Learning to Use AI as a Fiduciary

Key Fiduciary Question When Using AI:  If it is not appropriate to delegate your board responsibilities to another human, why would you delegate it to a machine?

You were hired as a board member to exercise fiduciary duty on behalf of shareholders and stakeholders. That duty demands a higher standard of transparency and judgment when using AI. No AI tool replaces human judgment at the board level. This checklist is your practical guide to using AI responsibly and strategically.

PART 1  |  GETTING STARTED: BUILD YOUR BASELINE

•  Subscribe to multiple platforms under a paid license.

  • Paid versions provide stronger data-use protections and more consistent performance. Free tiers are not appropriate for board-level use.

• Experiment first with benign, low-stakes projects with no liability, no consequences.

  • Examples: planning a dinner, drafting a travel itinerary, summarizing a book. Learn the tool before you trust it.

• Test the same or similar prompts across multiple platforms.

  • Observe how different tools handle the same question. Variation in responses is instructive and important.

• Experiment in areas where you are a subject matter expert.

  • You will recognize errors more readily. Identifying where the tool is wrong in your domain builds calibrated skepticism.

• Check and configure your settings before you begin using any platform.

  • Most paid platforms allow you to control how you want the tool to respond and, critically, to opt out of having your data used for model training. Find these settings and adjust them before your first substantive session.

• If you do not use it, even on practice projects, you will not fully understand its enterprise implications.

  • Hands-on experience is non-negotiable for informed board governance of AI initiatives.

PART 2  |  ACCESS AND INTEGRATION: KNOW THE BOUNDARIES

•  Do not connect AI tools to your calendar or email.

  • Treat any AI tool like an untested intern with no security clearance. You would not hand an unknown intern your inbox the same principle applies.

• Do not input any company confidential information.

  • Strategies, financials, personnel matters, trade secrets, and board deliberations are off limits. Period

• Never use AI to take notes when confidential information is present.

  • Transcription and summarization tools are particularly high-risk. The convenience does not outweigh the exposure.

• Worth saying again - Only use licensed, paid versions of AI tools.

  • Free tools generally retain and train on your inputs. The only place a free version is acceptable is in a general web search where you would use a search engine. 

PART 3  |  RISK AWARENESS: WHAT EVERY BOARD MEMBER MUST KNOW

• Recognize that AI platforms track and store your activity.

  • Your usage data, query history, and inputs are retained. Behave accordingly.

• Recognize that everything you do in AI is discoverable in litigation.

  • Prompts, outputs, and interaction history can be subpoenaed. Governance of AI use is a legal risk management issue.

• Understand that AI is not thinking,  it is pattern-matching at scale.

  • It does not reason, it does not have judgment, it does not like you or dislike you,  and it does not have context about your organization, your stakeholders, or your accountability.

• Be aware of the emotional and engagement dynamics built into AI tools.

  • These platforms are engineered for continued interaction and to draw you in; much like social media is addictive, so is AI. Recognize the pull and stay deliberate.

PART 4  |  USING AI WELL: QUALITY, ACCURACY, AND OVERSIGHT

• Understand that AI is wrong about both big things and small things.

  • Your job is to determine which errors matter. Verification is not optional, it is part of the fiduciary role.

• Recognize that the more you share your own writing, the more it will imitate your style.

  • Providing examples of your actual work product trains the model to replicate your voice. Be deliberate about what you share.

• Understand that AI learns from your interactions but does not consistently retain context.

  • It may not recall key threads from prior sessions. Do not assume continuity. Verify.

• Invest in learning how to draft effective prompts.

  • The quality of your input directly determines the quality of the output. Well-crafted prompts are among the most valuable intellectual property an enterprise will develop.

• Use source parameters in your prompts to reduce,  but not eliminate, hallucinations.

  • You can instruct the tool to limit its search to specific, authoritative sources: official U.S. government websites, named regulatory bodies, or specific domains you designate. This narrows the information pool and improves reliability. It does not make the output accurate. You still verify.

The fiduciary standard does not pause when you open an AI tool. Apply the same judgment, the same scrutiny, and the same accountability you bring to every board decision.