Most small charities don't need paid outside help to start using AI. A staff member or trustee with two to three hours a week, a discounted business plan and a one-page policy can run a safe first trial on drafting and admin. Outside help pays off when AI will touch beneficiary data, connect to your donor database or answer the public directly.
What changes the answer is less about the size of the charity than about two questions: how sensitive is the information you hold, and does anyone inside have the time and confidence to test things carefully? A three-person food bank with a tech-minded volunteer may need nobody. A two-person befriending service holding health notes on vulnerable adults probably needs someone to check its set-up, even if all it wants is help writing letters.
What a small charity can safely do on its own
These jobs involve little or no personal data, go wrong in visible ways, and can be tried in an afternoon:
- Drafting newsletters, appeals and social posts from notes you already have, then editing them into your own voice.
- Summarising trustee meeting notes into decisions and actions, with names of beneficiaries removed first.
- Rewriting volunteer role descriptions and policies in plainer language.
- Researching and outlining grant applications, with every factual claim checked against the funder's guidance.
Take the third job first. A volunteer co-ordinator pastes an old role description and asks:
Rewrite this volunteer role description for a charity that
drives older people to hospital appointments. Keep it under
200 words, use plain English, and put the time commitment
and the checks we need (driving licence, insurance, a
background check) near the top. Don't add any duties that
aren't in the original.
[paste original]An illustrative reply starts: "Volunteer driver, two to four hours a week. You'll take older people to and from hospital appointments in your own car, and wait with them if they'd like company. Before you start we'll check your driving licence and insurance and arrange a background check…" What you'd fix: the AI added "mileage paid at the standard rate", which wasn't in the original and may not be your policy. That kind of invented detail is the thing to read for every time. If you want to start with zero spend, how small charities can use AI with almost no budget covers the free and near-free tools.
Meeting notes need a different kind of reading. A trustee secretary pastes rough notes from a board meeting, with beneficiary names already replaced by initials, and asks for "decisions, actions with owners and deadlines, and anything left undecided, as three short lists". An illustrative extract of the reply:
Decisions: The board agreed to open a second weekly session from March. The board approved the new expenses policy.
Actions: Treasurer to apply to the community foundation by 30 January.
Undecided: Whether to buy a second minibus.
Two lines are wrong in ways only someone who was in the room would catch. The notes said the board agreed to explore a second session and cost it, not to open one, and a model tidying notes will often harden "look into" into "agreed". The expenses policy was approved subject to one change, which vanished from the summary. Read every line under "Decisions" against the notes, because that list becomes the minutes, and minutes are what the charity relies on later if a decision is ever questioned.
Cost isn't usually the barrier. Anthropic's nonprofit rate for Claude Team is $8 per user per month for eligible organisations verified through Goodstack, and OpenAI runs a similar programme for ChatGPT Business. Google Workspace for Nonprofits, the no-cost edition for verified charities, now includes the Gemini app and Gemini Notebook. AI tool discounts for charities lists what to claim and how.
Five signs you've reached the edge of do-it-yourself
- AI would handle beneficiary information. Health details, safeguarding notes, immigration or benefits situations, anything about children. Mistakes here harm real people and can breach data-protection law such as the GDPR. The set-up needs checking by someone who has done it before, even if the task itself is simple.
- You want systems to talk to each other. Web-form enquiries flowing into your donor database, thank-you emails triggered by donations, volunteer rotas updated automatically. Connections are where quiet failures happen: an automation that stops running and nobody notices for a month.
- AI would speak to the public for you. A chatbot answering supporter questions or a helpline triage tool. What it says becomes the charity's word, and it needs testing against the questions people really ask, including distressed ones. How charities use AI chatbots to answer supporter questions shows the testing involved.
- The board needs a decision it can defend. Trustees want a policy, a risk view and a clear line on what data may go where. Writing the policy yourself is realistic (see how to write an AI policy for a small charity); judging the risks of a specific set-up is harder.
- Nobody has two hours a week. If every staff member and volunteer is already stretched, a trial never gets past the first week. A few hours of outside time to set things up can be the difference between adoption and a forgotten login.
Score your charity in ten minutes
Give your charity 0, 1 or 2 points on each row. Be honest; the point is to find where help would be spent well.
| Factor | 0 points | 1 point | 2 points |
|---|---|---|---|
| Sensitivity of data AI would touch | None or public information only | Supporter contact details | Beneficiary health, safeguarding or financial details |
| Systems to connect | None; copy and paste is fine | One connection (e.g. form to spreadsheet) | Several, or the donor database |
| Public-facing AI | No | AI drafts replies a person sends | AI replies directly to the public |
| Internal time available | Someone has 2+ hours a week | An hour a week, irregularly | Nobody realistically |
| Internal confidence | Someone already uses AI tools well | Some curiosity, no experience | Nobody has tried |
A total of 0 to 3: go it alone and revisit in six months. 4 to 6: go it alone for drafting jobs, and get a few hours of outside review before anything touching the higher-scoring rows. 7 to 10: outside help will probably save money and risk, provided it's scoped tightly.
