Someone messages your church asking whether God can forgive something specific. Your pastor preached on exactly that, carefully, four years ago. Nobody can find it, so the person on the other end either retypes a shorter version from memory or says nothing.
That is the most common and least discussed waste in church digital ministry. It is not that the answers do not exist. It is that they exist in video, and video is not searchable, so functionally they do not exist at all.
Last week I said the honest use of AI is where a person had run out of hours. This is the version of that which costs the least and pays back fastest: you already own the material.
A transcript is an index, not content
The mental shift that makes this work is small and it decides everything.
Transcribing your archive does not give you new content. It gives you a searchable index of the content you already paid to produce — the staff time, the study, the preparation, the recording, all of it, already spent. What you are buying with transcription is the ability to find it.
That sounds modest. In practice it changes what a volunteer can do at 11pm on a Tuesday. Instead of a question they are not qualified to answer, they have a search box over twenty years of their own pastor’s teaching, and they can reply with something true that their church already said — in its own voice, with its own convictions, rather than whatever they half-remember.
This is also the honest version of “AI answers questions for your church.” The machine is not the authority. Your teaching is. The machine is the thing that can read ten thousand pages in a second and hand a human the right paragraph.
Do not publish the transcripts
Here is the instinct this triggers, and it is wrong: we now have hundreds of pages of text, let’s put it all on the site.
A few weeks ago I wrote about auditing my own site and finding that 61.9% of my pages read as near-duplicates of something else I had published. Dumping a transcript archive is the fastest way to reproduce that problem from scratch. Preachers repeat themselves across years, deliberately and rightly — the same illustrations, the same phrasing, the same core explanations. As spoken ministry that is faithfulness. As a few hundred web pages it is a pile of pages that all look approximately the same to anything trying to work out which one to show.
There is also the plain reading experience. Spoken language written down is not an article. It is full of asides, restarts and dependencies on a room you were not in. Publishing it does not make your church look thorough. It makes it look like nobody read it before it went up, which is exactly what happened.
So: keep the text where it is useful — searchable, internal, feeding whoever or whatever is answering questions. Publish only the pieces a person genuinely rewrites for readers, and expect that to be a small number.
The three things actually worth doing with it
Answer with it. Whoever handles questions online gets to search the archive first. This is the whole point, and it is the one that changes a person’s evening.
Find the gaps. Once your teaching is text you can ask what is not in it. Most churches discover a topic that comes up constantly in questions and has never once been addressed directly from the front. That is genuinely useful information and there was no way to get it before.
Rewrite the best of it, deliberately. One transcript, properly rewritten into something readable with a real argument, is worth more than the whole archive dumped. The AI guide lists drafting assistance as one of the uses that is clearly working, and this is its best application in a church: the raw material is already yours and already true, so the machine is helping with structure rather than substance.
Start with three months
Not twenty years. Three months, which is about a dozen messages.
Transcribe them. Read through what comes out. Then ask the only question that matters here: what can somebody on our team now do that they could not do last week? If the answer is “nothing, but we have a lot of text,” you have discovered something useful too — that the bottleneck was never the archive and your attention belongs somewhere else in the list.
For most churches the answer is not nothing. It is that the person answering messages stops guessing, and that the empowerment problem gets slightly easier, because a volunteer who can find what the church has already taught needs far less of a pastor’s time to do the work well. That is what raising a ceiling looks like — unglamorous, and it compounds.
Part of the weekly AI series. The full guide is AI for churches.
