The craft · AI-for-X · boardsFor chairs, NEDs, and corporate secretariesPairs with /comparison/advisor-vs-board

AI-for-X · boards and directors

AI in the board room: pre-read throughput, fiduciary stays human.

A serious director arrives at the meeting having read the pack, talked to two managers, and formed a position. AI helps with the read. It does not enter the meeting, does not see MNPI under loose terms, and does not vote. This manual is the line, the workflow, and the second-brain stack a director can run alone.

AudienceChairs, NEDs, corporate secretaries
First deploy~one week
Hard floorMNPI & fiduciary duty
See the scope diagram Founder-side governance
Director reading the board pack with AI support.

What this work actually is

AI for directors is throughput on the read-in. Fiduciary judgment cannot be delegated and is not.

AI for boards is the deliberate use of LLMs on the work that surrounds the meeting: digesting the pack, drafting the briefing note, researching governance and comparables, summarizing minutes, tracking actions. It runs on the director's tenant. It does not see MNPI under consumer terms. It does not vote, sign, or speak in the meeting.

The hard floor is MNPI handling, fiduciary duty, and the recordable governance log. Workflows that respect all three compound. Workflows that do not generate insider-trading inquiries.

Fiduciary duty is non-delegable. A director who outsourced the read outsourced the duty.

The scope diagram

Three columns. The third one is regulator-defined.

Human owns

Where the director sits and AI does not enter.

  • The vote and the rationale.
  • The fiduciary read on the management team.
  • Live judgment in the meeting.
  • Conflicts disclosure and recusal.
  • Direct conversations with management between meetings.
  • Anything that creates personal director liability.
Non-delegable · signed by the director
AI assists

Where AI multiplies the pre-read.

  • Pack digestion: turn 200 pages into a 4-page brief.
  • Comparable-company and peer-board research.
  • Governance and regulatory background reading.
  • Briefing notes from the director's own observations.
  • Post-meeting action tracking against the minutes.
  • Question-stack generation for management one-on-ones.
Drafted by AI · read and verified by the director
Never AI

What does not enter a model.

  • MNPI under consumer or shared terms, ever.
  • Internal investigations, whistleblower files, executive comp deliberations.
  • Live deal terms during a transaction.
  • Personal data of executives and employees.
  • Litigation strategy and privileged counsel communications.
  • Anything that breaches the company's information barrier policy.
Hard floor · MNPI rules & company policy
Figure 01 · scope diagram Drawn version sits here in v2: HUMAN (vote) · AI ASSISTS (pre-read) · NEVER (MNPI floor). Used as the cover page of the director's personal AI policy.

What changes when this is done well

~3h
Per meeting, reclaimed
From pack digestion and pre-read research without losing the deep section reads.
2x
Question quality
Question stack pressure-tested before the meeting; sharper exchanges with management.
0
MNPI in shared models
Tenant-only for board material; consumer chats never see the pack.
1
Personal AI policy
Director's written policy filed with the company secretary alongside D&O.

What you need before you start

Five prerequisites. The first one is filed before the next pack arrives.

01 · A personal AI policy filed with the secretary

One page. What you use AI for. What you do not.

Filed alongside your D&O and your conflicts declaration. Refreshed annually. The policy is what makes the workflow defensible if anyone asks.

02 · A tenanted model with no-train terms

The vendor terms decide whether MNPI is safe.

Consumer chats are out for board material. Default to vendors with a data-processing addendum and a tenant you control.

03 · A redaction discipline

Names, deal terms, and figures stripped before paste.

Even on the tenant. The discipline is what protects you if vendor terms change without you noticing.

04 · A trading-window awareness

You know the company's open and closed windows.

AI use does not change the rules; it can amplify the consequences. Pre-read prompts during a closed window deserve extra care, not less.

05 · A pack-handling rule

How and where you store the pack outside the secure portal.

Most directors mishandle the pack as soon as it leaves the portal. The rule covers download, AI processing, and deletion within a fixed window.

The split workflow

Six director workflows. Each shows what stays with you and what AI handles.

01

Human owns

Read the consequential sections of the pack yourself.

Audit committee report, going-concern, related-party, executive comp, litigation update. The director reads these end-to-end, no shortcut.

AI assists

Digest the rest of the pack into a four-page brief.

Routine sections summarized, KPI movements highlighted, prior-period diff surfaced. The director reads the brief and decides which routine section deserves a deeper read.

02

Human owns

Form your own position before the meeting.

What you think, why, what you would push on, where you would defer. Written by you, in the file, before any prompt.

AI assists

Pressure-test the position with the strongest counter.

Standing prompt: read the position as the strongest critic; name the weakest claim, the data you have not seen, the question you should ask. The director edits the position; AI does not write it.

03

Human owns

Talk to management between meetings.

Direct conversations with the CEO, CFO, head of audit. Owned by the director. Not summarized into a shared model after the call.

AI assists

Generate a question stack from the pack.

Top fifteen questions a serious director would ask, ranked by materiality. The director picks five. The list is not handed to management.

