Skip to content

Sermon Prep

The Last Read-Through Is Still Yours

“What if I just read the AI’s output straight off the page?” It’s a fair worry — but it assumes a use case the report was never built for. A Didymus Lab report isn’t meant to be read cover to cover and preached verbatim. It’s a reference structure: something you open and move through selectively, pulling the piece you need for this week’s message, not a manuscript you deliver as-is.

Why AI draft + human review actually works better

This isn’t a vague safety promise — it’s a studied pattern. A 2024 paper in the Journal of the Association for Information Systems (Drori & Te’eni, Human-in-the-Loop AI Reviewing: Feasibility, Opportunities, and Risks) empirically examined workflows where AI produces a draft and a human reviews and revises it on top. The core finding is straightforward: neither an AI-only output nor a human starting from a blank page performs as well as a structure where AI does the first pass and a human makes the final call. That combination captures both accuracy and speed in a way either extreme misses on its own.

Didymus Lab’s report pipeline is built on exactly that structure. A draft passes through an automated check that catches citations to Bible verses that don’t exist and journal DOIs that were never issued, then through a separate adversarial audit — a second AI process built to critically re-read the finished draft from scratch. That’s as far as the AI side goes.

Passing the audit doesn’t mean a report ships straight to your inbox, either. Every draft is first posted to an internal review channel, where someone on the Didymus Lab team who is pursuing a Ph.D. in biblical studies opens it and signs off before it goes out. If the audit flagged something, that flag travels along with the draft, so a person is looking at exactly what tripped the gate rather than starting from a blank read. That’s where our side of the process ends. What comes next is yours: which background note actually earns a place in this week’s sermon, which commentator’s reading you follow, which sentence gets rewritten in your own voice — all of that gets decided in the process of you reading and choosing.

What to actually do with a report

A few habits make this concrete:

  1. Don’t read start to finish. Open straight to whatever section this week’s sermon needs — background (§2), original-language notes (§3), whatever’s relevant.
  2. Check the footnotes. Confirming that a cited source actually exists and is quoted in context is worth doing yourself. We filter hard for this, but the pastor standing in the pulpit carries the final responsibility, and that doesn’t change no matter how many automated checks run first.
  3. Discard readings you don’t hold. Where a report includes a commentator’s interpretation that doesn’t match your theology or your congregation’s context, treat it as one voice among several — it’s there because multiple perspectives were included on purpose, for you to choose from, not because any one of them is the verdict.

AI gets you a fast, wide-ranging first pass. Deciding “is this actually right for Sunday” stays — now and going forward — entirely in your hands.

Next: where that saved time actually goes — the real shift from hours in commentaries to hours in prayer and visitation.

Comments

Loading comments…

Leave a comment

Comments are reviewed before being published. Only your name and comment are stored — no personal data collected.

You might also like

Sermon Prep

The Hours You Used to Spend in Commentaries — Where Do They Go Now?

AI's real value in sermon prep isn't finding more sources — it's the harder job of narrowing them down. Here's why more material usually means more prep time, not less, and why selection, not search, is the thing actually worth automating.

Sermon Prep

If AI Writes the Sermon, Is It Still Mine?

The deepest objection to AI sermon prep isn't laziness — it's identity. A prominent evangelical voice made that case publicly. We take the critique seriously, then draw the line that actually matters: between gathering material and deciding what to preach.

Sermon Prep

Afraid an AI Will Make Something Up and You'll Preach It?

The biggest fear around generative AI isn't laziness — it's a confident lie. Here's why AI 'hallucination' happens, why it's a reasonable thing to worry about, and how verified sources plus automated checking gates are built to catch it before it reaches a pulpit.