Playbooks · 6 min read
Nonprofit DMs: Four People Write, and One Needs Help
Start with the rule, because everything else follows from it.
Never let an automation ask for money from somebody who messaged about something else.
It is the single most tempting configuration in this vertical and the one that does the most harm. A donate line appended to every reply looks like good fundraising discipline. It is fine on a reply to "how do I give?" and it is grotesque on a reply to somebody asking whether your shelter has a bed tonight.
The inbox is at least four different people
A charity account, unusually, is written to by groups whose needs have nothing in common:
Wants to give you something
Wants something from you
They arrive in the same inbox, under the same appeal post, frequently using the same words. "Can you help?" is a donor offering and a person asking, and it is written identically by both.
An account that adds a donation link to every reply will eventually send one to somebody who wrote to you because they had nowhere else to write.
So the donation ask goes in exactly one automation — the one triggered by donation words on a donation post — and nowhere else in the account. Not as a signature, not as a P.S., not in the volunteer template.
The message from somebody in difficulty
This is the category that decides whether you should be automating at all.
People message charities in crisis. About money, housing, a child, violence, an addiction, an animal, a death. Sometimes at length, sometimes at 3am, sometimes in a way that makes clear they have been building up to sending it.
No automated reply is acceptable to any of those. Not a warm one, not a signposting one, not a "thanks for reaching out, someone will be in touch." A form response to somebody in trouble tells them what they were afraid of, which is that nobody is there.
And then somebody has to read that list, every day, at a known time, as a named person's job. Automation in this vertical is only defensible if that person exists. If they do not, you are building a machine that answers donors quickly and leaves people in difficulty in a queue nobody opens.
Never say anything about tax
A second hard line, and one that nonprofits cross without noticing because the information feels administrative rather than risky.
Do not put anything about tax treatment into a template. Not what a donation qualifies for, not what a donor can claim, not what a receipt entitles them to, not "tax-deductible" as a phrase. Whether any of that is true depends on where your organisation is registered, where the donor is, what kind of donation it is, how much, and rules that vary by jurisdiction and change.
An automation stating it is your organisation making a statement about a stranger's taxes to that stranger, at scale, unsupervised. Point instead: "our receipt process and the details on it are here", with a link to a page your finance people own and can correct in one place.
The same restraint applies to what a donation buys. "Your donation feeds a child for a month" is a claim, it will be quoted back to you, and it is one your programmes team should be writing, not a template author. If you use those lines, take them from published material rather than inventing them in the builder.
Campaign spikes, and the appeal that goes wrong
Fundraising traffic is not flat. An appeal, a disaster response or a piece of coverage produces a wall of comments in hours.
Set a rate budget before the appeal goes up. A comment stays answerable for seven days, which comfortably outlasts the spike, so a cap costs you nothing and stops a legitimate campaign from looking like a coordinated flood.
Leave takeover alone and understand what it is doing for you. The moment anyone on your team replies by hand, automation is out of that thread permanently. During a crisis appeal, when three people are in the inbox at once, that is the setting standing between a communications officer's careful reply and a cheerful template landing underneath it forty seconds later.
What is genuinely safe, and what does not exist
The automatable volume is real: how to donate and where, whether you accept goods and what you need right now, volunteer sign-up, where you work, what your programmes are, how to reach your office, where the annual report is, how to organise a collection drive.
What does not exist here, stated plainly. No payment links and no donation processing — the automation can point at your giving page and cannot take a rupee. No receipts. No donor CRM, no sync into a fundraising system; leads leave as a CSV. No scheduled appeals, no sequences, no broadcasts, and no way to message anyone who has not messaged you first. No WhatsApp. It does not post or schedule anything, so it will not run your campaign.
Official Graph API only: seven days on a comment, twenty-four hours inside a DM thread, restarted only by their next message. A donor who asks on Friday and comes back Monday is out of reach in between, which is the argument for putting the giving page in the first reply rather than the third.
Templated replies are free and unlimited and a credit is spent only when the AI writes a new reply — which for a small organisation means the entire donor-logistics tail costs nothing at campaign volume. Free tier is 100 credits, no card; packs start at ₹499.
Build the stop list, name the person who reads it, and only then build one automation on your donate post. What gets blocked and why is the page to read after the first week; the paced setup guide is the right order for everything after that.
