The Director's Playbook for Generative AI
Without a doubt, the value proposition of generative artificial intelligence (AI) is alluring. Opportunities for using it to enhance customer experiences, increase process efficiencies, innovate products and services, and improve productivity are immense—despite its risks and limitations.
To contribute value in boardroom discussions about implementing generative AI models, it is necessary to understand their opportunities, limitations, and risks. Directors should immerse themselves in learning about and getting hands-on with using this accessible technology. They should learn from experts inside and outside the organization and from published articles providing relevant content. Armed with a baseline understanding, directors should consider the following questions when engaging CEOs and their teams in strategic conversations regarding generative AI:
What is the business opportunity in deploying generative AI? This critical “Why should we care?” question should be considered strategically and tactically. Directors can ask the following five high-level questions to advance the conversation:
What are the implications of generative AI for our industry, and what are competitors doing with it?
Do we have a strategy for why, where, how, and when we will deploy generative AI? What use cases are we considering, and how are we selecting and prioritizing these opportunities and measuring the value contributed?
Are we organized appropriately to roll out our strategy? How are we empowering our people to build, train, and use generative AI?
Have we documented our organization’s guidelines and values for privacy, security, transparency, fairness, human versus machine responsibilities, and other matters related to our generative AI deployments? Do our policies account for the need to govern and manage this technology differently than nongenerative AI?
How do we know we are adhering to our guidelines and values? For example, do we have a cross-functional ethics committee that vets all plans and actions and monitors for unintended outcomes and consequences?
Insights gained from this discussion enable the board to understand how and why management intends to position generative AI in the business.
What are the legal, regulatory, and ethical issues we need to address? Generative AI is on the radar of regulators and policymakers at the national, state, and local levels as well as of other stakeholders due to the potential cyber, privacy, and societal risks. The environment varies by country and region. With legislative initiatives already underway and risk frameworks emerging around the world, directors should inquire how management keeps track of market developments.
How are we sourcing and managing the data used by our generative AI model(s)? Directors should obtain an understanding from management regarding whether the organization is using (a) publicly available models and domains, (b) foundation models that are fine-tuned with internal proprietary data, or (c) fully customized models. Whether a company uses its own data, third-party data, or data generally available in the marketplace will influence a model’s risk profile.
Do we have the talent we need to implement generative AI? Finding and onboarding the requisite talent and expertise is key to determining the mode of generative AI a company can deploy. While publicly available tools such as ChatGPT require no specialized expertise, they are far less secure, private, and reliable. That is why most companies will likely choose the middle road: fine-tuning a foundation model, which requires lighter data science expertise through a low-code interface.
Do we have a governance framework that enables experimentation? The board should inquire about the governance process and organizational structure for overseeing the company’s generative AI innovations and monitoring industry developments. Overall governance involves considerations relating to trust, ethical use, risk management, the third-party ecosystem, legal and regulatory compliance, and standards and controls. It entails a generative AI review and approval process. An adaptable governance framework could function through a small cross-functional, multidisciplinary team representing the data, engineering, security, and operational aspects of generative AI models.
What monitoring mechanisms and accountabilities do we have in place? Model owners—those responsible for their design, development, and operation—should be held accountable for their proper functioning. Human oversight supported by automated alerts and controls is an integral part of any generative AI solution, particularly when the model is connected to hardware or software, or there is a significant impact on sensitive decisions, e.g., employment matters. The board should inquire as to whether a process is in place to assure generative AI model outcomes align with intended results and comply with relevant regulatory requirements. Due to the complexity of the correlations in the model, it may be necessary to embed self-check mechanisms and conduct human reviews of AI-generated output. Internal audit can also serve as a check and balance. Models should be evaluated periodically for unreliable or biased content.
How do we manage the risks? Early implementations have exposed generative AI’s shortcomings: content source and provenance are not always evident, ownership rights are a major concern, bias and prejudice in text and images can be an issue, and images or videos appearing realistic can be deceptively false (deepfakes). Models can hallucinate or drift, that is, they can deliver results not backed by the data to which they have access. These issues can lead to misinformation—inaccurate and misleading content—and blatant plagiarism. They can also lead to disinformation (e.g., fake news; mimicking people or specific individuals through falsified photographs, videos, and voice scams). They open the door to more sophisticated cyber threats and deceptive social engineering strategies. Boards should ascertain how these issues are being addressed.
What are the change management issues? With resistance to change a formidable challenge for many organizations, management should communicate the following:
Generative AI technology’s strengths and limitations
The intention to deploy the technology thoughtfully, responsibly, and in accordance with applicable laws and regulations
The initial use cases planned and how those use cases align with broader strategic efforts, such as environmental, social, and governance as well as diversity, equity, and inclusion initiatives
The risks to be managed, including protection of the company’s intellectual property (e.g., trade secrets, other confidential information)
Reskilling and upskilling will be necessary for employees whose job functions are affected by generative AI.
The dawn of generative AI is yet another wake-up call for boards, another disruptive force for business. In this digital world framed by the Internet, digital devices, smart devices, the cloud, and ever-increasing connectivity, mobility, and computing power, directors rooted in the analog age and unable or unwilling to make the transition to be technology-engaged in the boardroom need not apply.
Protiviti is a NACD partner, providing directors with critical and timely information, and perspectives. Protiviti is a financial supporter of the NACD.
Jim DeLoach is managing director of Protiviti. DeLoach is the author of several books and a frequent contributor to NACD BoardTalk.