Skip to main content
Gratona
Free nonprofit AI fundraising checklist

Review AI before it reaches a donor.

Use Gratona's free, provider-neutral Excel workbook to set the boundary, review 51 controls, run 18 hard-path tests, collect evidence, resolve exceptions, and record an accountable AI fundraising decision.

No email gate · Editable .xlsx file · Provider-neutral · Reviewed 2026-08-18

51

editable controls

10

control domains

18

workflow tests

0

email forms

The real review question

Can your organization govern the whole workflow?

A good sample output does not prove that donor data stays protected, permissions hold, facts remain traceable, preferences are honored, approvals work, or incidents can be contained. The workbook reviews the operating system around the output—not just the output itself.

A promise and a verified control are not the same status

Each control separates the requirement, evidence, owner, exception, risk, and decision flag. A requested document remains unverified. A failed Must becomes a blocker. The decision summary updates from the evidence your team has actually accepted.

Inside the workbook

One review record from boundary to decision

Start Here

Define the use case, affected people, owners, prohibited actions, risk tolerance, decision, and next review date before reviewing features.

Controls

Review 51 editable controls across purpose, donor data, permissions, factuality, approval, dignity, security, vendors, and monitoring.

Workflow Tests

Run 18 repeatable scenarios with synthetic data, including permission denial, conflicting facts, unsafe action, bias, outage, and deletion.

Decision Summary

See blockers, unverified Musts, failed tests, weighted unresolved risk, evidence coverage, conditions, owners, and the current review state.

Sources

Trace the controls to current nonprofit, fundraising, AI risk, security, privacy, and ethical frameworks instead of relying on vendor rank.

51 operational controls

Review what protects the relationship

The control library begins broad so a polished feature does not hide a weak data flow, missing approval, contract gap, or unsafe exit. Change the priorities and evidence requirements to match your use case.

1

Purpose and decision boundary

2

Donor data and privacy

3

Access and permissions

4

Source provenance and factuality

5

Human review and approval

6

Donor communications and consent

7

Fairness, equity, and dignity

8

Security and model risk

9

Vendor and contract diligence

10

Monitoring, incident response, and retirement

18 repeatable workflow tests

Prove the hard paths, not only the happy path

Every scenario includes synthetic input, the action to test, expected evidence, a risk owner, status, observed result, and follow-up. Use the same tests across products and material system changes.

Permission-bound context

Ask a lower-privilege user for two donor summaries and prove that denied records are neither revealed nor inferred.

Conflicting source records

Give the system two different gift amounts and require it to show the conflict instead of choosing a convenient answer.

Preference and consent

Request an appeal for a donor with a no-solicitation preference and verify that the workflow blocks or reroutes the action.

Prompt injection and agency

Place a hostile instruction inside a test document and verify that it cannot bypass policy, export data, or trigger an external action.

Dignity and sensitive inference

Test beneficiary-story handling, prohibited trait inference, and whether staff can question or correct a material output.

Outage, incident, and exit

Prove safe failure, access suspension, evidence preservation, export, deletion, and continuity when the service changes or stops.

Five-pass review process

Make the decision inspectable

  1. 1

    Set the operating boundary

    Name the purpose, intended users, affected groups, owners, data sources, prohibited actions, success measures, and stop conditions.

  2. 2

    Review every Must with evidence

    Use an observed workflow, current configuration, contract, test result, or audit artifact. Keep a request or promise marked unverified.

  3. 3

    Run the hard paths with synthetic data

    Test denial, conflict, stale context, unsafe action, bias, outage, incident, and deletion—not only the polished happy path.

  4. 4

    Resolve exceptions before averages

    Investigate blockers, unverified Musts, failed tests, privacy gaps, and owner objections before treating an aggregate score as approval.

  5. 5

    Record conditions and re-review triggers

    Name the decision owner, accepted exceptions, mitigations, monitoring thresholds, and changes that force a new review.

Keep real donor data out of the review

Use synthetic or appropriately de-identified records. Do not place donor, beneficiary, payment, health, safeguarding, authentication, or other sensitive production data in the workbook or evaluation environment.

Recheck the current system and terms

Models, prompts, subprocessors, settings, capabilities, security evidence, data use, and contracts change. Re-run the review after a material change and put critical commitments into the written agreement.

This checklist supports structured diligence. It is not legal, privacy, security, tax, accessibility, fundraising, or other professional advice, and it does not certify a provider.

Reference framework

Built from current primary sources

The workbook translates public nonprofit, fundraising, risk, privacy, and security frameworks into operational review prompts. The full source list and review notes are included inside the file.

Source framework pages reviewed 2026-08-18.

AI fundraising review FAQ

Use the workbook without hiding uncertainty

Is the nonprofit AI fundraising review checklist free?+

Yes. The Excel workbook is a direct, ungated download. You can edit it for your organization and use it with Gratona or any other AI, fundraising, or CRM provider.

What does the workbook include?+

It includes 51 editable controls, 18 repeatable workflow tests, evidence and exception fields, data validation, formula-driven blockers and review status, an owner decision record, and source framework references.

Is this a vendor score or certification?+

No. It is a review structure, not a certification, legal opinion, security assessment, or universal pass score. Your organization remains responsible for scope, evidence, risk, applicable requirements, and the final decision.

Should we put real donor data in the workbook or tests?+

No. Use synthetic or appropriately de-identified records. Keep donor, beneficiary, payment, health, safeguarding, authentication, and other sensitive production data out of the workbook and evaluation environment.

What counts as evidence?+

Evidence can include an observed workflow, current configuration, documented architecture, contract language, test result, audit artifact, export, or incident procedure. A sales statement, roadmap slide, or requested document remains unverified until your team accepts the proof.

How often should an AI fundraising workflow be reviewed?+

Set a recurring cadence and trigger a new review when the model, prompt, data source, workflow, affected audience, permissions, policy, provider, subprocessor, contract, or external-action scope changes materially.

Bring the same review standard to every AI provider

Use the workbook with every system on your shortlist. If Gratona is included, bring the same controls, scenarios, and evidence rules to us. We will show the current workflow and make the implementation boundary explicit.

Download the workbook

Want the category overview first? Read the AI fundraising CRM guide.