Where AI belongs in each movement

This is the guide you keep open while you work. The first lesson in this module sets a single test: a tool may prepare the inviter and it may serve the record, and it must not stand in for the relationship or decide who a person is. That sentence is easy to agree with and hard to apply at ten o'clock on a Tuesday when a draft is due. What follows takes the test movement by movement, so that when you sit down to use a tool for something specific, you already know which side of the line that task is on.

Two habits carry through every movement below. First, a task belongs to a tool only if you would still own the outcome with your name on it once it reaches a donor. Second, if you would not want to tell the donor how a piece of work was made, that reluctance is information, not paranoia, and it usually means the work does not belong to the tool.

None of this is a claim that tools are dangerous by nature. A spreadsheet formula, a mail-merge, a search engine are all tools that changed fundraising and none of them decide who a donor is or say the words of an invitation. The newer generation of tools is more capable, which is exactly why the line matters more, not less. The movements below draw that line in the same place the Framework already draws it: preparation and the record on one side, relationship and judgment on the other.

Pause

Pause is preparation. It is the movement where you get your own head right before you ask anything of another person, and preparation is largely mechanical: notes, drafts, rehearsal.

Use it for: drafting your own preparation notes ahead of a visit or a call, so you walk in with your thinking organized rather than scattered. Summarizing your own meeting notes and call records from earlier in the relationship, so you are not re-reading six months of history the morning of. Building a short talking-points outline from those notes, the kind you glance at once in the parking lot and then set aside. Rehearsing a hard conversation by having a tool play the donor's likely questions or objections back at you, so you hear your own answers out loud before the donor does, and can smooth the parts that sound stiff.

Not for: deciding whether you are ready. Readiness is a judgment about your own clarity, nerve, and standing with this particular person, and that judgment is yours to make; a tool has no access to what you actually feel walking toward that door.

The rule: use a tool to get your words and your notes in order, but never let it tell you when you are ready to walk in.

See

See is where a fundraiser learns who a person is, not who a file says they are, and it is the movement where a tool does the most good and the most quiet harm.

Use it for: summarizing public information and first-party records the donor would expect you to hold, things like their own giving history with you, their stated interests, prior conversation notes, and public materials such as a company bio, a foundation's published priorities, or a news item they would recognize as public. Pulling a long record into a short, readable brief before a visit is a legitimate use of a tool's time-saving, and so is turning a wealth-screening export into a plain-language summary a fundraiser can actually read in five minutes instead of scanning a spreadsheet.

Not for: inferring protected traits, health status, family conflict, or wealth from thin signals. A tool asked to guess at a person's health, marital trouble, or net worth from a few data points will produce something confident-sounding and often wrong, and a wrong guess about a person's life does real damage even when nobody outside the office ever sees it. Also not for pasting a donor's giving history or your own notes about them into a tool whose terms let it train on what you send it; that donor data is not yours to hand to a system that keeps it. Confirm what a vendor's terms actually allow before anything donor-specific goes in, the same way you would before sending a file to any outside party. Consent and custody rules apply here exactly as they do to any other disclosure of donor information: know what you agreed to tell the donor about how their data is used, and keep the data inside the systems you told them it would live in. The companion guide on screening and verification in the prospect research module covers the research discipline itself; this guide only marks where a tool sits inside it.

A capacity score or an affinity flag from a screening tool is a starting point, not a fact. It opens a question a human has to ask, about whether the number matches what you know, what the board knows, and what the relationship has actually shown so far. Treat the score as a prompt to verify, not as a conclusion to record.

The rule: a tool may organize what you already know and are entitled to hold; it must never guess at what a person has not told you, and donor data does not leave your custody for a system that keeps it.

Discern

Discern is the test itself: is this invitation true, and is it ready. That test is a human judgment built from relationship, timing, and mission fit, none of which a tool can weigh.

Use it for: listing what you do not yet know. A tool can be useful at surfacing gaps, the open questions in a donor's file, the readiness signals you have not checked, the parts of the mission fit you have not confirmed, the last conversation nobody logged. A clear list of unknowns, turned into three or four questions you go find answers to before you move forward, is a genuinely good use of a fast, tireless assistant.

Not for: answering the readiness question itself. Whether an invitation is true to the mission and ready for this person right now is a judgment that draws on relationship history, board knowledge, and timing that lives in your head and your colleagues' heads, not in a data set. The readiness tests belong to the fundraiser and the team around the fundraiser.

The rule: let a tool find the gaps in what you know; the discernment that closes those gaps stays with the person who will make the ask.

Invite

Invite is the movement where the mission connection gets named out loud and the donor is left genuinely free to say yes, not yet, or no. The words matter because they are a promise, and a promise needs an author who will stand behind it.

Use it for: tightening your own draft. Once you have written the invitation in your own words, a tool can help you cut it down, catch a sentence that runs long, or check whether the reading level matches how this donor actually talks. That is editing, applied to something you already wrote.

