AI automation for non-profit and public sector: the Goede Doelen Loterij benchmark

Non-profit and public sector organisations face a customer service paradox. The people they serve often have the most complex, emotionally sensitive needs. The budgets available to meet those needs are the most constrained. Adding headcount is rarely an option. Not serving people well is also not an option.

Goede Doelen Loterij, the Dutch charity lottery that funds over 80 good causes, resolved this paradox in 2025 by deploying an AI digital employee named Jennifer. The result was the strongest customer satisfaction result in Freeday's entire 2025 deployment cohort. That is worth examining carefully, because it challenges the assumption that automation trades off against care.

Why non-profit customer service is harder than it looks

Goede Doelen Loterij's customer base is not a typical B2C product audience. Many of their participants are older, engaged with the lottery because of genuine connection to the charitable causes it supports, and have high expectations for being treated well when they contact the organisation.

The contact mix reflects this. Participants call about prize notifications, payment queries, subscription changes, and occasionally the emotional dimensions of charitable giving. These are not purely transactional queries. They require a tone that is warm, clear, and respectful.

This is exactly the kind of customer service that AI vendors historically avoided claiming they could handle. The assumption was that automated systems were appropriate for transactional queries and inappropriate for anything requiring empathy or sensitivity. Jennifer's result challenges that assumption directly.

What made Jennifer the top performer in the cohort

The results across the 2025 Freeday cohort varied widely. Understanding what separates the top and bottom performers is more useful than the average.

Three factors separated Jennifer from lower-scoring deployments:

Knowledge base quality. Goede Doelen Loterij invested in getting the knowledge base right before go-live. Jennifer had accurate, complete information about the lottery structure, prize categories, charitable causes, payment processes, and common participant queries. When participants asked questions, they got correct answers. This sounds basic but it is where many deployments fail.

Conversation design. Jennifer's conversational style was designed specifically for the Goede Doelen Loterij audience. Warm, clear, appropriately formal, and never rushed. The tone matched what participants expected from the organisation. This is a design decision, not a default AI behaviour.

Escalation quality. When Jennifer encountered a query that required human involvement, the handover was clean. The participant did not need to repeat their situation. The human agent received the context already prepared. Good escalation design contributes to satisfaction because the overall experience is coherent, not fragmented.

The public sector case for AI automation

Public sector organisations have been slower than commercial enterprises to adopt AI customer service automation, for understandable reasons. Regulatory caution, political sensitivity around automation and employment, and legitimate concerns about digital exclusion have all slowed adoption.

But the pressure to improve citizen services without increasing headcount is intensifying. Dutch municipalities, housing corporations, and public utilities face growing contact volumes with flat or declining service budgets.

The housing corporation sector is a specific example. Woonbron's AP automation deployment (35,000 invoices annually, approximately 80% automation) demonstrates that AI works in the public housing sector. Eigen Haard, another Dutch housing corporation, has also deployed AI. The Freeday governance industry page covers these public sector deployments in more detail.

The evidence from Goede Doelen Loterij is directly relevant for public sector decision-makers who are concerned that their citizens or tenants will respond poorly to automated service. A top result from an audience with high expectations is not a warning sign. It is a validation.

The employment question answered honestly

Every conversation about AI automation in non-profit and public sector organisations eventually arrives at the employment question. Will AI replace the people currently doing this work?

The honest answer from the 2025 cohort data is: it depends on how the organisation chooses to use the freed capacity.

Freeday's 2025 cohort freed 95 FTE equivalents across six deployments. In most cases, that freed capacity was redirected to higher-complexity work rather than used to reduce headcount. For a non-profit customer service team, "higher-complexity work" typically means the emotionally demanding cases that require genuine human empathy and relationship skills, the cases where a human makes the most difference.

An AI that handles routine prize payment queries, subscription updates, and payment plan information frees the human team to spend their time on the participants who are calling because they are struggling financially or because they have a complex situation that needs genuine attention. That is not a worse service. For the participants who need it most, it is a better one.

What the implementation looks like for resource-constrained organisations

Non-profits and public sector organisations often assume that AI deployment requires large IT departments and significant budget. The ATAG 14-day go-live timeline applies here too.

The Freeday deployment model is designed for organisations that do not have dedicated AI engineering teams. The integration work is handled by the deployment team. The knowledge base is built from existing documentation. The organisation's role in the implementation is providing access to their knowledge and their systems, and being available to test and review before go-live.

The ongoing operational requirement is knowledge management. Someone needs to own the knowledge base and keep it current. This is a content management function, not a technical one. Most non-profit and public sector organisations already have people who manage their website and communications. Knowledge base maintenance is a comparable responsibility.

The Freeday contact page is the starting point for organisations assessing whether their contact volume and contact mix justify an AI deployment.

Healthcare as an adjacent sector

The public benefit sector includes healthcare, which has specific characteristics that are relevant to this discussion. Healthcare AI automation is subject to additional regulatory requirements and carries higher consequence risk than lottery or housing queries.

The Freeday healthcare AI agent page covers the specific deployment considerations for healthcare settings, including the human oversight requirements and the contact types that are within scope for automation versus those that require clinical involvement.

FAQ

Can AI customer service automation work for vulnerable customer groups?

Yes, when designed carefully. Goede Doelen Loterij's top cohort result, from an older audience, demonstrates that AI can serve populations that might be considered less comfortable with technology. The key is conversation design that matches the audience's communication style and expectations, not a generic chatbot tone.

Does AI automation reduce costs in non-profit organisations?

Yes, though the primary benefit for many non-profits is service quality improvement and capacity reallocation rather than cost reduction. Freeing human agents from routine queries to focus on complex, emotionally sensitive contacts improves both staff experience and service quality.

How do public sector organisations address concerns about digital exclusion?

Well-designed deployments include easy access to human agents for participants who prefer or need human contact. The AI is not a barrier to human service; it is a faster path to resolution for participants who are comfortable with it. Human agents remain available for those who need them.

What is a realistic implementation budget for a non-profit AI customer service deployment?

Budgets vary significantly by contact volume and scope. The right starting point is a conversation about contact volume and mix. Freeday's SaaS pricing model is accessible to organisations that do not have enterprise IT infrastructure budgets.

How long does it take to go live?

Standard Freeday deployments go live in two to four weeks. The knowledge base preparation period is typically the longest phase for non-profit organisations, as documentation is often spread across multiple systems and formats.