Customer description‍

A municipal implementing organization focused on debt counseling and early detection, active in a diverse urban region with a high degree of socio-economic inequality. The team works with housing associations, energy companies and healthcare providers to provide early help to risk groups.

Challenge‍

The organization noticed that traditional signaling mechanisms missed many risk cases, especially among residents with low digital skills or a language delay. Many aid processes only started when problematic debts had already arisen.

Solution‍

A solution was developed that provides a risk model that uses signals from rent arrears, energy consumption, payment behavior and contact history — and explicitly takes language use, reading skills and digital accessibility into account. Based on this, the form and tone of communication were automatically adjusted.

Approach

1. Collection and combination of signaling data
   Integration of data from housing associations, energy suppliers and municipal counters. Includes data about unopened emails, non-callbacks, and language used in previous contacts.
2. Development of risk profiles
   AI model grouped residents based on behavior and signals, and calculated the risk of debt escalation — with attention to language barriers and reading level.
3. Personalizing communication
   The first outreach was tailored based on profile: understandable language, alternative channels (such as WhatsApp or telephony) and adapted conversation.
4. Monitoring effectiveness and further development
   Response to communication, trajectory acceptance and successful referrals were measured and used to adjust the model.

Results

• 26% increase in effective relief interventions within three months
• Significantly higher response among residents with low language skills
• Better connection between care providers and target group
• Lower entry threshold through understandable and accessible communication

Learnings‍

By explicitly including language skills and digital skills in the AI model, a fairer and more effective assistance process was created. The approach shows that technology is not only scalable, but can also be socially inclusive.

‍