When the Melody Drifts

AI, Sermons, Governance and Christian responsibility

This is the second part of a two-part reflection on AI, church life, and Christian responsibility.
Part one introduced the image of AI as jazz improvisation: useful variations are possible, but only when we keep returning to the melody.


When the Melody Drifts

AI, sermons, governance, and Christian responsibility

Last month, I suggested one basic tune for Christian thinking about artificial intelligence:

AI may assist human work, but it must not displace human responsibility.

That remains the melody. Everything else is variation.

Two areas deserve particular care because they sit close to the church’s identity.

The first is proclamation: sermons, teaching, worship material, and pastoral communication.

The second is governance: minutes, reports, proposals, policies, consultation papers, funding applications, and recommendations.

They look very different. One belongs near the pulpit. The other belongs in the board pack, the parish council papers, the committee folder, or the email chain no one quite remembers agreeing to. But both shape how the church understands itself, speaks, decides, and acts.

That is why both need the same discipline. AI may help. It must not take over the human responsibility.

Sermons, shortcuts, and the preacher’s vocation

Sermon preparation is one of the obvious pressure points.

AI can help a preacher explore ideas, test structure, clarify language, or notice where an argument has gone wandering off into the theological shrubbery. Those can be useful forms of assistance, especially when time is thin and Sunday is approaching with its usual lack of mercy.

But AI cannot carry the preacher’s vocation.

The preacher remains responsible for Scripture, theology, truthfulness, context, and pastoral care. A sermon is not merely religious content. It is an act of witness offered by a particular person to a particular community before God.

This matters because some people are already using AI for sermons. The question is not whether it will happen. It is whether it will be used badly, shallowly, invisibly, and without formation.

A lazy prompt can produce a sermon-shaped object very quickly: “Write a sermon on Luke 10.” The result may sound fluent, warm, and vaguely churchy. It may even contain some useful thoughts. But it has not prayed, listened, visited, grieved, argued with the text, or loved the congregation.

The danger is not only that AI may produce a shallow sermon. The deeper danger is that it may spare the preacher the very wrestling through which faithful preaching is formed.

Used well, AI may not save as much time as people hope. A responsible process still requires study, prayer, judgement, revision, and pastoral attention. It may become less like a ghost-writer and more like an over-eager study assistant: useful, fast, sometimes insightful, and absolutely not to be left alone with the pulpit.

The test is not whether AI was involved. The test is whether the preacher has still done the work only the preacher can do: listening to Scripture, attending to the community, wrestling with truth, praying honestly, and standing behind the word preached.

A polished draft is not the same as a faithful word.

The same applies to pastoral communication. AI may help us find words when we are tired, stuck, or facing a difficult message. But care cannot be outsourced. If a grieving person receives a beautifully phrased message that no human heart has actually attended to, something has gone wrong.

The machine may help with wording. It cannot do love of neighbour for us.

Minutes, reports, and the quiet shaping of decisions

The less-discussed risk is governance.

This is where AI may become deeply influential without looking especially dramatic. A system might summarise a meeting, draft a report, turn messy consultation notes into themes, prepare a funding proposal, polish a policy, or convert a board discussion into action points.

That sounds helpful. Often, it will be helpful. Many church committees are already drowning in papers, minutes, compliance work, and the sacred tradition of “can someone just tidy this up before it goes out?”

But governance documents do more than record information. They frame reality.

Minutes decide what is remembered. Reports decide what is emphasised. Proposals decide what seems possible. Policies decide what becomes normal. Recommendations decide what appears wise.

If AI is used carelessly in that space, it may not produce one spectacular error. It may do something subtler. It may flatten the tensions, overstate the consensus, omit the awkward minority concern, or make a weak recommendation sound inevitable.

Over time, that can move a church, charity, school, or committee away from the melody.

Not because anyone intended harm. Not because the machine “took over”. But because the humans trusted a clean draft more than they should have.

Good governance depends on more than tidy documents. It depends on faithful memory, honest disagreement, clear accountability, and responsible decision-making. AI can assist with the paperwork. It must not become the hidden author of institutional judgement.

A set of minutes can be grammatically perfect and still fail to tell the truth.

The point is not that AI is always unreliable. The point is that its errors are easiest to miss when the document looks polished, plausible, and administratively tidy.

That is why governance artefacts need human review not only for factual accuracy, but for framing, emphasis, omissions, tone, and accountability. The question is not merely, “Did the AI get the words right?” The question is, “Has this document helped us remain truthful, wise, and responsible before God and one another?”

When the tool starts acting

So far, many churches have mostly encountered generative AI: tools that produce text, images, summaries, slides, or ideas.

But another kind of AI is becoming more important: agentic AI.

The difference is simple enough:

Generative AI asks, “What can I produce?”

Agentic AI asks, “What can I do?”

That shift matters.

A tool that drafts an email is one thing. A tool that sends the email is another. A system that summarises invoices is one thing. A system that approves payments, updates records, contacts people, or acts with your login credentials is another matter entirely.

Once an AI system can act on your behalf, the governance questions become sharper.

Who authorised the action? What data could it access? What systems could it change? Could it email members, alter records, expose private information, or make commitments in the organisation’s name? Could someone stop it quickly if it began doing the wrong thing?

“The AI did it” is not an adequate excuse.

Delegation is not abdication.

You cannot responsibly delegate action to a system you do not understand well enough to stop.

Better cautionary tales

Churches need better AI cautionary tales.

Not horror stories. Not “the robots are coming for the organist” nonsense. We need practical, accessible stories that help ordinary leaders recognise risks early.

A church office does not need a PhD in machine learning. It does need enough wisdom to ask good questions.

What information are we putting into this tool? Can the result be checked? Could the system invent, distort, or misattribute details — and what would the consequences be if it did? Has it changed the framing, not just the wording? Has it left out dissent, uncertainty, or pastoral complexity? Is this appropriate for preaching, pastoral, financial, legal, safeguarding, or governance matters? Are we using AI to support human judgement, or to avoid it?

These are not anti-technology questions. They are stewardship questions.

Returning to the tune

AI can generate variations at astonishing speed. Some will be useful. Some will be silly. Some will be dangerous in precisely the way church technology often becomes dangerous: not because anyone meant harm, but because a tired person under pressure clicked the convenient button and hoped for the best.

We know that story. It usually begins five minutes before the service.

The church’s task is not to reject every new tool, nor to baptise every new platform with a cheerful “how wonderful”. Our task is to keep returning to the melody.

Behind every AI tool sit human choices about data, power, labour, profit, access, and whose voices are treated as normal. Christian responsibility is not only about checking an output. It is also about the systems we accept, fund, normalise, and resist.

We are called to truthfulness, wisdom, neighbour-love, stewardship, and accountability before God. We are called to care for people who may be helped by these tools and people who may be harmed by them. We are called to use power with humility, not merely efficiency.

AI may help us play some variations.

It cannot tell us what song we are meant to be playing.


Illustration created using AI image-generation tools for d|c|t.

Peter Lane is Principal Consultant at System Design & Communication Services and has over 30 years of experience with Technology systems.    We invite your questions, suggestions and ideas for articles.   These can be submitted either through the editor or by email to dct@dct.org.nz We also operate a website focused on building a community of people interested in improving how we use technology in churches, located at dct.org.nz