BarnOwl Info Sharing Insight: The State of AI and AI in Corporate Governance with Michael Judin and Lionel Moyal

BarnOwl Info Sharing Insight: The State of AI and AI in Corporate Governance with Michael Judin and Lionel Moyal
BarnOwl Info Sharing session: 17 August 2023
Presented by: Michael Judin (Partner at Judin Combrinck Inc. Attorneys, Johannesburg, and Lionel Moyal (the SI and Channel Partners Lead at Microsoft Africa Region.

Introduction

Thank you very much Michael and Lionel for your most informative presentation on ‘The state of AI and AI in corporate governance’ at the BarnOwl info-sharing event held on 17 August 2023. Thank you too, to all those who attended the session.
We have entered a new era of technology revolution with the mainstream emergence of generative AI, with disruptive impact already across multiple industries. This session discussed the state of AI, its potential to accelerate growth and innovation as well as its impact to business. The session covered risks and governance considerations for boards of directors, including the Business Judgement Rule and the Directors’ and Officers’ liability Insurance.
The following are summary paragraphs generated by ‘Read AI’ which one of our delegates kindly sent to me. I took Michael’s advice and read through it, stress tested it and made sure that there is no bias and also made sure as per Lionel’s advice that the technology did not decided to use its ‘creative’ side and come up with any hallucinogenic ideas. Although I had to make one or two very minor changes, ‘Read AI’ really has come up with a brilliant summary from the recording:
The meeting focused on the state of artificial intelligence (AI) and its impact on corporate governance. Lionel Moyal and Michael Judin provided an overview of AI and its use in the workplace, including the importance of governance and legal implications when using AI in an organizational context.
Lionel demonstrated how to use Chat GPT and Bing Chat Enterprise to analyze large amounts of information, such as public audit reports and risk management reports, and discussed the intricacies of AI language models.
The meeting emphasized the need for actionable insights and contextual interpretation of risk reports, as well as the importance of responsible AI, fairness, reliability and safety, privacy and security, inclusivity, transparency, and accountability.
Michael Judin discussed the challenges and liabilities that directors, committee members, and prescribed officers face under the New Companies Act, and introduced the Business Judgment Rule as a way to shield individuals from liability provided they act ethically in the best interests of the company. He also stressed the importance of stress testing AI generated opinions for proper governance and compliance with the business judgment rule.

The AI evolution

Everybody has heard about AI and how it’s changing the world. We are all aware of the release of ChatGPT and its massive impact on technology, on business and in fact on society as a whole. It’s important to understand what AI can do, what it can’t do, how it’s progressing and how we’re going to start to see it manifest more and more inside organisations and everyday life.
Lionel took us through a brief roadmap of AI journey from 2016 onwards, although the initial origins of artificial intelligence were in the early 1950s.

The Era of generative AI

GPT-3:
The GPT-3 model is generally around large language models which essentially uses very large bodies of text that have been fed into a very capable training model with billions of parameters. So it literally gets to learn how this text actually relates to each other and is able to then generate or predict the next word. It has a probability model and it predicts what should come next in a sentence at its very basic core.
But what is very interesting is that through scaling and through giving it larger and larger amounts of data and larger and larger amounts of compute, we’ve seen emergent behaviour. Emergent behaviour is when you get a different qualitative output by changing just the quantitative inputs. The more you put into something, you suddenly get a very different qualitative response. It’s interesting to see what emerges inside of these training models.
Codex:
Codex is a code generation model, which has become very popular amongst developers, because computer code is actually just text, and being able to train on large volumes of text of code means that models can actually start to predict based on what it is you’re trying to do.
DALL-E:
DALL-E is image generation where for example, you can say, I need a picture of a ‘horse on the moon’ and it will actually generate a real picture. It uses very large bodies of content and media to be able to generate this through things like diffusion models, which is a technique which has recently been developed and is incredibly powerful and able to create realistic images and which is now also being applied to video.
ChatGPT:
You can think of ChatGPT, as a large language model and reasoning engine in your pocket. You can ask it questions, you can feed it information, you can have it assist you to answer questions.
Imagine that you have hundreds or even thousands of documents that you need to consume and you need to find trends, patterns, etc. It could take you days or even months with a team of people to go through and analyze this information. Whereas it’s easy to use a tool like ChatGPT or a large language model and feed it large amounts of information and use its capabilities. You don’t have to train it on this new content. It has already been trained, it has the abilities now and it has a depth of knowledge. It has the ability to take in new content, which is your own content and start to analyze this and give you patterns and insights which are very actionable.
Lionel took us through two practical examples being an analysis of a reports from the Auditor General relating to the PFMA as well as using the IRMSA 2023/204 Risk Report. Lionel demonstrated how to get more and more detailed output and insights by posing additional questions to ChatGPT.

