AI automation for NGOs and charities, and what it must not touch

The paperwork side of an organisation, and the line it does not cross

AI automation for NGOs and charities works on the paperwork side of the organisation and nowhere near the decisions: donor records and receipting, recurring giving, grant and funder reporting, beneficiary case files, volunteer coordination, and the fund accounting that keeps restricted money separate from general money. It does not decide who receives help, and it does not write about a beneficiary without a person reading what it wrote.

Charities arrive at this from one of two directions. Either fundraising has grown faster than the admin behind it, so money is coming in and nobody is being thanked properly, or the funders have grown more demanding, so a small team spends its evenings rebuilding numbers into somebody else's template.

The unglamorous half of this work is the half that pays. A receipt that goes out in the same minute as the gift, a restricted fund that reports itself and a case record two caseworkers can both see remove the reason your best people spend Saturdays on administration.

• The donor side: giving pages, receipts and certificates, recurring gifts, lapsed donor contact, appeals
• The funder side: grant pipelines, reporting packs, monitoring data collected in the format it will be reported in
• The programme side: beneficiary records, eligibility routed to a human committee, case notes, follow up
• The people side: volunteer rotas, shift confirmations, onboarding packs, staff and trustee reporting
• The money side: restricted and unrestricted funds, reconciliation, supplier payments, cost per programme

The week you actually have

No solutions here. Count the ones describing your last seven days.

• A donation lands and the receipt goes out days later, typed by whoever had a free hour
• Someone gave generously last year, has heard nothing since, and you notice while writing the annual report
• Donor records sit in a spreadsheet, an inbox and a bank statement, and the three do not agree
• Restricted money and general money are held apart in somebody's head rather than in the accounts
• A funder wants the report in their own template, so the same numbers get rebuilt by hand
• The grant deadline was known months ago and the report is still written in the final week
• Beneficiary records are paper forms in a box, and finding one case history takes a morning
• The same beneficiary is registered twice under two spellings of one name
• Volunteers are coordinated in a group chat nobody can search and nobody can hand over
• A volunteer arrives for a cancelled shift, or a shift goes uncovered because one message was missed
• Field staff send photographs and voice notes, and turning those into a report is a job in itself
• The monitoring data a funder asked for was never collected in the shape they wanted
• Every appeal goes to the whole list, because segmenting it would take a week nobody has
• Nobody can say what a programme costs per person served without a day of digging

The arithmetic nobody does

No industry averages here, and no benchmark for what a charity your size ought to raise. Take your own figures and do three sums. The third is the one boards find uncomfortable.

• Sum one: your average donation, multiplied by the number of people who gave once in the last two years and have not been contacted since.
• Sum two: the hours your team spends each week on receipting, reconciliation, funder reports and volunteer coordination, multiplied by what an hour costs you, multiplied by fifty two.
• Sum three: the value of a grant you did not apply for, or did not reapply for, because the reporting burden on the last one used up the time. Count it once, at the full value of the grant.

If the first sum is larger than what you spend on fundraising staff, the problem is not that too few people give. It is that the people who already gave were never asked again.

What gets built, and where a charity starts

The first build is whichever of those three sums came out largest. In practice it is usually the donor path, because it pays for everything after it: give, receipt, thank, report back, ask again.

The list below is what an organisation accumulates over a year or two, not a shopping list for month one.

• Donation and appeal pages that record one off, monthly and pledged giving against a named donor rather than into a total
• Receipts and tax or zakat certificates issued automatically, in the same minute as the gift
• Restricted, unrestricted and designated funds tracked separately and reportable without a rebuild
• Abandoned donation recovery, which works the way an abandoned basket does on a shop
• Donor journeys: thank you, then what the money paid for, then the next appeal
• Lapsed donor contact, segmented by what they gave to rather than only by when they last gave
• A shared record of donors, funders, volunteers and cases, with permissions set by role
• Grant pipelines, with deadlines, conditions and reporting dates held in one place
• Reporting packs assembled from the records: trustee papers, funder returns, annual report figures
• Beneficiary intake that captures the monitoring fields your funders will ask for, at intake
• Volunteer rotas, shift confirmations and reminders, and an onboarding pack that sends itself

What stays human when the records are people

Donor data and beneficiary data are the most sensitive records in the building, and they carry different risks. A mishandled donor record embarrasses you. A mishandled beneficiary record can put somebody in danger.

Nothing about an individual beneficiary is automated without a person in the loop. The system holds the record, prompts the follow up, flags the missing field and assembles the numbers. It does not decide eligibility, it does not write a case note, and it does not send a message about somebody's circumstances that no human has read.

• Eligibility, assessment and safeguarding decisions, which stay with your committee or your caseworkers
• Any message to or about an individual beneficiary, which a person reads before it leaves
• Consent, checked before any appeal or contact sweep runs, and where the rules are unclear the cautious reading wins
• Reporting obligations to funders, which are constraints on the design rather than an afterthought: what you promised to collect, keep and delete
• Any story used in an appeal, which needs written consent and needs telling with dignity
• Anything published in the charity's name, and anything that commits the organisation contractually

What changes with the country you fundraise in

Charity rules are among the least portable rules there are. Four things change the build depending on where the organisation is registered and where it raises.

