In my day job, I lead engineering teams that build software. We use AI constantly. And engineers tend to talk about AI very differently from how it’s marketed.
Here are a few things my engineering colleagues take for granted that every pastor should know.
1. It predicts. It doesn’t know.
A language model is trained to predict what text should come next, based on patterns in a huge amount of writing. That’s a simplification, but it’s the right mental model. When it answers your question about Romans 9, it isn’t looking up the truth. It’s producing a likely-sounding response.
Often, likely-sounding and true are the same thing. Sometimes they aren’t.
2. It makes things up, confidently.
Engineers call it hallucination. The model produces a quote, a citation, a statistic, or a historical event that doesn’t exist, and states it with total confidence. It isn’t lying. It has no idea it’s wrong.
For pastors, this shows up as fake Spurgeon quotes, nonexistent journal articles, invented Greek word meanings, and Bible verses that are slightly off. Always check quotes, citations, and Scripture against a real source.
3. Grounding helps.
Tools that answer from a specific set of documents, like Logos’s Study Assistant answering from your library, or a workspace AI answering from your team’s notes, are much more reliable than a model answering from memory. They show their sources, which means you can check them. (I wrote about using the AI in Logos.)
When it matters, use grounded tools, and click the sources.
4. Garbage in, garbage out.
Vague questions get vague answers. Give context: who the audience is, what you’re trying to do, what you already believe, what format you want. Engineers spend real time on how they ask. Pastors should too.
5. Review is not optional.
On a good software team, no code ships without someone reviewing it, including code AI wrote. That’s not because AI code is bad. It’s because responsibility can’t be delegated to a tool. A human puts their name on it.
Ministry should work the same way. Every sermon, letter, and post that goes out has a human who owns it. If something is wrong, “the AI said so” isn’t an answer. That’s why I say on this site that I’m the final editor of everything under my name.
6. Data is the real risk.
When engineers worry about AI, they don’t mostly worry about robots. They worry about data: where it goes, who sees it, how long it’s kept, and whether it trains the next model. In a regulated industry like healthcare, that’s a legal matter. In a church, it’s a matter of trust. Treat a prayer list like a medical record.
7. It changes monthly.
What’s true about these tools today will be partly outdated in six months. Build habits and principles, not dependence on one product’s features. Review your policy every year.
“Test everything; hold fast what is good” (1 Thessalonians 5:21). That verse would make a fine sign above any engineering team’s door. It belongs above the study door too.
Translating between technology and ministry is what I help churches do. I’m a pastor with more than twenty years in business and technology. If that would help, here’s how I work.
Take this with you
Enjoyed this article?