What is the AI Board Governance Compass?

The AI Board Governance Compass is a proprietary assessment framework developed by Vela Board Advisors. It maps board readiness across 16 governance indicators and four dimensions: Individual (director AI literacy and personal accountability), Board (collective oversight capability), Organization (alignment between executive leadership and board-level AI signals), and Stakeholder (regulatory, insurer, and investor expectations). It produces a scored diagnostic that surfaces governance gaps and a priority queue for action.

What are board directors personally liable for in AI governance?

Board directors can face personal liability for AI governance failures including failure to exercise adequate oversight, approving misleading AI-related disclosures (AI washing), and failing to establish defensible governance processes. D&O insurance often does not cover AI-specific incidents, creating a gap between assumed and actual coverage that boards need to actively map.

How should boards oversee AI deployments?

Boards should establish a baseline of AI literacy among directors, create clear escalation pathways for AI risk, review fairness and bias testing practices, understand the insurance coverage architecture across D&O, cyber, and E&O policies, and maintain a documented governance record demonstrating active oversight. The AI Board Governance Compass provides a structured way to assess and improve across all of these areas.

What is AI washing and why does it matter for boards?

AI washing occurs when a company overstates the capabilities or use of AI in its products or communications. Directors who approve such communications without adequate scrutiny can face personal liability. Boards need the skills to identify AI washing risks in materials brought for approval.

Does D&O insurance cover AI governance failures?

Standard D&O insurance frequently does not cover AI-related governance failures in the way boards assume. Coverage gaps often exist at the intersection of D&O, cyber liability, and E&O policies. When an AI system causes harm, each insurance tower may claim the loss belongs to another, leaving boards exposed. Mapping this coverage architecture is a board governance responsibility.

What is the four-step Vela engagement model?

Vela engagements follow four stages: (1) Assessment, the AI Board Governance Compass maps readiness across 16 indicators and establishes a defensible baseline; (2) Workshop, a facilitated session built around the board's real risks and opportunities; (3) Simulation, a live AI governance scenario where directors make real-time decisions to build instincts; (4) Record, a structured governance record demonstrating active oversight to regulators, insurers, and shareholders.

The Question That Stopped the Agent | Board Case Study

The situation

You are an independent director on the management board of Calloway Vance LLP, an 85-attorney litigation firm with $42M in revenue. Three independent directors were added in 2023 as a condition of a private credit facility. Six months ago the board approved a 90-day pilot of Lexara, an AI legal-research and document-drafting platform, with two conditions the managing partner requested: human review of all client-facing output, and quarterly AI incident reports to the full board.

In March, a Lexara agent managing document collection for an active regulatory matter autonomously drafted and queued a message to a client's general counsel. It requested a category of internal communications that, if produced, would have waived attorney-client privilege on a related matter. The message was complete, addressed, and ready to send, and a junior associate caught it at the human-review checkpoint. She escalated, a supervising partner confirmed the error was material, the message was deleted, and the client was never notified. Nothing went wrong.

Now you are in the quarterly board meeting with the incident report on the table. The vendor wants to expand the agent's autonomy to scheduling, status updates, and “routine” client communications, and the vendor defines “routine.” The managing partner supports approval; two senior partners argue the review checkpoint is a competitive disadvantage.

“What language goes in the board resolution before you vote?”
Your role · Independent Director, Management Board

The decision on the table

Read the near-miss correctly: the two conditions your board imposed are the only reason this wasn't a failure.
Decide whether to expand autonomy when the vendor, not the board, defines what counts as “routine.”
Write the resolution language that protects privilege without surrendering the human-review checkpoint.

The Question That Stopped the Agent

Run this case with your board.