A child sponsorship org considered using AI to write more engaging profile descriptions for sponsors. They rejected at Reject — automation would distance them from the children.
Rejected at the framing stage. Reinvested in field-officer training to write profiles by hand.
Field officers in Honduras and Rwanda are writing 80 child profiles a week for new sponsorship listings. They're tired. The profiles read flat because there's no time to sit with each child. Sponsors complain that profiles all sound the same. The children — most of all — are reduced to bullet points: age, school, favorite color.
Christ would see that the children have stopped being children in this system. They've become marketing assets. He'd see that the answer most leaders are reaching for — make the profiles more 'engaging' — accepts the premise that engagement, not relationship, is what matters.
You named something hard: the profiles are flat because the system asks officers to produce them at a pace that prevents knowing. AI will not fix that. It will hide it. A model that makes profiles 'engaging' from thin data is not narrating a child — it is fabricating one. The sponsor will feel closer to the child. The child will be exactly as far away as before. That gap is the problem.
A generative AI tool that takes basic child data (age, family situation, school year, hobbies) and produces a warm, 200-word profile description for sponsorship listings.
Field officers like Marcos in Tegucigalpa are writing 80 child profiles per week. They're rushed; profiles are formulaic. Marketing team wants higher sponsor conversion. Real people affected: 12,000 sponsored children whose stories are being told secondhand, and ~340 field staff carrying the writing load.
Yes — unambiguously. The right answer is fewer profiles per officer, more time per child, and structural change to our intake process. We've been treating writing as production rather than as bearing witness.
The stated need is 'better-converting profiles.' The real need is 'we have industrialized a process that should be slow.' AI would optimize the wrong thing.
The children themselves. A profile written by a model — however warm — is a lie of intimacy. If sponsors later discover this, trust collapses. If they never discover it, the child is permanently misrepresented. Either outcome harms the most vulnerable person in the transaction.
Reduce the number of new profiles per quarter. Train officers in narrative writing as spiritual practice. Pair each officer with a child for a full afternoon before writing. Accept that some children will wait longer to be matched — and tell sponsors why.
This team named the deepest temptation of AI in mission contexts: efficiency that severs the connection between the helper and the helped. A sponsor reads a profile because they want to know a child. A model that fabricates engagement gives the sponsor an experience and the child nothing. The rejection here is theological, not technical — and that is exactly when Reject is most faithful.
Start from this example as a draft. Every field will be pre-filled — edit freely. Your own context will surface as you go.
These examples are illustrative. Real discernments will be more complex, more painful, and more specific to your context. The library helps you see the shape of the work.
Examples paraphrased from common patterns observed in faith-based development AI projects. Not based on any single real organization.