركات لا يخلو من التحديات والقيود، التي يجب العمل على التصدي لها. وبالرغم من تبني بعض الدول إستراتيجيات وطنية خاصة بتطبيق تقنيات الذكاء الاصطناعي في مختلف المجالات، إلا أن أغلبها لا ترقى إلى المستوى المطلوب. كلمات مفتاحية: ذكاء اصطناعي، حوكمة الشركات، إدارة البيانات، اتخاذ القرارات.

AUDITING WITH AI: A THEORETICAL FRAMEWORK FOR APPLYING MACHINE LEARNING ACROSS THE INTERNAL AUDIT LIFECYCLE

Article

Feb 2024

Vivek Shivram

Bibliography

Chapter

Nov 2023

Filippo Annunziata

The Use of Artificial Intelligence in Corporate Decision-Making at Board Level: A Preliminary Legal Analysis

Article

Jan 2023

Floris Mertens

La aplicación de tecnologías REGTECH en el cumplimiento del deber de información de los emisores de valores en Colombia

Article

Aug 2021

David Guillermo Ortiz Rey

IBM Watson: How Cognitive Computing Can Be Applied to Big Data Challenges in Life Sciences Research

Article

Full-text available

Apr 2016
CLIN THER

Life sciences researchers are under pressure to innovate faster than ever. Big data offer the promise of unlocking novel insights and accelerating breakthroughs. Ironically, although more data are available than ever, only a fraction is being integrated, understood, and analyzed. The challenge lies in harnessing volumes of data, integrating the data from hundreds of sources, and understanding their various formats. New technologies such as cognitive computing offer promise for addressing this challenge because cognitive solutions are specifically designed to integrate and analyze big datasets. Cognitive solutions can understand different types of data such as lab values in a structured database or the text of a scientific publication. Cognitive solutions are trained to understand technical, industry-specific content and use advanced reasoning, predictive modeling, and machine learning techniques to advance research faster. Watson, a cognitive computing technology, has been configured to support life sciences research. This version of Watson includes medical literature, patents, genomics, and chemical and pharmacological data that researchers would typically use in their work. Watson has also been developed with specific comprehension of scientific terminology so it can make novel connections in millions of pages of text. Watson has been applied to a few pilot studies in the areas of drug target identification and drug repurposing. The pilot results suggest that Watson can accelerate identification of novel drug candidates and novel drug targets by harnessing the potential of big data.

A Room Where Executives Go to Get Help from IBM's Watson', MIT Technology Review (online)

Aug 2014

See Tom Simonite

See Tom Simonite, 'A Room Where Executives Go to Get Help from IBM's Watson', MIT Technology Review (online), 4 August 2014, <https://www.technologyreview.com/s/529606/a-room-where-executives-go-to-gethelp-from-ibms-watson/>. IBM Watson first came to the attention of the general public when it won the American game show Jeopardy! in 2011, while competing against Ken Jennings and Brad Rutter, two of the most successful human contestants in history.

IBM Computer Watson Wins Jeopardy Clash', The Guardian (online)

Feb 2011

See Adam Gabbatt

See Adam Gabbatt, 'IBM Computer Watson Wins Jeopardy Clash', The Guardian (online), 17 February 2011, <https://www.theguardian.com/technology/2011/feb/17/ibm-computer-watson-wins-jeopardy>;

Jan 2013
94-95

David Ferrucci

David Ferrucci et al, 'Watson: Beyond Jeopardy!' (2013) 199-200 Artificial Intelligence 93 at 94-95.

The system is described as a 'massive abduction machine', meaning that it would come up with an initial answer based on the information (symptoms) provided, followed by eliminating some answers less supported by evidence

95-96

David Ferrucci

David Ferrucci et al (n 98) 95-96, 104. The system is described as a 'massive abduction machine', meaning that it would come up with an initial answer based on the information (symptoms) provided, followed by eliminating some answers less supported by evidence. powering-cognitive-rooms>.

Alexa and Watson are Taking Over Your Conference Room', Commercial Integrator (online)

Jun 2017

Craig Maccormack

D Craig MacCormack, 'Alexa and Watson are Taking Over Your Conference Room', Commercial Integrator (online), 30 June 2017, <https://www.commercialintegrator.com/blogs/alexa-watson-control-conferenceroom/>.

An Actionable Guide to Effective Data Management and Data Governance

Jan 2014

David Plotkin
Data Stewardship

David Plotkin, Data Stewardship: An Actionable Guide to Effective Data Management and Data Governance (Elsevier, 2014).

considering the regulation of the use of cloud computing and other substantial outsourced services for the financial services industry in Singapore. See further Basel Committee on Banking Supervision, Sound Practices: Implications of Fintech Developments for Banks and Bank Supervisors

Jan 2017
36-38

See For Instance Maziar
Peihani

See for instance Maziar Peihani, 'Financial Regulation and Disruptive Technologies: The Case of Cloud Computing in Singapore' (2017) Singapore Journal of Legal Studies 77, considering the regulation of the use of cloud computing and other substantial outsourced services for the financial services industry in Singapore. See further Basel Committee on Banking Supervision, Sound Practices: Implications of Fintech Developments for Banks and Bank Supervisors (2017) 36 -38, where recommendations are made for best practice where outsourced services are extensively relied upon by financial institutions.

See further Australian Securities and Investments Commission v Hellicar (2012) 286 ALR 501 at [138] where the resolution adopting a misleading ASX announcement in the minutes of a board meeting was held to have been a 'contemporaneous record of proceedings at the meeting

Nov 2018

See Ellie Chapple
Elisabeth Sinnewe

See Ellie Chapple and Elisabeth Sinnewe, 'So What's a Secretary to Do? Banking Royal Commission Raises Questions About What's in Minutes' The Conversation (29 November 2018) (online) <http://theconversation.com/so-whats-a-secretary-to-do-banking-royal-commission-raises-questions-aboutwhats-in-minutes-107509>: 'Company secretaries are already acutely aware that every set of minutes of every board meeting might one day end up as evidence.' See further Australian Securities and Investments Commission v Hellicar (2012) 286 ALR 501 at [138] where the resolution adopting a misleading ASX announcement in the minutes of a board meeting was held to have been a 'contemporaneous record of proceedings at the meeting'. See further Austin and Ramsay (n 123) [7.570].

Perspectives on the current and imagined role of artificial intelligence and technology in corporate governance practice and regulation

Abstract

Recommended publications

The Corporate Board in an Age of Collaborative Intelligence and Complex Risk

The corporate board in an age of collaborative intelligence and complex risk

Tourism Recommender System Based on Natural Language Classifier

Uma discussão sobre a utilização da inteligência artificial no judiciário brasileiro