Volunteer Onboarding Checklist for Charities: Build One With AI

Build the checklist in stages: have the AI draft a checklist from your role descriptions, add an AI-written message for each stage (welcome, references, induction, first-month check-in), and give every safeguarding step to your safeguarding lead to approve before use. AI drafts the paperwork and the messages. People decide who is suitable and what is safe.

Blank checklists online tend to list generic items for every charity. What they leave out is the sequence, the wording of each message and who signs off each step, which is where a small team loses volunteers or, worse, skips a safeguarding step. The fix is to treat the checklist as a record with owners and dates, not a poster: each item has an owner, a done-when test and, where it helps, an AI-drafted message.

Follow me on Instagram@sagnikteaches

Start with the role pack, because every other step hangs from it

A role pack is a one or two page document for each type of volunteer role. It answers what the person will do, when, who they report to, what they must complete before a first solo shift, what you expect of them and what they can expect from you. Without it, onboarding becomes a series of favours remembered by the coordinator. With it, the checklist almost writes itself.

Connect on LinkedInSagnik Bhattacharya

A good role pack holds: the purpose of the role; the tasks in plain words; the time commitment and shift pattern; who supervises; the checks and training required before starting (as decided by your safeguarding lead, insurer and funders); the code of conduct; expenses and how to claim; who to ring if you are ill or worried; and the policies to read. Give the AI your notes, not a blank page:

Subscribe on YouTube@sagnikteaches

Draft a one-page role pack for volunteers at a community food bank. Use ONLY these facts. Plain English, friendly, short sections.
Role: Saturday sorter. Tasks: check donated food dates, sort into categories, pack boxes, tidy the hall. Shifts: Saturdays 8:30 to 12:30, at least twice a month. Supervisor: the session lead. Before the first solo shift: induction, safeguarding briefing, signed code of conduct. Checks: [to be decided by the safeguarding lead]. Expenses: travel within the charity's policy, claim within 30 days. Who to call if ill: the session lead's mobile (number on the rota).
Do not add requirements that are not listed. Do not state any legal requirement.

Two corrections are typical (illustrative). The AI adds "must be able to lift 15 kg" because it sounds sensible for a food bank, but you did not decide it, and a lifting requirement can exclude willing people who could do other tasks. And it writes "a background check is required", when the checks line deliberately says the safeguarding lead decides. Every requirement in a role pack should survive the question "is this essential, or just a preference?" If it is essential, it needs a reason you can state.

Once the pack is approved, ask the AI for the sequence: "From this role pack, list the onboarding steps in order. For each step give an owner, a done-when test and whether it needs the safeguarding lead." That list is your first draft of the checklist, and the stages below show what it should contain. For finding people in the first place, ChatGPT for volunteer recruitment posts and rota messages covers the stage before this one.

Stage 1: from enquiry to a confirmed offer

This stage is where enquiries most easily go cold, because they arrive at a busy moment and the reply comes a fortnight later. Set a reply target you can keep (two working days is realistic for a small team), and make every step below visible in your tracker.

A fair process asks every applicant for the same role the same questions. Ask the AI for the question list, then trim it to those that bear on the role: reliability, why they want to help, what they would do if a regular they know was upset, how they handle an instruction they disagree with. Do not ask the AI to score answers or rank applicants. That is a judgement about a person, and the charity is accountable for it.

Reference requests that get answers, and what the AI must not do

A reference request fails when it is vague ("please provide a reference") or when it arrives with no deadline and no way to reply quickly. Make the request short, specific and easy to answer by return email or a five-minute call. The reference request in the messages section further down does that. Four questions are enough: how long and in what capacity the referee has known the person; how reliable they are; how they work with the kind of people the charity supports; and whether there is anything the charity should know that bears on their suitability for this role.

The rule for AI at this step is narrow. It may draft the request, and it may remind you who has not replied. It must not decide whether a reference is good, and it must not be the first reader of the reply. References contain one person's opinion of another, which is sensitive, and a summary that rounds a lukewarm answer up to a warm one can change a decision. In an illustrative case, a referee writes "reliable, generally on time, a bit quiet". A summary that says "excellent reference, highly recommended" has added what was not there. So a person reads the original, and if you want help with a summary, paste only the answers with names removed and ask for exact quotes, not a verdict.

