How to Brief Your Board on AI in One Page

The board packet locks Friday and a director has asked for “something on AI.”

You could drop in a vendor deck, a glossary, or a product demo that never asks the board to decide anything. Directors would leave unclear on risk, on pilots, and on who decides, or they would green-light spending on hype. What fiduciaries actually need is a briefing they can read in three minutes and staff can talk through in five: what you are using, the privacy fences, what is in and out of scope, which decisions stay human, one ninety-day pilot ask, and what you will not claim.

What the sector already recommends

NTEN’s AI resource hub notes that AI has likely already come up in board discussions, not only in staff meetings. The same hub observes that nonprofits that rushed to adopt AI ran into unintended risks, and NIST’s Generative AI Profile notes that people can be unnecessarily averse to generative AI and miss its beneficial uses, so the brief should read as neither a sales pitch nor a scare sheet.

Independent Sector’s guidance for nonprofits adds a point about pace, which is to start slowly. NTEN’s Equitable AI Project Planning worksheet is a good frame for the pilot ask, since it tells teams to start from a single goal rather than a technology, say who will be involved and how, state the budget, and plan for long-term financial and staffing investment.

For the memo, ask for staff AI literacy plus one pilot tied to a real member or mission problem, not a vague strategy slide. ASAE’s own organizational AI policy makes the reserved-decisions block non-negotiable: humans verify and own outputs, protect member PII, use approved tools, be transparent when AI shapes member-facing work, and never rely on AI for final decisions about hiring, personnel, or member standing. The policy states the principle plainly: at ASAE, people make the decisions, not AI.

NIST’s AI Risk Management Framework gives directors shared vocabulary (Govern, Map, Measure, Manage), and the Generative AI Profile names confabulation, harmful bias, and automation bias. The framework also treats risk management as more than avoidance: it aims to reduce harms while making room for benefits, and it warns that trying to eliminate every negative risk can be counterproductive. One glossary card is enough. The FTC’s Operation AI Comply is the “will not claim” block: no invented productivity lifts, no member-growth promises, no “AI-powered” labels you cannot substantiate.

One number puts this in context. In a November 2024 survey by FSAE and the Association Societies Alliance, only 13% of the 305 respondents had an organizational AI policy, while marketing and communications was already the top reported AI use area. The tools are in use whether or not the policy exists. Your board is not behind for lacking a forty-page policy. It is behind if nobody has put one page in front of it.

The one-page template

Print or paste this shell into the packet. Fill it with your real tools, your real fences, and one real pilot. The sample entries in the shell are a teaching example, not a description of any real association’s setup. This is not legal advice.

TO: Board of Directors
FROM: [CEO / board liaison]
RE: AI at our association: status and 90-day ask
DATE: [packet date]

1) What we are using today
- [Tool A]: drafting member emails from approved website copy (staff review before send)
- [Tool B]: summarizing public agenda packets for staff prep (not confidential exhibits)
- Unknown personal-account use exists; we will bring it under an approved-tool rule

2) Data and privacy fences
- No member PII, AMS exports, donor notes, certification records, or confidential
  board exhibits in consumer AI tools
- Organization-approved accounts only for association work
- Humans review outputs before publish, send, or act

3) In scope / out of scope this quarter
- In: drafting and summarizing from approved or public text; internal brainstorming
- Out: member standing; dues changes; hiring; advocacy position language; legal strategy

4) Decisions reserved for humans
- AI may draft or summarize. The board and named staff decide. No automated final
  decisions on money, membership status, personnel, or public positions.

5) 90-day pilot ask
- Pilot: [one sentence]
- Owner: [name] | Success metric: [one measurable] | Stop rule: [when we halt]
- Budget: $[x] plus [y] staff hours | Report back: [next board date]

6) What we will not claim
- No unverified time-saved or growth numbers tied to AI
- No "AI-powered" member program labels without a truthful description
- No treating model output as fact without a human check

Next review: [date]. Staff acceptable-use note follows separately.

The five-minute talking sequence runs in this order: why the association is being intentional rather than panicked, what is in use, the fences, the reserved decisions, the one pilot ask, what you will not claim, and the next review date. Invite exploration questions about pace and about the risk of inaction as well as action. If your culture expects it, dry-run with the board chair.

Write the page before the packet locks

Ask three people (membership, communications, operations) which AI tools they touched this month and whether on a personal or organizational account. Cap the memo at the top three and note unknowns as shadow use. Open the six-heading shell. Write the fences in plain words and point to ASAE’s policy themes without pasting its text. List the reserved decisions: dues and pricing, member standing and discipline, hiring and personnel, advocacy positions, legal strategy, any fiduciary vote. Pick one pilot that is mission-aligned, off the critical path, and measurable, and name its owner, metric, stop rule, and budget even if the budget is zero dollars plus hours. Write the “will not claim” block. Add the page to the packet with an optional half-page NIST glossary card. Time the talking sequence aloud. Save the memo with a version date.

How a board brief loses the room

A vendor deck is not a brief. A glossary with no ask is not a brief either. Promising ROI you cannot measure will come back to you at the next meeting; use attributed figures or none. Confidential exhibits, AMS exports, and donor notes stay out of consumer chat even when you are summarizing for the board. If the memo never says who owns the final call, the room fills that gap with hype or fear.