Next generation AI

The Microsoft Bing search engine has recently released a Bing chat function which is the public version of ChatGPT. It takes the ChatGPT engine, and combines it with the Bing search engine so that all of your information, the questions you ask are not just up to 2021 (which is when ChatGPT stopped its training) but uses fresh searchable information. However, this is a public tool and these public models have retained the right to use your content for learning.
What’s interesting is that it’s not like ChatGPT where it will just give you the answer, but it’ll give you references as well. It augments what it searches or what it gives you with information from the web. So it will show that this particular piece of information or reference came from the document and this particular piece of information it actually gathered from a public source (i.e. the Web) with a reference. So it becomes a much more useful as a business tool.
Microsoft has just released ‘Bing Chat Enterprise’ where all of the information that you send to it, remains the property of your company and doesn’t get sent or shared publicly.
A word of caution: Context becomes very important. It’s very important to remember that ChatGPT is a large language model and is essentially a tool that predicts the next word. Predicting the next word means that there’re a range of possibilities of what that word could be. However ChatGPT retains the right to be creative so it’s important to understand that an AI model can be a little bit creative and take it in a different direction. This is why it’s very important to prompt it correctly and give it the right context.

The era of the copilot

AI is an incredible productivity tool and does not replace the human. It’s what we call the era of the copilot. And the best way to explain it is to differentiate what we mean by the copilot.

The foundation models are very large, very complex, very difficult to train, and costly to train.
GPT and GPT 3 and GPT 4 is what ChatGPT uses and can be used in your own context and environment without ever having to train such big models. They are used as a reasoning engine and as a tool for additional productivity. Organisations are starting to build in these models and creating what we call copilots. They connect into your organisation and into your applications. These are conversational and expert helpers that are scoped to assist you with specific complex tasks. Using the foundational capabilities of the model, you start to train your own copilot with your own organisational content, templates and intellectual property. We are seeing copilots emerge in sales and marketing and productivity, customer service, competitive intelligence. Copilots can be used in a generic way but can also be used as very specific helpers such as legal assistance.

Next Steps

The playing field has changed. This technology is here today and it’s groundbreaking. It provides the opportunity to re-imagine your business and not to just think incrementally. Think about how you do things and how you could be doing things differently. Think about how this could completely change the way you work, not as a replacement, but as an able assistant or copilot.

The state of artificial intelligence and its relationship to corporate governance

Business Judgment Rule
In the context of this exciting new technology, there are important risk and governance considerations for boards of directors, including the Business Judgement Rule and the Directors’ and Officers’ liability Insurance. The following is an extract from the Institute of Directors paper on ‘Guidance for Boards – Exercising Business Judgment’:
Directors, whether executive or non-executive, are required to exercise their judgement as to the best decisions or courses of action available to a company. However, sometimes even the best laid plans fail, and, in such instances, the assessment of the appropriateness of the decisions taken by the directors is not based solely on the manner in which the decision turned out but are also based on the process that the directors followed in arriving at their decision.
This is the essence of the ‘business judgement rule’1 which entered the South African legal lexicon with the arrival of the Companies Act, 20082 (‘the Act”). It has been heralded by many as the key form of protection for directors and allows them to make informed, reasonable judgements without the threat of liability for a potential negative outcome hanging over their heads.
However, to avail themselves of this protection afforded by the Act, a director’s conduct must conform with the requirements for the protection in that: –

• reasonably diligent steps were taken to become informed;
• the director must not have had a material personal financial interest in the matter, and;
• the director must have had a rational basis for believing that the decision was in the best interest of the company.

Reasonably diligent steps are taken to become informed
“Reasonably diligent steps were taken to become informed” is important in context of our AI discussion.
As Michael points out:
You have to be properly informed and whilst the rule allows one to rely on the opinions of experts, you are required to have stress tested it to ensure that you can rely upon it before you give an opinion or advice. So to take an example which is probably familiar to all of you, a lawyer or an advocate gives you an opinion. You can’t simply take that opinion and present it to the company, to the board, to any other stakeholder and say this is what the position is. You are required under this rule to read it, to understand it, to question it. In other words, stress test it before you take it on as your own opinion. Otherwise, it’s not your opinion, it’s the opinion of someone else, which you are repeating.
And that plays very much into what Lionel has so eloquently explained to you today is that you can’t simply relax for the weekend before your board presentation on Monday and late on Sunday night, after consuming much food and wine, go to ChatGPT and ask her to prepare your presentation for the board and arrive the next morning and dish it out and say here’s my presentation, it’s not yours. It’s a report which has been generated, but it’s an automated response. And for it to be your opinion, you have to have stress tested that opinion, to have questioned it and to make sure that you can refer to it as your opinion. It’s been generated from an automated system and it’s not your opinion. You may have prompted it, so in a proper governance world, you have to disclose that.
This very important governance aspect was very well explained in probably one of the most leading governance cases in the world, an Australian case called The Centro case which highlighted that “directors cannot substitute reliance upon the advice of management for their own attention and examination of an important matter that falls specifically within the board’s responsibilities…” You can find more information on the Centro case in the Institute of Directors paper on ‘Guidance for Boards – Exercising Business Judgment’.
So what AI tools can generate is half the journey and we need to complete the journey from a governance perspective to have the benefit of the business judgment rule. You’ll have the same issue if you’re being questioned by the auditors regarding liability insurance as to your conduct, how you questioned, how you interrogated and how you were satisfied that you understood and that you could convey it as your own opinion, the information you had used from an AI tool.
In that questioning, in that understanding and in that interrogating you will be bringing to that content what are known as this three human super talents of curiosity, humility and emotional intelligence, which is absolutely important because it’s not sufficient for it just to have intellectual intelligence.
Disclosure
So the big lesson from what Lionel has explained to us this morning is that these AI tools are of enormous assistance, however from a governance perspective, it doesn’t stop there. In order to comply with the business judgment rule, which shields you from liability if things go wrong, you need to be able to tell a court that “I used ChatGPT or I used Bing, but when I relied upon it, it was after I had, to the best of my knowledge, independently, with no conflict, had questioned what I didn’t understand. I had sought expert opinion and I honestly and reasonably believed this to be correct.”
Another point Michael raised is the use of AI in HR and the recruitment process where you need to be very careful about algorithmic bias.
One needs to be very careful that what you are presented with and rely upon and assume is somebody’s own work, was not generated by somebody who didn’t know how to prompt AI effectively. It is very important to make sure that you have policies and procedures in your organisation regarding the use of AI and the importance of employees being obliged to disclose where AI has been used.