Stage 2: safeguarding steps, drafted by AI and approved by your lead

Every item in this stage is a draft until the safeguarding lead has read it and signed it off. The AI's part is translation: turning your policy into a briefing script, a one-page "if you are worried" card, or a quiz. It does not decide what the thresholds are, who the lead is or how concerns are reported. Paste your own policy into the prompt and tell the AI to use nothing else.

The "if you are worried" card shows why the lead must read every line. In one illustrative draft, the AI told volunteers to "ask the person for more detail about what happened". The charity's policy says the opposite: listen, write down what was said in the person's own words, do not question them, and pass it to the lead the same day. The AI wrote a kind, plausible instruction that contradicted the policy, which is exactly the kind of error a signed approval block exists to catch.

APPROVAL BLOCK (add to every safeguarding document before it is used)
Document: [title]            Version: [1.0]
Drafted with AI assistance from: [name of policy and date]
Reviewed and approved by: [name], Safeguarding Lead
Date approved: [date]        Next review: [date]
Changes made by the lead to the AI draft: [list]

That last line, "changes made by the lead", is useful evidence for trustees that a person read the draft closely. If your charity works with children, the template in a safeguarding policy template for a children's club shows how AI helps with the policy itself, and the same caution applies there.

Stage 3: the induction session, in running order

An induction works best as a single session before the first shift, or as the first hour of it, in an order that puts the most important things first while people are fresh. The table gives a running order for a first shift of about four hours: roughly an hour of induction, then supervised practice. Adjust the timings to your role.

The AI saves most of the time in the preparation. Ask it to produce a one-page "first shift" sheet with the running order and the names to fill in, a briefing note for the buddy ("what to show, what to say, what to watch for") and a short understanding check. Keep the check to five questions, make sure the safeguarding answers come from your policy, and have the lead approve the key. A check that reads "Who is our safeguarding lead and how do you contact them?" and "What do you do if someone tells you something that worries you?" is more use than a long quiz, and it gives you a record that the briefing landed.

Stage 4: the first month and the three check-ins

Volunteers often leave quietly. A short, human check at three points helps you catch problems while they are small. The AI drafts the invitation, and a person has the conversation.

Two habits protect the relationship. First, when a volunteer misses a shift or goes quiet for two weeks, send a kind check-in before you assume they have gone. The AI can draft three versions in different tones for you to choose from. Second, ask what the charity could do better, and act on one thing each quarter, so that feedback visibly changes something.

Messages to copy for each stage

Each of these is a starting point: personalise the first line, and check every bracket before you send. They follow a simple pattern: what happens next, by when, and how to reach a person.

Acknowledgement and welcome

Subject: Thank you for offering to volunteer at [charity name]

Hi [first name],

Thank you for getting in touch. We would love to have you involved. I have attached a short pack that explains the [role name] role: what you would do, when, and what we ask of everyone who volunteers with us.

The next step is a short chat so we can find the right fit. Are you free for 20 minutes on [date] or [date]? If neither works, tell me what does.

Best wishes,
[your name], Volunteer Coordinator

Reference request

Subject: Reference request for [applicant name]

Dear [referee first name],

[Applicant name] has offered to volunteer with [charity name] as a [role name] and gave your name as a referee. A short reply by [date] would help us a great deal.

1. How long have you known [applicant name], and in what capacity?
2. How reliable are they?
3. How do they get on with the people our service supports ([group])?
4. Is there anything we should know that bears on their suitability for this role?

You can reply to this email or ring me on [number]. We will keep your reply securely and share it only with the people who decide on this role.

Thank you,
[your name]

Induction reminder, the day before

Hi [first name], a quick reminder that your induction is tomorrow, [day], at [time], at [address or room]. It lasts about [length]. Please wear comfortable shoes. Ask for [name] at the door. If anything changes, ring [number]. See you then!

After the first shift

Hi [first name], thank you for coming in on [day]. We were glad to have you. How did it feel, and is there anything that was unclear or that you would like more help with? A one-line reply is plenty. Your next shift is [date]. [your name]

Invitation to the four-week review

Hi [first name], you have been with us for about four weeks, which is a good moment to check in. I would like 15 minutes to hear what is going well, what could be better, and whether the role still suits you. Could we speak on [date] or [date]? [your name]

Kind decline, for when a role is not the right fit

Hi [first name], thank you for the time you gave us and for offering to help. After talking it through, we do not think the [role name] role is the right one for you at the moment. That is about the role, not about you. If other roles suit you better, such as [alternative], I would be glad to talk about them. Thank you again. [your name]

Never put the reason for a decline in writing unless the lead has agreed the wording, and never include anything from a reference. If a decision is difficult or sensitive, a phone call from a person is kinder than any message.

