"GPT-4, Write a Blog Post About Director Skills and AI"
Key findings
- GPT-4 estimates that just 2.4% of directors at constituents of the MSCI ACWI Index could be considered direct experts in artificial intelligence (AI), though transferable and ancillary AI expertise were more common.
- AI experts were naturally more common on boards in the information-technology sector, though these boards struggled in terms of ancillary experts, suggesting some of these companies may face challenges in overseeing the effects of AI products on external stakeholders.
- Due to the small number of estimated AI experts, increased demand from nomination committees for AI knowledge could place experts at risk of being overburdened by board commitments.
Board expertise is one lens through which investors can scrutinize companies' efforts to manage the risks and opportunities of AI. But the sudden emergence and rapid evolvement of AI as a business-ready technology suggests that boardrooms may lack expertise that could help directors and companies navigate these challenges. Despite some experiments giving generative AI a direct voice in the boardroom, human directors remain corporations' ultimate decision makers.[1] Given this fact, it's important for investors to have a sense of how well those human directors understand this new technology, as this could have an effect on how well companies are managed.
Using GPT-4 to evaluate AI expertise
To understand how well-prepared boards are to address AI today, we used OpenAI's GPT-4 model to read the biographies of over 24,000 directors and categorize these directors under three definitions of AI expertise: direct, ancillary and transferable.
Definitions of AI expertise
Type | Description | Examples |
|---|---|---|
Type Direct | Description Expertise specifically related to the development and application of AI. | Examples - Professional experience in machine learning. - An advanced degree in computational linguistics. |
Type Transferable | Description Expertise that can provide a foundational background for understanding the development and application of AI. | Examples - Professional experience in data science. - An advanced degree in mathematics. |
Type Ancillary | Description Expertise related to potential risks and consequences of AI. | Examples - Professional experience in data privacy. - An advanced degree in intellectual property law. |
Source: MSCI ESG Research.
We applied these definitions to 24,035 individuals who served on the board of a constituent of the MSCI ACWI Index as of March 25, 2024.[2]
GPT-4 estimated that direct AI expertise was scarce, with just 2.4% of individuals meeting our definition. This is a lower frequency than even risk-management expertise — the rarest skill we evaluate for directors in our current corporate-governance methodology, with just 3.8% of individuals in this sample qualifying as risk experts.
Estimated percentage of over 24,000 individuals with AI expertise, MSCI ACWI Index
Includes individuals who served on the board of a constituent of the MSCI ACWI Index in our corporate-governance research and MSCI ESG Ratings coverage as of March 25, 2024, excluding individuals with no biography (n=24,035 individuals). Source: MSCI ESG Research.
Though few in number, GPT-4 found that directors with AI-related experience were widely distributed across sectors. One in five companies had a direct AI expert on hand, comparable to risk-management expertise, which was found at 23% of boards. Half of companies had access to a director with transferable or ancillary AI expertise. While higher than the other categories, this was still well below rates for financial expertise (found on 97% of boards).
Companies with at least one estimated AI expert per sector, MSCI ACWI Index
Includes constituents of the MSCI ACWI Index in our corporate-governance research and MSCI ESG Ratings coverage as of March 25, 2024, where at least 50% of directors had a published biography associated with their profile (n=2,729; communication services: 142; consumer discretionary: 278; consumer staples: 213; energy: 104; financials: 455; health care: 228; industrials: 443; information technology: 336; materials: 266; real estate: 128; utilities: 136). Sector classification based on Global Industry Classification Standard (GICS®), the global industry classification standard jointly developed by MSCI and S&P Global Market Intelligence. Source: MSCI ESG Research.
The results produced by GPT-4 have not been verified by a human. The figures should therefore be treated as imperfect estimates and used with caution.
We undertook a model validation exercise involving keyword searches and a manual review of 254 individual profiles. Relative to our validation sample, GPT-4 had the following accuracy rates:
- Direct expertise: 93% accuracy
- Transferable expertise: 84% accuracy
- Ancillary expertise: 84% accuracy
- Overall: 88% accuracy
Load management
The relatively small number of AI experts suggests that these individuals may be especially sought after by board-nomination committees and could thus become vulnerable to "overboarding" — that is, serving on an excessive number of boards. Overboarded individuals risk seeing their overall effectiveness as a director on any one of their boards reduced due to the total volume of work to which they are expected to contribute.
Overboarded individuals by estimated AI-expertise category, MSCI ACWI Index
Includes individuals who served on the board of a constituent of the MSCI ACWI Index in our corporate-governance research and MSCI ESG Ratings coverage as of March 25, 2024, excluding individuals with no biography (n=24,035 individuals). Individuals were considered overboarded if they served on four or more boards and were not an executive at any of those companies, or if they served on three or more boards and were an executive at any of those companies. Source: MSCI ESG Research.
To date, direct AI expertise has been something of an exotic commodity, with estimated experts having lower overboarding rates than any other expertise category — including directors with no AI expertise — despite the small number of estimated experts. In contrast, ancillary experts had higher overboarding rates than any other type of director we evaluated. These existing board commitments suggest that ancillary expertise may be hard to acquire through director recruitment alone.
For all types of AI expertise, director recruitment is only one approach available to boards. Boards can also rely on director education — whether with external consultants or internal experts — to leverage existing expertise and upskill boards organically. This approach could help to minimize risks from overboarding, as well as other potential risks from a reliance on "specialist directors."[3]
Preparing for the future of business
Reviewing the expertise of boards is just one way investors can analyze companies' efforts to manage the risks and opportunities of AI. Our recent paper, AI Engagement: A Stakeholder Approach, sets out to help investors engage with companies on a broad set of AI risks. It provides sample engagement questions and evaluation approaches addressing risks related to four key stakeholder groups: the board and management, customers and data subjects, companies' workforce and the environment and climate.[4]
1 For example, Abu Dhabi's International Holding Co. gave an AI model the role of non-voting board observer in February 2024.2 Among constituents of the MSCI ACWI Index as of March 25, 2024, 24,778 individuals served on the boards of 2,747 companies in our ESG Ratings and corporate governance research coverage. Of these individuals, 743 (3.0%) had no published biography associated with their profiles. These individuals were excluded from analysis, leaving a final count of 24,035 individuals. Of these companies, 18 (0.7%) had boards where more than half of directors lacked a published biography associated with their profile. These companies were excluded from analysis, leaving a final count of 2,729 companies.3 Roy Shapira and Yaron Nili, “Specialist Directors,” Yale Journal on Regulation, forthcoming, November 2023.4 Yoon Young Chung, Siyu Liu and Harlan Tufford, “AI Engagement: A Stakeholder Approach,” MSCI ESG Research, May 2024 (client access only).
The content of this page is for informational purposes only and is intended for institutional professionals with the analytical resources and tools necessary to interpret any performance information. Nothing herein is intended to recommend any product, tool or service. For all references to laws, rules or regulations, please note that the information is provided “as is” and does not constitute legal advice or any binding interpretation. Any approach to comply with regulatory or policy initiatives should be discussed with your own legal counsel and/or the relevant competent authority, as needed.