• Registration and public reporting differ. What must be filed, how often, and how much is public varies by country, and that decides which numbers the system has to produce on demand.
• The giving instruments differ. Zakat and sadaqah carry eligibility and separation rules of their own. Elsewhere the equivalent question is tax relief and the certificate that proves it. Both need the fund kept apart at the moment the money arrives, not at year end.
• Payment rails differ. Card giving, bank transfer, mobile wallets and cash collected in the field leave different trails, and the reconciliation work is proportional to how many you accept.
• Contact rules differ. Where WhatsApp is the main channel, business messaging runs on opt in, on templates approved in advance, and on a window that closes after the donor's last reply. Where email leads, consent rules are stricter than most fundraising teams assume.
• Seasonality differs. Ramadan, Eid and year end giving concentrate much of the year's income into a few days, and a receipting process that copes in March will not cope in the peak week.

When a charity should spend this money on people instead

Some organisations should not buy any of this yet, and saying so is more useful than another paragraph of persuasion.

One more, and it is the awkward one. If your donor list was collected without a clear record of consent, the first job is cleaning it, and cleaning it will make the list smaller. That is the correct outcome.

• If the whole administrative load genuinely fits inside one person's week, the money belongs in the programme. This is worth building when the paperwork is displacing the work, not before.
• If intake is disorganised, automating it produces disorganised data faster. Fix what gets asked and recorded first, on paper if necessary, then build around the version that works.
• If you cannot yet say who is allowed to see a beneficiary record, do not start with case management. Start with the permission rules, because a record system built before that decision has to be rebuilt after it.
• If nobody internally will own the system, even for an hour a week, it will be installed and then quietly abandoned.

What a trustee should ask, and the decision no system gets to make

No price is published for this, and for a charity the reason belongs in the same breath as the number. The figure moves on how many funders you report to and in how many templates, whether your accounting software will let anything read the fund split, how many donor records have to be reconciled out of three places into one, and who maintains it after the first month. What that means for a board weighing it is that the comparison is not with a software budget. It is with the third sum above, which is the grant you did not reapply for because reporting on the last one used up the time.

The line here is a design rule rather than a disclaimer. Nothing decides who receives help. Eligibility is collected, checked for completeness and routed to a human committee, and a system that scored an applicant would be doing the single thing a charity cannot delegate. Nothing is written about a beneficiary and sent anywhere before a person has read it. A distressed caller hands it to a person at once. Any appeal, any public statement and anything issued in the organisation's name waits for human approval, because a charity's name is most of what it owns.

A charity with a capable administrator can build the receipting in-house, and that is often the right place to start, since a receipt going out in the same minute as the gift needs a payment notification and a template rather than a project. Running it with your own team also surfaces the reconciliation problem honestly, which is where the real work turns out to be. What is harder to hold internally is fund split reporting, because restricted money held apart in somebody's head is a governance risk and moving it into the accounts is exacting, unglamorous work. Haad, the co-founder who owns growth and client solutions, does that diagnosis and will say when an organisation is too small for any of it.

Wobble works with SOS Children's Villages of Sindh, where storytelling and tightly targeted paid social connected the organisation with donors at a competitive cost per conversation, and with LRBT USA, the fundraising arm of a Pakistani eye care institution running twenty hospitals, where a donor acquisition engine has just launched and is honestly described as ramping rather than delivered. The company is based in Karachi, is answerable for what it operates, bills month to month and carries 25 engagements across six countries. Both nonprofit engagements are fundraising rather than case management, and that distance is worth naming before you weigh them.

Common questions

Will this decide who gets help?

No. Eligibility, assessment and safeguarding decisions stay with your committee or caseworkers, and nothing about an individual beneficiary goes out without a person reading it first. What the system does is hold the record, prompt the follow up, flag the missing field and assemble the reporting.

Our funders want reports in their own formats. Can that be automated?

Largely yes, and the saving is real because it is the same numbers retyped into different shapes. The work happens at the other end: the monitoring fields a funder will ask for at the deadline are captured at intake, so the report is assembled rather than reconstructed.

How are restricted funds handled?

As a separate fund from the moment the money arrives rather than at year end. Zakat, sadaqah, designated appeals and grant money each keep their own balance and their own spending record, so a funder asking what happened to a specific gift gets an answer that does not require reconstruction.

Is donor data safe with an outside supplier?

It sits in your accounts under your logins wherever the tools allow, access is by role rather than a shared password, and every access is logged. If the engagement ends you keep the accounts, workflows, documentation and data, and our access is removed.

We are volunteer run with no budget. Is this for us?

Probably not yet, and that is a straight answer rather than a polite one. If the administration fits inside one person's week, the money is better spent on the programme. This pays once recording the work has started to displace doing the work.

Can it write our appeals and impact updates?

It can prepare them, and a person approves them before anything goes out in the charity's name. Where an appeal uses somebody's story, that needs written consent and needs telling with dignity. The drafting saves time. The approval is not optional.