AI volunteer enquiry handling.

Why volunteer enquiries need a workflow

Volunteer enquiries are time-sensitive and repetitive, which suits AI drafting with clear limits. Most ask the same things: which roles exist, when shifts run, whether there is a minimum age, where to park and how to start. A part-time coordinator answering between other jobs often takes days, and keen people move on.

The workflow sorts each enquiry, answers routine questions from approved information, matches the person to open roles, and books an induction or hands over to a coordinator with a summary. It does not decide who is suitable; AI for nonprofits covers the wider sector.

The workflow in detail

Inputs

Web form submissions, the volunteer inbox, social messages where the platform allows access, the current role list with requirements, the induction calendar, and a coordinator-approved FAQ.

Systems

Microsoft 365 or Google Workspace mail, a volunteer database (Airtable, a spreadsheet or volunteer software), a booking calendar, and a model limited to the approved FAQ and role list.

Steps

Classify the enquiry into one of the five types in the example below. Extract contact details, availability and interests. Check role requirements, draft a reply, offer induction times and update the volunteer record.

Human review

For the first month a coordinator approves every reply. After that only low-risk categories send automatically; groups, roles with children or vulnerable adults, disclosures, complaints and anything unclear go to a person.

Outputs

A sent or queued reply, an updated volunteer record, an induction booking, and a daily coordinator summary.

Measures

Median time to first reply, enquiry-to-induction conversion, share of replies sent without edits, and errors found in a weekly sample.

Risks

Promising a placement that does not exist, quoting an out-of-date age rule, collecting sensitive information in chat, and answering a safeguarding disclosure with a template.

Illustrative example: a food rescue charity

Illustrative example: a food rescue charity in Tāmaki Makaurau receives about 140 volunteer enquiries a month, more after media coverage. Assumptions: a coordinator on 0.6 FTE spends about 9 minutes per enquiry (roughly 21 hours a month) and the median first reply takes three working days. The split below is synthetic.

If the 101 automatic cases take a minute each for spot checks and the 39 others still take 9 minutes, the month needs about 7.5 hours instead of 21, and most people hear back within minutes.

New Zealand rules the replies must respect

• What you tell people at sign-up. Information privacy principle 3 requires reasonable steps to tell people why their information is collected, who will receive it, whether giving it is optional and how to access or correct it, ideally before collection. A chat assistant that gathers details must show or link that notice first.
• Volunteers as workers. Under section 19 of the Health and Safety at Work Act 2015, a volunteer who works regularly, on an ongoing basis and as an integral part of the organisation is a “volunteer worker”, though fundraising activity is excluded. The section 36 primary duty of care, including necessary training and instruction, then applies, so regular volunteers go through induction rather than just turning up. An all-volunteer group that employs nobody is a “volunteer association” and not a PCBU (section 17).
• Safety checks for roles with children. The Children’s Act 2014 safety-checking rules define work as paid work, or unpaid work done as part of an educational or vocational course, so most unpaid volunteers fall outside them. Many organisations vet child-facing volunteers anyway. The NZ Police Vetting Service is open only to authorised agencies and, as listed on 19 October 2026, was processing most requests within 31 working days, above its 20-working-day target. Replies should give that honest timeframe and never suggest a child-facing role can start before vetting clears.

What the coordinator decides

Coordinators decide who is placed where, especially in roles with children, older people or clients in crisis. They handle disclosures, whether someone mentions their own hardship, a health condition or a concern about another volunteer, because a templated reply to a disclosure does harm. They also decline people, chase references and manage group capacity.

Failure modes and controls

• Phantom roles. The model offers a Saturday shift that filled last week. Control: roles and shifts come from the live database at reply time.
• Over-collection. A friendly assistant asks about health or criminal history. Control: a fixed field list; anything beyond it is declined and routed.
• Missed escalation. A message mixing an offer to help with a personal crisis gets a cheerful template. Control: a separate check for distress and safeguarding language, tuned to over-escalate, plus weekly sample review. See managing AI mistakes and escalation.
• Reply loops. Out-of-office messages trigger replies. Control: header checks and one automatic reply per thread per day.

Cost drivers and a sensible start

Costs follow channels and data, not model fees. Email and web forms are simple; social messaging adds access and review work. A clean role list matters more than prompt wording, and an Airtable base or spreadsheet is enough for most organisations. Build the FAQ from six months of real enquiries using our support-ticket FAQ method.

Start with email and the website form only, with every reply approved for a month. Our AI customer support work covers routing and review design. Logged volunteer hours can feed grant evidence assembly as in-kind support. Volunteers who also give belong under the consent rules in donor communication drafting, not on appeal lists by default.

Written by the Levy Street team in Wellington. Last updated 19 October 2026.