Eight new volunteers at a food bank: the time it saves

In the example, with invented figures, a food bank takes on eight new Saturday volunteers in a month. The coordinator, who is also a volunteer, used to spend about two hours twenty minutes on each. With the role pack, the messages above and a tracker, the work looks like this. The figures are estimates, so swap in your own timings.

74 minutes across eight volunteers is 592 minutes, or about ten hours in the month. The chat is untouched on purpose. The bigger gain is consistency: all eight got the same pack, the same safeguarding briefing and the same check-ins, which is what a trustee or an insurer would want to see if they asked. If the charity later adds a Sunday session, the checklist needs a new row for the role rather than a new process.

One checklist, five different roles

The same skeleton bends to fit different roles. Your safeguarding lead decides what each variation needs; the AI's job is to adapt the wording once they have.

• One-day event volunteers. Shorten the process: a brief form, a 15-minute briefing on the day, an emergency contact and a named supervisor. Skipping references may suit a low-risk role if your lead agrees, but the safeguarding briefing and the health and safety basics stay, however short the role.
• Volunteers under 18. Expect extra steps: parental or guardian consent, tighter supervision, limits on hours and tasks, and early involvement of the safeguarding lead before any offer. The rules differ by role and organisation. Ask the AI to draft the consent letter, and have the lead approve it.
• Remote or digital volunteers. The risks shift from contact to data: what they can see, where files live and which accounts they use. Give them accounts on the charity's own tools rather than letting them use personal AI accounts for charity work, as discussed in an AI policy for volunteers using their own accounts.
• Befriending or phone-support volunteers. Add more checking before the first call, supervision after it, rules about contact outside the service, and a way to record calls or visits. The role pack needs a "what to do if you are worried" section written with the lead, and a monthly supervision slot.
• Charity shop volunteers. Add till and cash handling, lifting and stock-room safety, who opens and locks up, and lone-working arrangements. A buddy for the first three shifts replaces most classroom induction, and a short written list of "what to do if" covers the rest.

A tracker the AI can read without seeing names

A shared spreadsheet is enough for most charities. One row per volunteer, with an ID in place of the name for anything you paste into an AI tool. Columns that work: ID, role, stage, date applied, references received (0, 1 or 2), checks outcome (pending or recorded), safeguarding briefing date, induction date, two-week check-in done, four-week review done, and a notes cell for administrative notes only. Keep sensitive detail out of the notes. If you want a rota and sign-up layer on top, a volunteer rota app built with Google AppSheet shows one route.

Then use the AI as a checking clerk, with the rules written down:

Today is [date]. Below is a tracker with volunteer IDs only. Rules: references must be received within 10 days of the request; no volunteer may do a solo shift until the safeguarding briefing date is filled in; the two-week check-in is due 14 days after the first shift; the four-week review is due at 28 days.
List, by ID: (a) overdue items, (b) anyone with a solo shift booked but no briefing recorded, (c) check-ins due this week. Do not guess missing dates. Say "not recorded" instead.

[paste rows]

A response (illustrative) lists V-014 for references outstanding 12 days against a 10-day rule, V-017 for a Saturday shift booked with no briefing recorded, and three check-ins due this week. The second finding is the valuable one: it catches the gap that matters before it turns into a risk. Treat anything the AI flags as a prompt to look at the tracker, and any "all clear" with suspicion until you have glanced at the rows yourself.

Keeping volunteer data in the right place

Most of this checklist needs no personal data in an AI tool at all: the role pack, the messages, the briefing script and the check are drafted from policies and templates. Keep names, references, identity documents, health information and anything about the people you support out of prompts. What donor data must never be pasted into ChatGPT makes the same point for supporters, and the same list applies to volunteers.

When you do need a shared place for templates, a Project (in ChatGPT or Claude) or a Gemini Gem, which is becoming a skill from November 2026, lets every coordinator draft in the same voice. Nonprofit programmes can lower the cost, with eligibility rules you should confirm before applying. As of October 2026, Anthropic's Claude for Nonprofits page lists its Team plan at $8 per user a month with a two-seat minimum, with verification through Goodstack. OpenAI's help centre lists discounted ChatGPT Business Standard seats for eligible nonprofits at $8 a user a month billed annually ($10 monthly; Premium seats are excluded), also verified through Goodstack, and larger organisations can ask its sales team about bigger discounts (its February 2026 update mentions up to 75%). Google Workspace for Nonprofits has been offered at no charge to eligible organisations and includes Gemini features, but confirm both points on Google's page. Whichever you choose, put the rule in writing: AI drafts from templates, and people read anything that concerns a person. A short charity AI policy, such as the one in an AI policy template for charities, gives trustees something to approve.

Decisions that never go to the AI

These stay with named people, whatever the tool can do:

• Whether someone is suitable to volunteer in a particular role.
• What a reference means, and whether it raises a concern.
• Which checks a role requires, and what to do with the outcome.
• Safeguarding thresholds, referrals and the handling of any concern or disclosure.
• Ending a volunteer's placement, and the words used when you do.
• Any complaint about, or from, a volunteer.

Everything else on the checklist is a good candidate for a faster draft. If you keep those six decisions firmly with people, and a record of who made each one, the AI makes onboarding quicker without making it less careful.

Questions charities ask about AI and volunteer onboarding

Which AI tool is safe to use with volunteer information?

Use AI to draft from your own policies and templates, not to process personal data. Keep names, references, identity documents and health details out of prompts where you can. If you must paste personal data, use a business or nonprofit plan whose terms say content is not used for training, and ask your data-protection adviser first.

How many references should we ask for?

That depends on the role, your insurer, your funders and your safeguarding policy. A role with unsupervised contact with children or adults at risk usually needs more checking than a one-off litter-pick. Decide the number for each role type with your safeguarding lead, write it in the role pack and apply it the same way every time.

Can AI write our safeguarding policy?

It can draft wording from your own policy and from the guidance your regulator or funder publishes, but it cannot decide your thresholds, reporting routes or lead roles. Have your safeguarding lead write or approve every rule and sign the final version. Treat the AI as a drafting aid and a plain-English translator.

How long should onboarding take?

It depends on the role: a few days for a one-off event, several weeks where checks and training must be finished before a first solo shift. Set a target for each role, tell applicants the timeline in your first reply, and use the tracker so nobody waits in silence.

Further reads

• Community Centre Volunteer Form: AI Sign-Ups and Rota Reminders — Collect sign-ups and send rota reminders once volunteers are cleared.
• Claude AI for Charities: Five Admin Jobs Set Up With Projects — Keep role packs and message templates in one shared Project.
• AI Tool Discounts for Charities and Non-Profits: What to Claim — Check which AI tool discounts a charity can claim, and the catches.
• AI Governance for Charities: A Trustee Risk Register to Copy — Give trustees a risk register covering AI use across the charity.
• Charity Complaints Policy Template: Draft One With AI and Check It — A drafting prompt, a sceptical second AI pass and a made-up complaint walked through day by day, so a small charity's policy works in practice.
• Charity Shop Pricing: Price Donated Items Faster With AI — Price donated items from two phone photos, a band card and a sold-price check, with a 90-second target per researched item and the brand mistakes AI makes.
• Museum Accreditation Checklist: Policies and Evidence With AI — Copy your scheme's requirements into an evidence tracker, let AI find the policy gaps, and keep trustees as the approvers.
• Sample Church AI Policy: One Page on Staff, Volunteers and Member Data — A one-page sample policy for a church: the admin and pastoral line, colour-coded member data rules, approval and disclosure, with clauses to copy.
• Sports Club Code of Conduct: Player and Parent Versions With AI — Player and parent versions, a five-step sanctions ladder and a safeguarding link, drafted with AI and approved by your committee.
• AI for Thrift Stores: List Donated Items From a Phone Photo — A donation-intake routine for thrift store staff: five photos, a fixed prompt, then brand, condition and price checks before any online listing goes live.
• How to Manage a Community Centre: Where AI Helps and Where It Cannot — Four tests that sort community-centre jobs into AI drafts and person-only decisions, with worked examples for bookings, rotas, newsletters and grants.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Anthropic's Claude for Nonprofits page, OpenAI's help centre article on OpenAI for Nonprofits and its announcement with the February 2026 update, read in October 2026, and Google's nonprofit programme pages for Workspace. Confirm each programme's eligibility rules before applying.