Four charities, four different answers
A volunteer-run food bank: score 3
It wants help writing rota emails, funder updates and a newsletter. Referral forms mention household circumstances, but AI won't touch those. One volunteer uses ChatGPT at work. Answer: no outside help. The volunteer sets up a nonprofit business plan, writes a half-page rule ("never paste referral forms"), and runs a four-week trial on the newsletter and funder updates.
At the end of the trial, the volunteer compares a few plain numbers rather than a general impression. In this illustration, the monthly newsletter took about three hours before and an hour and a quarter with AI drafting from the co-ordinator's notes; the funder update fell from two hours to just under one. Two invented details were caught in four weeks: a volunteer count that was out of date, and a quote attributed to "a regular visitor" that nobody had said. That's a trial worth continuing, with one rule added: every number and every quote in a draft is checked against the notes before it goes out.
Six months later the committee asks whether AI could sort incoming referral forms by urgency. The score changes: sensitivity moves from 0 to 2 because referral forms describe household circumstances, and connecting the form to a spreadsheet adds a point. The total goes from 3 to 6, which is the signal to get a short outside review before building it, even though the newsletter work stays in-house.
A youth football club: score 5
Drafting match reports and fixture emails is easy. But the committee also wants AI to sort registration forms, which include children's medical notes and emergency contacts, into team lists. Answer: do the drafting alone now; get a short outside review of the registration idea before building it, because children's data and safeguarding raise the stakes even when the task sounds like admin.
A befriending service for older people: score 8
Two paid staff, forty volunteers, case notes that include health and family details, and a wish to have AI summarise volunteer visit reports for the co-ordinator. Nobody has used AI beyond curiosity. Answer: outside help is worth it here, scoped to three things: which tool and settings are safe for case notes, a tested summary workflow with names stripped out, and a handover so the co-ordinator can run it alone.
An animal rescue: score 4
It wants a chatbot on its website for adoption questions ("Do you rehome cats to flats?", "What's the adoption fee?") and help drafting adopter home-check reports. Answer: the chatbot is the part to get help with or to hold off on, because it answers the public; home-check drafting can start in-house with addresses removed from what goes into the AI.
Which kind of outside help fits which problem
| Type of help | Good for | Watch out for |
|---|---|---|
| Skilled volunteer | Drafting workflows, simple automations, enthusiasm | Builds under personal accounts; leaves with the knowledge |
| Corporate volunteering or pro bono team | A defined project over a few weeks | Their priorities and timelines, not yours; hand-off at the end is often thin |
| Your software supplier's onboarding | Turning on AI features in a donor database or CRM you already use | Advice limited to their own product |
| Peer charity that's done it | Honest lessons, templates, policies to adapt | Their data and risks may differ from yours |
| Freelancer | Building a specific automation to a written spec | Scope creep; ask for a fixed price and documentation |
| AI implementation consultant | Deciding what's worth doing, checking data risk, setting up and handing over | Paying for general advice instead of a defined result |
Whichever you choose, agree a fixed scope and a fixed fee or number of hours before work starts. Open-ended "help us with AI" arrangements are how small charities spend more than they meant to and end up with a slide deck instead of a working process. For a broader view of when paid help is warranted, eight signs it's time to get help applies to charities too.
Questions to put to anyone offering to help
Charities attract offers of help, some excellent and some that mostly want a case study. Five questions separate them quickly:
- "What will we be able to do without you at the end?" A good answer names a working process and who runs it. A weak answer is "you'll understand AI better".
- "Which of our data will you need to see, and in which tools?" Anyone who asks for a full export of your donor database before they understand the problem is starting in the wrong place.
- "Have you set this up for an organisation holding data like ours?" Experience with sales teams doesn't transfer neatly to safeguarding notes or health information.
- "Whose accounts will it run on?" The only acceptable answer is the charity's.
- "What will it cost us each month after you've gone?" Automation and AI subscriptions keep billing long after the project ends, and a clever build on three paid tools can cost more than the time it saves.