Not for: generating the ask itself. An invitation is not a paragraph that reads well; it is a specific person's honest statement of why this gift, from this donor, matters to this mission, and that statement has to come from someone who will sit across the table and mean it. A generated ask can sound sincere without being sincere, and a donor who senses that, even without naming it, trusts you a little less afterward. This is the one place in the Framework where the difference between a well-written sentence and a true one decides whether the invitation works at all.

The rule: use a tool to sharpen your own words, not to supply the words you have not written yet.

Receive

Receive is where every answer gets honored, including no and not yet, and where the donor learns whether your gratitude was real or procedural.

Use it for: acknowledgments that follow a template, as long as a person reads and approves each one before it goes. A template gets the tax language and the gift details right every time; a human read catches the acknowledgment that is technically correct and actually tone-deaf, the naming error, the gift that deserves more than the standard line, the donor whose circumstances this month call for a phone call instead of a letter.

Not for: automated replies to a no or a not-yet. A donor who declines or asks for time has just told you something about the relationship, and an automated response tells them the message did not reach a person. Every no and every not-yet gets a human reply, even a short one, and the reply is where you find out whether the no was final or the timing was wrong.

The rule: templates may carry the routine language, but a person reads each acknowledgment before it ships, and the thank-you has to be true, not just correctly formatted.

Accompany

Accompany is the promise kept after the gift, the movement most exposed to shortcuts because impact reporting is slow, real evidence is scattered, and a tool can produce something polished in a minute.

Use it for: drafting impact reports built from your organization's real data, real outcomes, real numbers, with a named human owner who checked the facts before anything went out. A draft assembled from verified figures and then reviewed by a person who knows the program is a legitimate time-saver, and it can shorten the gap between a program's real results and a donor actually hearing about them.

Not for: synthetic stories. No composite "beneficiaries" presented as real people, no invented quotes attributed to someone who did not say them, no fabricated scene dressed up as a case study. A donor who later learns that the family in the newsletter was not a real family does not just lose trust in that one story; they lose trust in every report you send after it. When an image or a story in a report was generated rather than photographed or recorded, say so, plainly, in the piece itself.

The rule: report real impact from real data, own it with a named person's name, and disclose anything that was generated rather than documented.

Multiply

Multiply builds capacity in others: staff, board, volunteers, the next fundraiser who has to carry this work when you move on. This is the movement where a tool is least risky, because the output is infrastructure, not a relationship.

Use it for: training staff and board on the standard in this guide. Using a tool to draft a first-pass checklist, a policy outline, a slide deck for a board briefing, or an onboarding document that a person then edits, owns, and issues. Building the scaffolding of a good practice is exactly the kind of task a tool speeds up without touching a donor relationship, and a shop that never uses it for this kind of internal-facing work is probably spending staff time it does not have on documents no donor will ever read.

Not for: skipping the sign-off. A policy or a checklist that nobody in leadership reviewed and approved is not yet a standard; it is a draft with a title on it. Multiply only works when it builds something durable, and durability comes from a named leader taking ownership of the document, not from the document existing.

The rule: use a tool to speed up the drafting of standards and training materials, and let the leader who will enforce them sign what goes out.

The read-aloud test

Here is the test to run on anything you are not sure about, in every movement above. Picture the donor in the room. Explain, in one or two sentences, how the piece in front of you was made: which parts you wrote, which parts a tool drafted or organized, what data went in, who checked it before it reached them. If that explanation embarrasses you, or if you find yourself reaching for a vaguer version of the truth to say instead, do not send the piece as it stands.

This is not a legal standard and it is not a data-privacy standard, though it will usually keep you inside both. It is a standard about whether the relationship stays honest. A donor does not need to know the mechanics of every tool in your shop, any more than they need to know which mail house folded the appeal letter. They do need to be able to trust that the thank-you was felt, the ask was yours, and the story about the child who benefited from their gift was a real child. If a use of a tool would survive being said out loud to the donor, plainly and without softening, it probably belongs in your practice. If it would not, it does not belong there yet, no matter how much time it saves this week, and the time saved this week is rarely worth the trust spent to get it.

Keep this guide next to the policy template in this module. The policy names what your shop will and will not do; this guide is what you consult in the moment, movement by movement, when a specific task is in front of you and you are deciding which side of the line it falls on.

Evidence and adaptation note

This is a working tool, not a universal benchmark. Replace every bracketed field and example number with your organization's facts. Composite cases are labeled; their figures illustrate the method and should not be cited as sector results. Check legal, tax, privacy, employment, and accounting language against current guidance and your jurisdiction before adoption.

Primary references for review

Use these as verification starting points. The named reviewer still owns the final interpretation and must confirm that each source is current.

• AFP Code of Ethical Standards: https://afpglobal.org/ethics
• AFP Donor Bill of Rights: https://afpglobal.org/donor-bill-rights
• Apra, Statement of Ethics: https://www.aprahome.org/
• NIST AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework