01Only your real data
The AI drafts only from records in your organisation's workspace: registers, field submissions, grant data and consented field notes. It does not invent numbers, people, quotes or stories, and it does not bring in data from other organisations.
02Every number has a source
Each figure in a draft links to the dataset, round and filter it came from. Reviewers can check any claim before approving.
03Limits are stated
- Who collected the data, and whether it was independently verified.
- How many people were missed or not reassessed.
- Where results compare different groups instead of the same people.
- That changes observed are not, on their own, proof of what caused them.
04Consent and dignity
Stories and photos are used only where consent is recorded. Names are changed by default and identifying details are removed. Children's stories require a parent's or guardian's consent.
05People approve every report
No report is shared until a named person in your organisation approves it. The approver and date are recorded with the report's version history. Overstated words are flagged for that person's attention.
06No training on your data
Your records are used to draft your organisation's reports and for nothing else. They are not used to train AI models.
07Hosting
Program data is encrypted and hosted in India by default.