Question 5 is worth doing as a sum, because the answer is rarely zero. Suppose the befriending service receives a proposal to automate its visit summaries: each emailed report triggers a Zapier workflow with an AI step, and the summary is written into a shared document. Forty volunteers each filing a weekly report makes about 160 reports a month. Zapier's AI step uses 1, 3 or 5 tasks a run depending on the model tier chosen, plus one task to write the result, so the workflow uses between 320 and 960 tasks a month. The Professional plan (listed from $29.99 a month on monthly billing) includes 750 tasks, so on the top tier the charity would be pushed onto a bigger plan. The proposal also puts the two paid staff on Claude Team at the nonprofit rate of $8 a seat for their other drafting, so the running cost is somewhere around $46 a month, or $550 a year, before anyone has checked a summary.
The alternative proposal is plainer: the co-ordinator pastes the week's reports, names removed, into Gemini, which comes with the charity's Workspace for Nonprofits at no extra cost, using a saved prompt, and spends 45 minutes a week reading and correcting the result. Neither answer is wrong. But a charity that asks the question sees the choice between $550 a year and three hours of staff time a month, rather than finding the subscription on a bank statement after the helper has gone.
If the answers are vague, keep the work in-house for now and come back to the offer once you know what you want from it.
A one-page brief for whoever helps you
Write this before the first conversation. Here it is filled in for the befriending service above:
WHAT WE WANT: Volunteer visit reports summarised weekly for
the co-ordinator, highlighting anyone who seems more isolated
or unwell than last time.
WHAT WE HOLD: Case notes with health conditions, family
contacts and home addresses. About 120 people supported.
CURRENT TOOLS: Google Workspace for Nonprofits, a donor
database, visit reports sent by email.
WHO WILL RUN IT AFTERWARDS: The volunteer co-ordinator,
about 3 hours a week of admin time.
LIMITS: No names or addresses in any AI tool unless it's
covered by our business agreement. Nothing sent to the
people we support without a person reading it.
DONE MEANS: A written weekly routine, tested on 4 weeks of
real reports, that the co-ordinator can run alone, plus a
one-page note the trustees can approve.
BUDGET: Fixed fee, agreed up front.Keeping the work yours when the helper leaves
An illustrative failure: a skilled volunteer sets up an automation that copies donation notifications into a spreadsheet and sends thank-you emails. It runs on their personal automation account and their personal email. Eight months later they move away, the account lapses, and the charity discovers three months of unthanked donors. Nobody else knew the automation existed.
Before any helper, paid or unpaid, finishes, check off:
- Every account is owned by a charity email address, with at least two people holding the login.
- There's a one-page description of what each workflow does, what triggers it and how to tell if it has stopped.
- Prompts and templates are saved in a shared folder, not only in someone's chat history.
- Someone inside has run the workflow alone, start to finish, at least twice.
- The trustees know which AI tools are in use and what data goes into them. Charity AI risks around data and donor trust covers what they should be asking.
If those five things are true, outside help has done its job, and the charity can carry on without it.
Other questions trustees ask about AI help
Are trustees responsible if an AI tool causes a data breach?
Trustees are generally responsible for overseeing how the charity handles personal data and risk, whichever tool or person caused the problem. That's why the board should approve which AI tools are allowed and what data may go into them. If you're unsure where your duties start and stop, check your charity regulator's guidance or ask a solicitor who advises charities.
Can we use free ChatGPT or Claude for donor emails?
For drafting a generic appeal with no names in it, a free plan with the model-training switch turned off in privacy settings is workable. For anything containing donor or beneficiary details, use a business plan that doesn't train on your content by default. Nonprofit discounts make those plans cheap, so the free tier rarely makes sense for real supporter data.
How long before a small charity sees any benefit?
Drafting jobs such as newsletters, role descriptions and meeting summaries usually save time within the first two weeks. Anything that connects systems, such as moving web-form enquiries into a donor database, takes longer: plan on four to eight weeks including testing, whether you do it yourself or bring someone in.
Further reads
- Donor Thank-You Letters With AI That Still Feel Personal — A good first do-it-yourself job with a clear before and after.
- How to Use AI to Write Grant Applications, Step by Step — Step-by-step grant drafting, one of the biggest time savers for small teams.
- Turning Charity Data Into Impact Reports With AI — Turning service data into impact reports without exposing beneficiaries.
- What an AI Consultant Can't Do for You, and What You Must Own — What stays your job even when you do bring in help.
- AI Consultant vs Agency vs Freelancer: Which Should You Hire? — How the different kinds of paid help compare on cost and fit.
- How to Anonymise Client Data Before You Paste It Into AI — The practical method for stripping names before text goes into AI.
- AI Grant Writing Tools Compared for Small Non-Profits — Specialist grant tools against general assistants on non-profit pricing: what each costs, what it does well, and which suits 4, 12 or 30 bids a year.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Claude for Nonprofits page (claude.com); Google for Nonprofits help pages on AI features in Workspace for Nonprofits; published summaries of OpenAI for Nonprofits (checked September 2026).