Conclusion

This technology is here today and it’s groundbreaking. It provides the opportunity to re-imagine your business and not to just think incrementally. Think about how you do things and how you could be doing things differently. Think about how this could completely change the way you work, not as a replacement, but as an able assistant or copilot.
The huge advantage and the excitement of what this technology offers us, is the ability to interrogate in depth, which we couldn’t do before, vast amounts of information, gather it together and use that as the basis for our opinion. It has enabled us to give an opinion which would not humanly have been possible to do before.
This must be combined with the three human super talents of curiosity, humility and emotional intelligence because it’s not sufficient just to have intellectual intelligence.
Use this incredible technology in your own organisation responsibly, ethically, effectively, in a conscious way, with all of the guardrails, following the templates that the Microsofts’ of this world are producing.

Presentation and video links

Please see attached presentation here, and the info sharing recording here.
Please see the Institute of Directors paper on ‘Guidance for Boards – Exercising Business Judgment’

The following are screen dumps of output produced by ‘Read AI’:

Related links

https://api.barnowl.co.za/integrated-grc-featured/king-iv-now-more-than-ever/
https://barnowl.co.za/knowledge-centre/

Contact us

Cheryl Keller | BarnOwl: cheryl@barnowl.co.za
Michael Judin| Michael@elawnet.co.za
Lionel Moyal| limoyal@microsoft.com

Thank you

Once again, thank you Michael and Lionel for your time and for your informative presentation and thank you to all those who attended our info sharing session. We look forward to seeing you at our next info sharing session. Please keep a look out for our upcoming events at: https://api.barnowl.co.za/events/

Kind regards
Jonathan Crisp
Director – BarnOwl GRC and Audit software

About BarnOwl:

BarnOwl is a fully integrated governance, risk management, compliance and audit software solution used by over 150 organisations locally and internationally. BarnOwl is a locally developed software solution and is the preferred risk management solution for the South African public sector supporting the National Treasury risk framework.
Please see https://api.barnowl.co.za for more information.

About our guest speakers

Biography Michael Judin

Michael is a non-executive ex officio director of, and legal advisor to, the American Chamber of Commerce in South Africa (“AmCham”). Michael is also the non-executive chairman of the Advisory Board of the AmCham. On November 22, 2006, Michael was inducted as an Honorary life Member of the AmCham in recognition of exemplary leadership and dedicated service.

Michael is Co-Chairman of the Corporate Governance International Development sub-Committee of the American Bar Association’s Business Law section.

Michael is a member of the e-Commerce Advisory Committee of the Financial Sector Conduct Authority (formerly Financial Services Board) of the Republic of South Africa.

Michael was a member of the Task Team which wrote the original Code for Responsible Investing in South Africa (“CRISA”) Committee and is a member of the Task Team that wrote the new code.

Michael is a member of the advisory panel of experts of The Good Governance Academy NPC, chaired by Prof Mervyn King. Michael is the non-executive Chairman of the Conscious Leadership and Ethics Institute NPC.

Biography Lionel Moyal

Lionel is the Channel Partners Lead at Microsoft Africa Region.

Lionel has a degree in aeronautical engineering as well as an MBA from Wits University.

Lionel is an advocate for digital transformation, responsible AI and organisational innovation through people and technology.

Lionel is a subject matter expert on aligning technology driven transformation with corporate governance as per King IV.

Prior to joining Microsoft Lionel was a co-founder and Managing Director of Intervate, a leading Microsoft partner focused on productivity, cloud and mobile solutions for enterprise customers.

Lionel is passionate about technology and innovation and engages to support the local technology and start-up community.