Keeping church member names private when you use AI
AI tools can help a busy church office reword a reminder or summarise a long email thread. But a church holds some of the most personal information people ever share: who is in the hospital, whose marriage is struggling, who gave what. Here are habits that let you get the help without giving that away.
Know what happens to what you type
When you paste text into an AI tool, it is sent to the company that runs the model. Depending on the product and your settings, it may be stored, reviewed by people, or used to improve future models. Consumer chat apps and business or API services often have different rules. Before using a tool for church work, read its data and retention terms, and prefer a service that offers zero data retention, meaning your request isn't kept after the answer comes back.
Replace names with placeholders
The simplest protection is never to send the name at all. Swap each person for a label before you paste, and swap them back yourself afterwards:
Before: Hi Maria, could you cover nursery on the 22nd?
Sent: Hi [Volunteer A], could you cover [Role] on [Date]?
After: (you put "Maria", "nursery" and "the 22nd" back yourself)Keep the list of who is [Volunteer A] on your own computer, not in the chat. Remove email addresses, phone numbers and street addresses entirely; the AI doesn't need them to help with wording.
Send only the context the task needs
Placeholders don't help if the rest of the text gives the person away. “[Person 1], our worship pastor's wife, who just had surgery” identifies someone in any small church. Ask whether the AI really needs each detail. For rewording a scheduling note, it almost never needs health, family or giving information.
Things to never paste
- Prayer requests, pastoral care notes or counselling notes.
- Giving records.
- Anything about children, beyond what a scheduling note needs.
- Membership lists, rosters or exports from your church database.
- Anything someone told a leader in confidence.
Let the AI help with words, and keep decisions with people
AI is useful for wording. It is a poor judge of people. Don't ask it who is “reliable”, who seems to be pulling away, or who should be asked first. Those are pastoral judgements, and they belong to people who know the person.
Check the output before it goes anywhere
Read every AI draft before you send it, as if you wrote it. Make sure it says only what you meant, and that nothing was added that you didn't provide.
Where Ozzie fits
Ozzie builds these habits in. When it uses AI, the AI sees placeholders instead of the names Ozzie read from Planning Center, and names are put back on your Mac. A second check refuses any request that still holds a name Ozzie has shown you, an email address or a phone number. On the managed route every AI request requires zero data retention, and you can use a model on your own Mac instead. Ozzie can only recognise names it has read, so the same advice applies: don't type other people's details into a question. Read more about privacy and safety.