04

Human owns

The meeting itself.

The room is the work. AI is not in the room, not transcribing in the background, not running real-time summary. The director shows up and listens.

AI assists

Briefing notes for the chair before the meeting.

Where the chair, AI digests management's prior commitments, surfaces the open items, and drafts the agenda flow. Chair reads, edits, owns.

05

Human owns

The vote and the rationale.

The director's vote is the director's. Always. The rationale is in the minutes and in the director's personal file.

AI assists

Personal voting log against prior decisions.

Track how this vote sits against your prior positions. Flag inconsistencies. The log is the director's, not the company's.

06

Human owns

Action follow-through with management.

The chase between meetings. Owned by the chair and the corporate secretary in formal terms; owned by every director on the items they sponsored.

AI assists

Post-meeting action tracking against the minutes.

Pull the minutes, list the actions, surface what slipped from the prior meeting, prepare the chase note. Chair signs.

Figure 02 · pre-read swim lanes Six-step swim-lane diagram lives here in v2. Top lane: director. Bottom lane: AI. Meeting block in the middle is closed to AI.

How to know AI is hurting the seat

Six tells the duty has slipped.

Tell 01

A consequential section was read only via the AI summary.

The duty has slipped. Read the section end-to-end before the next meeting and restore the section-by-section discipline.

Tell 02

MNPI reached a consumer model.

A regulatory exposure. Disclose to the company secretary; review the leak; tighten the redaction discipline and the storage rule.

Tell 03

The position you brought to the meeting was generated, not chosen.

The throughput reversed. Reverse the flow: write your position first, then ask AI for the strongest counter.

Tell 04

The question you asked management was clearly assembled by AI.

You handed AI's question to management. Pre-read prompts surface candidates; you choose, edit, and own the question in the room.

Tell 05

The pack was processed during a closed trading window without extra care.

The window does not change the AI rules; it raises the cost of any leak. Tighten the storage rule and the deletion timing for window periods.

Tell 06

The personal AI policy has not been reviewed since you joined the board.

The policy is stale. Refresh against current vendor terms, current company policy, and current regulatory guidance.

Tools and tactics

A second brain shaped for a director seat.

The brain holds the pack history, the position log, and the action follow-through. It is the director's continuity across years.

The Second Brain · director seat

Stan's adapted stack for boards

One folder per board. Personal AI policy at the top. Per-meeting subfolder: pack digest, position memo, question stack, voting log, action follow-through. MNPI never enters the brain; redacted equivalents only. Annual review by the director against the policy.

  • One folder per board, per-meeting subfolder.
  • Personal AI policy filed with the company secretary.
  • Position memo before every meeting, written by the director.
  • Question stack generated, narrowed, and owned by the director.
  • Voting log with rationale, kept personally.
  • Annual review of the policy and the pack-handling rule.

Documented in full inside the engagement · teaser here

Tactic 02

The pack-handling rule

Download, process, delete within a fixed window. Storage location named. Redacted equivalents only into the brain. The rule survives a vendor change.

  • Download from secure portal only.
  • Process inside the tenant.
  • Delete within the window stated in the policy.

Tactic 03

The standing critic prompt

One paragraph used after every position memo. "Read this as the strongest critic. Name the weakest claim, the data I have not seen, the question I should ask before the vote." Used to sharpen the director's own writing.

  • Same prompt every time.
  • Used on your own writing, never on management's.
  • Refined when it lets a soft position through.

Tactic 04

The personal voting log

Date, item, vote, rationale in one sentence. Kept by the director, not the company. Reviewed against new votes to surface inconsistencies.

  • One row per vote.
  • Rationale in one sentence.
  • Reviewed quarterly against new votes.

Coming soon

Three rooms held open inside this manual.

Built for directors who want the policy and the stack pre-assembled.

In build

The Director's Personal AI Policy Pack

Policy template, pack-handling rule, position memo template, voting log, briefing-note template. Drafted with risk counsel.

Scoped

The Chair's Briefing Library

Named brief blocks for the chair: pre-meeting agenda flow, executive session, succession review, audit committee. Released when the personal pack has been stable for one quarter.

Scoped

The Annual Director AI Audit

A small structured engagement: read of three meetings of pre-read and follow-through, gap report against MNPI rules, fix list for the personal policy.

What this work is not

A board governs. AI does not.

This page makes the director sharper. It does not change who is liable.

The throughput is on the read-in. The fiduciary duty is on the seat. The comparison page sets the structural difference between governance authority and an advisor's read.

Read advisor vs. board →
Pull AI back out of the workflow when
  • A consequential section was read only via summary.
  • MNPI reached a consumer model.
  • The position was generated, not chosen.
  • The personal policy is stale.

When the seat question is the wrong question

Run the policy and the stack for one full board cycle.
If the read is still soft, the question is structural.

Application-gated. Personal reply within 48 hours.

Apply for advisory

Tier 01 from $2,500 · Tier 02 from $4,500 / month · All three tiers