Social sponsoring, the other way to do marketing
A campaign buys attention for a few weeks. Social sponsoring buys a lasting association between your brand and a concrete result. What it gives, what it costs, who it suits.
AI for associations: where to start
An association does not need an AI strategy. It needs to get time back on the tasks that do not deserve it: writing, repetitive answers, sorting. Generative AI can do those three things today, provided it is used with a few rules. Here is where to start, without budget, without risk for your data, and without turning your team into an IT department.
In short
- Start with a tedious task with no judgement at stake, not with a project
- Three uses work from the first week: writing from your notes, answering repetitive questions, sorting and summarising
- No personal data of your beneficiaries in a free consumer tool; proofread every figure, name and date
- An AI policy fits on one page: allowed tools, forbidden data, mandatory proofreading, disclosure when useful
- Free tools are enough to start; an assistant linked to your website is quoted at cost in Kern4Good
Where associations stand
Sector surveys all say the same thing: the will is there, the framework is missing. In TechSoup's AI Benchmark Report 2025, run mostly among North American non-profits, 70% of respondents think AI can reduce their workload, 60% say they lack in-house expertise, and 86% have no AI usage policy. Belgian figures are still missing, but the situation of small structures is the same: a few people already use a consumer tool on their phone, without a common rule.
That is the point that matters: AI is already in your association, in a diffuse way. The question is not to introduce it, it is to frame it and make it useful.
Rule number one: a task, not a project
Choose a task the whole team finds tedious and that requires no judgement: writing the monthly call for volunteers, summarising the meeting minutes, preparing the answers to the questions you are asked ten times. Equip it, measure the time saved, then move on to the next one. An "AI project" with a committee and a specification is the best way never to start.
Three uses that work from the first week
Writing from your notes
Give the tool your raw notes and your old texts, ask for a first draft in your tone. Proofread, cut, publish. The tool does not replace the person who writes, it spares them the blank page. It also helps to adapt: the same message in a short version for social networks, a long one for the newsletter, in Dutch or in English to be proofread.
Answering repetitive questions
Gather your reference documents (registration conditions, opening hours, procedures, fees) and ask the tool to prepare template answers you keep at hand. To go further, an assistant linked to your website answers visitors with those same documents, at any hour, and says when it does not know. It is the use that gives the most time back to associations whose mailbox overflows.
Sorting and summarising
A thirty-page report to read before tomorrow, a mailbox to sort by urgency, minutes to produce from a meeting recording, a twenty-page call for projects whose criteria must be extracted: AI does the first pass, you do the second.
The rules on data
Generative AI does not change the law: as soon as personal data (a name, an address, a situation) goes into a tool, it is processing within the meaning of the GDPR, and the association remains responsible for it. The CNIL, the French authority, wrote it in black and white in its recommendations, and points out that the main risk is the individual, diffuse use of consumer tools, not large projects. The European AI regulation adds two obligations that also concern small structures: making sure the people who use AI are trained in what it does and does not do (article 4, applicable since February 2025), and saying when a content or an assistant is generated by an AI (article 50, applicable since 2 August 2026).
- No personal data of your beneficiaries or members in a free consumer tool: some use your conversations for training
- Read the "do not use my data for training" setting when it exists, and turn it on
- For regular use with internal documents, prefer a professional tool with a contractual framework and servers in Europe
- Proofread everything that contains a figure, a name or a date: AI can invent, confidently
- Say so when an assistant answers instead of a person: it is an obligation, and it is also what your publics expect
A one-page AI policy
The word is scary, the document is simple. One page, validated by the board, that says:
- The allowed tools, and which ones are forbidden for the association's data
- The data that never goes into a tool: beneficiaries, members, case files, passwords
- Proofreading: who proofreads what before publication, and what is never published without proofreading (figures, commitments, answers to a person in difficulty)
- Disclosure: when and how you say a text or an answer comes from an AI
- The referent: the person to ask when in doubt
This page protects your volunteers as much as your beneficiaries: it says what is allowed, and nobody has to guess any more.
What it costs
Consumer tools have a free version that is enough to start, and subscriptions at a few dozen euros a month for regular use. An assistant linked to your website is quoted. In the Kern4Good programme, the half-day discovery session for the team and the assistant are offered at cost: the AI for non-profits page presents what the programme does today, and the glossary defines generative AI.
The assistant on your website: what to demand
An assistant on an association's website answers visitors' questions from your documents. Before installing one, demand four things: that it only answers from your texts and says "I do not know" otherwise, that it always gives a path to a person (phone, email, drop-in hours), that it records no personal data without saying so, and that it announces it is an automated assistant. An assistant that invents an opening hour or an eligibility condition does more damage than an FAQ page.
And the digital divide?
An association serving publics far from digital tools must keep a human path. In Belgium, according to Statbel, 61% of people aged 16 to 74 have at least basic digital skills in 2025; the others will not write to an assistant. AI then serves the team, to answer faster, not the public, to replace the front desk. The Digital divide page details what the programme offers on this point.
Where to start, in practice
- Choose the task and the person who does it
- Gather the association's reference texts in a shared folder
- Test the tool on three real cases, compare with what you would have written
- Note the time saved and the errors seen
- Write the rules page, have it validated, display it
- Decide: keep, adjust, stop; then move on to the next task
The Digital training page describes the half-day discovery session for teams; the guide Redoing your association's website explains where an assistant fits in a website.
Sources
Pages consulted on 19 September 2026. Rates, ceilings and percentages change: check the update date of each page before building a budget on it.
Do we need to train the whole team?
No. Train the person who does the chosen task, let them show the result to the others. The will to get started follows. The half-day discovery session sets the common framework, it does not make everyone a specialist.
Can AI write our grant applications?
It can prepare the structure and a first draft from your reports, and extract the criteria from a call for projects. The substance, the figures and the commitment remain yours, and that is what the reader of the file expects.
Will our volunteers feel replaced?
Tell them what AI does (the first draft, the sorting) and what it does not do (the relationship, the decision, the front desk). The time saved goes back to the field, that is the argument that reassures, and the rules page makes it tangible.
Which tool should we choose?
The one your referent finds simple, whose data setting is clear, and whose paid version costs no more than a few dozen euros a month. We impose none; the scoping compares two or three options on your real cases.
Do we have to say we use AI?
For an assistant answering visitors, yes: the European regulation requires it since August 2026, and your publics prefer it. For a text proofread and published under your name, disclosure is not mandatory; your rules page decides.
