How to Turn Staff AI Questions and Concern Emails Into a Nonprofit AI Risk Register With NotebookLM

Many nonprofits are already using AI in small, unofficial ways, but the rules around donor data, confidential notes, public statements, and staff review often lag behind. This workflow helps you turn real internal material into a usable governance artifact instead of starting from a generic template. It is for nonprofit executives, operations leads, development leaders, board administrators, and policy owners who need something practical that staff can actually follow.

Editorial guardrail: Use AI to sort and summarize internal concerns. A staff member should review any output that affects privacy, donor handling, public messaging, HR, beneficiary records, or board oversight before it is treated as policy.

What you need

• A set of staff AI questions, concern emails, chat excerpts, or meeting notes
• Any existing privacy, donor-data, confidentiality, or acceptable-use policies
• A Google account for NotebookLM, with desktop preferred
• Someone on your team who can confirm whether the concerns listed are current and real

How to capture or gather the source material

• Export the relevant email threads as PDFs, or copy the messages into a clean Google Doc with dates and senders removed if you want to reduce noise.
• If concerns came up in meetings, combine those notes into one short document called AI Questions and Concerns.docx.
• Add one file with your current rules, even if it is incomplete. The point is to compare worries against real policy, not just summarize complaints.
• Remove personal details that are not necessary for the policy discussion. Keep the scenario, the task, and the risk.

The fastest workflow

1. Create one notebook and upload the concern emails or notes plus your existing policy material.
2. Ask NotebookLM to group the concerns by workflow, data type, and risk level. Good categories include donor communications, grant writing, volunteer management, HR, public content, board materials, and spreadsheets.
3. Ask for a risk register with columns for workflow, example concern, likely harm, current policy coverage, recommended control, and owner.
4. Use a second pass to separate items that need a policy rule from items that only need staff training or better templates.
5. Export the final register into your governance notes and assign owners for review.

Tool-specific instructions

Primary path: NotebookLM

• NotebookLM works well because it can compare raw concern material against the policy files you upload and point back to those sources with citations.
• Treat the concern file as source material, not truth. The goal is to capture patterns and edge cases, then compare them against real rules.
• Ask for a table or structured register, not a narrative summary. Governance artifacts are more useful when they are sortable and assignable.
• If one email contains several issues, tell the model to split them into separate rows.

Fallback options

Claude fallback

• Upload the concern files and policy files into a Claude Project and ask for a structured risk register with one row per issue.
• Claude is a good fallback when you want to refine the wording over several iterations inside the same project context.

ChatGPT fallback

• Upload the files to ChatGPT and ask for a risk register in plain text or CSV-style rows.
• This path is useful when you want quick sorting and then plan to clean the final table manually.

Copy and paste prompt blocks tailored to the workflow

Primary prompt

{
  "task": "Create a nonprofit AI risk register from uploaded concern emails and policy documents.",
  "required_columns": [
    "Workflow",
    "Example concern",
    "Data involved",
    "Likely harm if mishandled",
    "Current rule or missing rule",
    "Recommended control",
    "Owner"
  ],
  "instructions": [
    "Use only uploaded sources.",
    "Treat emails and notes as examples, not final policy.",
    "Split combined concerns into separate rows.",
    "Flag items that need policy, training, or simple template fixes."
  ],
  "final_request": "End with the five issues leadership should review first."
}

Fallback prompt

{
  "task": "Turn the attached staff AI concerns into a practical governance memo.",
  "required_headings": [
    "Most common concern categories",
    "High-risk workflows",
    "Policy gaps",
    "Training fixes",
    "Recommended next decisions"
  ],
  "instructions": [
    "Keep the language plain and operational.",
    "Do not invent legal requirements.",
    "Make a clear distinction between a missing rule and a missing example or training aid."
  ]
}

Quality checks

• Check that every row describes one concrete issue, not a broad category that no one can act on.
• Make sure the register separates missing policy, missing review, and missing staff training. Those are not the same problem.
• Verify that the highest-priority issues really are the ones involving sensitive data or public-facing risk.

Common failure modes and fixes

• The output turns into a long essay: Ask for a risk register table with fixed columns and one row per concern.
• Too many rows are duplicates: Tell the tool to merge identical issues but keep distinct examples in the notes column.
• The register feels abstract: Add one more source with actual current workflows so the model can tie each concern to a real task.
• The team cannot tell what to do next: Ask for a final sort by policy change, training change, and low-priority backlog items.

Sources Checked

• Google NotebookLM Help, Learn about NotebookLM - Computer. https://support.google.com/notebooklm/answer/16164461?co=GENIE.Platform%3DDesktop&hl=en. Accessed 2026-03-27.
• Google NotebookLM Help, Add or discover new sources for your notebook - Computer. https://support.google.com/notebooklm/answer/16215270?co=GENIE.Platform%3DDesktop&hl=en. Accessed 2026-03-27.
• OpenAI Help Center, File Uploads FAQ. https://help.openai.com/en/articles/8555545-file-uploads-faq. Accessed 2026-03-27.
• Anthropic Help Center, How can I create and manage projects?. https://support.claude.com/en/articles/9519177-how-can-i-create-and-manage-projects. Accessed 2026-03-27.
• Candid, Getting started on a responsible AI use policy for nonprofits. https://candid.org/blogs/how-to-create-responsible-ai-use-policy-for-nonprofits/. Accessed 2026-03-27.
• BoardEffect, Nonprofit leaders share their thoughts on AI. https://www.boardeffect.com/blog/leaders-thoughts-ai/. Accessed 2026-03-27.

Quarterly Refresh Flag

Review this article by 2026-06-25. Re-check product features, upload flows, and nonprofit workflow references before updating or republishing.