AI for charities and non-profits

AI for charities means building analyses and models using data you already have: donations, direct debit mandates, responses to mailings and campaigns, and posts about your campaigns. This helps you see who may stop donating, forecast income or measure what a campaign achieved. These are among the AI applications we build.

Many of the same questions arise for cultural organisations, which also want to show the social value of their activities.

Donor analysis groups your donors by their giving behaviour. A common starting point is RFM: how recently someone donated, how often and how much. You can add factors such as the acquisition channel, the length of a direct debit mandate and responses to previous campaigns.

The most common question is who is likely to stop donating. That works much like subscription churn; read more about customer analysis and reducing churn. The pitfall: a model that selects only generous donors overlooks loyal donors who give smaller amounts.

By source, because each behaves differently. Regular direct debits are relatively easy to track, one-off donations depend on campaigns and the news, and legacies arrive irregularly. A forecast for each source, with a range, tells you more than a single total.

Watch out for years when a major campaign or emergency appeal skewed the figures: mark those years, or the model will learn the wrong pattern. Read more about forecasting at forecasting.

Only with a legal basis under Article 6 of the GDPR, such as consent or legitimate interest. Answers to parliamentary questions from August 2026 state that the organisation must assess this itself first, with oversight from its data protection officer (DPO) and the Dutch Data Protection Authority. Your privacy notice should explain which data you use and why.

Systematic, large-scale profiling often requires a DPIA. The CBF may ask you to substantiate a claim of legitimate interest. Your lawyer or DPO makes the decision; we help you take stock of what is involved. See Responsible AI.

According to the ACM, charities may still call current and former donors to make a request, but anyone who says they do not want to be called again must be removed from the list. Since 1 January 2025, recognised charities have been subject to a standard requiring cancellation to be easy, overseen by the CBF.

For a model, this means objections, requests not to be contacted and cancellations are strict filters applied before any selection. That information must be kept up to date from the donor system.

By collecting posts and responses and determining whether the tone of each is positive or negative. For the Respect Foundation, we built a social media monitor and an impact analysis for the Week van Respect; our LLM analysed the sentiment.

This lets you see not only a campaign’s reach but also how people talk about it. Read how these techniques work at NLP and text analysis.

Across multiple systems. The donor system or CRM holds direct debit mandates, the payment provider holds collection records, web forms capture online donations, and agencies recruiting donors door to door and on the street supply their own files. The same donor often appears more than once.

Linking these sources to a single donor identifier and removing duplicates provides the foundation for any analysis. Read how this comes together at data and AI platform and data quality.

If your donor list is small enough for you to know donors personally, donations cannot be linked to individual donors, or a breakdown of income by source is enough. In those cases, a Power BI dashboard is more useful than a model.

If your data is scattered and full of duplicates, start with the data foundation. For smaller organisations, there is data and AI for SMEs. If a simpler solution is a better fit, we’ll say so.