Most of the time, when asked, I have no hesitation in rattling off a list of applications of AI and all the ways it can go wrong. I might also go on at some length about why it's always a bad idea to talk about AI in the royal sense (i.e., “AI”); it's practically a (bad) party trick. But recently, when I was asked to characterize the main ethical issues with AI, I completely blanked out. It was an uncomfortable and instructive moment on many levels.
On the positive side, the awkward silences and tangled words raised a question: How many employees find themselves in this exact situation every day, faced with questions and pressures around the use of AI? This may be why employees are hesitant to propose new ideas or speak out against questionable practices, or why they see AI governance as a dark quagmire to be avoided at all costs.
Given today's inflated AI expectations, employee empowerment must be a top focus. AI-enabled systems permeate most aspects of business today, yet often remain invisible or unnoticed. For example, when vendor systems are purchased without full disclosure or inspection of embedded AI components, or when individuals informally extend operations using readily available tools such as ChatGPT. Add in the market frenzy for all things AI everywhere, and the situation becomes bleak. Even the most well-defined policies and standards cannot work in isolation.
Good governance prescribes limits to ensure that the products developed are appropriate for use and do not violate consumer expectations or regulatory or compliance requirements. But good governance prioritizes the additional goal of increasing organizational literacy and self-awareness. It empowers all employees, regardless of level, to think critically about when and where an AI-enabled system makes sense. Just as important, it empowers them to raise their hand when it doesn't make sense.
Building broad, functional literacy requires investments — not just in time and talent, but in tools. A comprehensive blueprint is beyond the scope of this article, but to kickstart your AI literacy program, consider these three types of materials:
Think of this as a basic guide to your analytics and AI toolbox: the tools themselves, not their outputs. Our goal is not to drown you in technical details or turn you into an expert on any particular technology. Rather, our goal in this catalog is to familiarize all employees with the broad range of analytics and AI tools available, and to further equip them with the foundational knowledge to determine whether a particular tool (e.g., algorithmic technique) is worth considering.
- A hammer is useful for…
- When using a hammer, always… (place your thumb between the hammer and the head of the nail)
- Consider using a screwdriver if…
The materials catalog should also provide instructions on what resources to refer to and what processes to invoke when using each tool.
Initially, I wanted to call this a product catalog, but that would be confused with an index of an organization's finished AI products. Such accounting may soon become mandatory as part of emerging regulations in the EU and elsewhere. Therefore, a comprehensive AI reference library would catalog both raw algorithmic elements and finished products.
Organizational governance and ethical principles often sound clear in concept but become unclear when the practical context comes into play. This is a well-known problem, and one that jurisprudence guides: laws and statutes establish rules and rights, and in practice are interpreted through case law.
Similarly, AI case references provide valuable insight into how principles and regulations manifest in practice. Such reference libraries can be particularly useful in highlighting clear red lines for your organization and illustrating the trade-offs to consider relative to situational reputational or ethical boundaries. This is especially true when, as is often the case, regulations, laws, reputational, or ethical codes conflict.
As mentioned above, such references are not immutable, fixed precedents. Nevertheless, they are one of the best ways to concretely demonstrate the company's application norms related to AI. Therefore, your AI reference library should include applications that have been approved with constraints and applications that have been discontinued. External examples are useful in the early stages, but they usually highlight negative examples rather than positive ones.
- AI Quick Start Evaluation
AI applications cover a wide range of regulatory, compliance, and legal obligations that are in addition to any internal company rules or codes of conduct. To ease the burden on your team at every stage of development, consider creating a guided, quick-start assessment that quickly uncovers and scopes applicable governance requirements.
As Chris McClean explains in Episode 26 of Pondering AI, this type of guided investigation helps remove barriers and improve engagement with ethical scrutiny and regulatory compliance: Each assessment is a short scoping investigation built into each gate of the development lifecycle, from ideation to deployment, ensuring that important considerations aren’t overlooked through lack of awareness or a frenzied rush to deployment.
