Formation, agency and public communication
Artificial intelligence is here to stay, and that is both good and bad news for the wordsmiths among us. On one hand, generative AI can strengthen preaching and public communication when it serves as a critical partner to established human competence. On the other hand, reliance on AI as a substitute for foundational skills, truth-seeking, creative agency and personal voice risks weakening both the formation and responsibility of the communicator.
In this article, I suggest that generative AI should be used as disciplined augmentation rather than substitution. The technology offers the possibility of vastly extending human capacities for research, criticism, creativity and communication while preserving the intellectual, relational, ethical and spiritual practices through which competence, judgment, authentic voice and—especially in preaching—the formation of the communicator are developed.
Professional and business people alike ignore generative AI at their peril. The challenge for all of us is to master the tools it offers without surrendering agency and authenticity.
Authorship, authenticity and disclosure
There is an obvious difference between asking AI to identify weaknesses in a sermon one has written and asking it to produce an entire sermon that is subsequently delivered substantially unchanged. Between these extremes lies a continuum of practices including research assistance, brainstorming, structural advice, editing, rewriting and substantial co-generation.
Public communication has always involved collaboration. Political leaders employ speechwriters; authors use editors; academics receive comments from colleagues; preachers consult commentaries and draw upon theological traditions. AI does not create collaborative authorship, but it changes its scale, speed and opacity.
Congregations nevertheless reasonably expect preaching to represent the preacher’s own theological, pastoral and spiritual labour. Audiences similarly assume that a speech attributed to a public figure substantially represents that person’s considered position, even when professional speechwriters have assisted. Extensive undisclosed substitution of AI-generated material may therefore raise legitimate questions about authenticity and intellectual honesty.
Not every use of AI requires disclosure. Asking a system to detect repetition or suggest alternative headings is not equivalent to presenting an AI-generated sermon as one’s own. The ethical question is proportional: at what point has assistance become sufficiently substantial that an audience’s reasonable assumptions about authorship become misleading?
Lived experience and human relationship
Another risk is the separation of public communication from lived human experience. This is especially significant in congregational preaching, where there is a strong expectation that the preacher’s content and tone will pair well with audience needs.
A sermon is not simply a theological essay that is orally delivered. It is an ecclesial and liturgical act shaped by relationships among preacher, biblical text, Christian tradition, congregation and the wider world. Preachers speak among people whose lives they know, however imperfectly: people who have experienced bereavement, illness, injustice, reconciliation, doubt, joy, moral failure, disappointment and unexpected grace. Responsible preaching emerges partly from sustained attention to those realities.
Generative AI has no equivalent participation in human life. It can produce sophisticated language about grief, prayer, reconciliation, injustice and hope because it has learned from human discourse concerning these realities. It does not itself grieve, pray, reconcile, suffer injustice or hope. This does not make AI useless for pastoral or theological reflection. It can compare perspectives, formulate questions and help a preacher test language. But it cannot replace the forms of human presence from which pastoral wisdom develops.
Prayer, silence, pastoral conversation, attentive reading, worship, observation, friendship and participation in community are therefore not inefficient preliminary stages that technology can conveniently eliminate. They belong to the ecology from which authentic preaching emerges. The same principle applies beyond the pulpit. Political speeches, advocacy, lectures and ceremonial addresses require attention to particular communities, histories, anxieties and aspirations. Effective public speaking begins with patient and attentive listening.
Formation of the preacher
For Christian preaching, however, the argument goes further. Sermon preparation is not merely a production process designed to generate religious content. It is itself potentially a spiritual and theological practice. Christian traditions differ in their precise theology of preaching, but commonly understand it as involving Scripture, proclamation, vocation, community, prayer and dependence upon the Holy Spirit. The preacher does not simply extract information from a biblical text and transfer it to an audience. The preacher also stands under the text, encounters its demands, tests personal assumptions against it and seeks to embody what is proclaimed.
This gives AI-assisted preaching a theological dimension not necessarily present in ordinary discussions of productivity. If part of the purpose of sermon preparation is the formation of the preacher, automating that process may remove something valuable even when the resulting sermon remains rhetorically competent. AI can explain the theology of Paul, Augustine, Barth or Bonhoeffer; it can compare interpretations of Romans or Isaiah; it can analyse a difficult Greek construction, or identify an exegetical controversy. But AI cannot experience repentance, obedience or spiritual formation on the preacher’s behalf.
The sermon is not the only product of sermon preparation. The preacher’s intellectual and spiritual growth is also one of its products.
AI can also enlarge human agency
A balanced assessment must acknowledge that AI does not invariably diminish agency. For some users it may significantly enlarge it. It can assist people communicating in a second language, support some people with disabilities, offer formative feedback to inexperienced writers, and give ministers with limited access to theological libraries or professional editorial assistance resources that were previously difficult to obtain. Used pedagogically rather than substitutionally, such assistance may contribute to the acquisition rather than erosion of competence.
This qualification shows why the objective cannot be preservation of difficulty as such. Some difficulties form us; others merely exclude us. Removing unnecessary barriers can be liberating. The important question is whether the activity being delegated is merely instrumental or is itself part of the capacity we are trying to develop.
This yields a useful governing principle: we should not ask simply which tasks AI can perform, but which human capacities the performance of those tasks is supposed to develop. Formatting references, detecting repetition, organising notes or generating alternative headings may often be delegated with little loss. Interpreting Scripture, constructing an argument, exercising moral judgment, discovering one’s voice, listening pastorally or determining what one actually believes all deserve greater caution because the process of performing these often-mundane tasks may itself be formative.
The wider ethical context
AI should not be imagined as an immaterial or ethically neutral assistant. Generative systems are largely developed by powerful technological organisations, require substantial computational infrastructure and energy, and depend upon enormous quantities of human-created material. Their development raises wider questions concerning environmental impact, intellectual property, labour, privacy, access and the concentration of technological power.
These issues lie beyond the immediate question of sermon and speech preparation and should not be exaggerated into an argument against AI. Nevertheless, ethically serious users should recognise that apparently effortless digital assistance rests upon substantial material, economic and cultural infrastructures. Responsible use includes attention not only to what AI does to individual users but also to the systems within which these technologies are developed and deployed.
From substitution to disciplined augmentation
The appropriate response is neither technological enthusiasm nor technological abstinence. Generative AI is best understood as an instrument whose value depends upon the practices within which it is embedded. Communicators might preserve AI-free stages of preparation, particularly at the beginning of important projects; continue practising foundational skills independently; draft significant arguments and passages themselves; verify important factual claims and sources; and periodically undertake complete tasks without AI assistance. These practices help determine whether AI-assisted performance continues to rest upon independently possessed competence.
AI may be especially valuable when used adversarially rather than merely generatively. Instead of repeatedly asking, “Write this for me,” the preacher or speechwriter might ask: “What is weak in my argument? What evidence have I overlooked? What would an informed critic say? What alternative interpretation deserves consideration? Which assumptions am I treating as self-evident?” Used in this manner, AI may increase rather than diminish intellectual effort by creating additional opportunities for criticism, comparison and revision.
Perhaps the most useful discipline is to ask, from time to time, whether one could still perform a particular task competently without AI. If the answer gradually becomes no, it’s likely that what you assumed was augmentation may have become dependency. The deeper danger is not simply that AI sometimes produces factual errors or bland prose. It is that powerful technologies enable us to produce the appearance of intellectual competence while bypassing practices through which competence, judgment, creativity and wisdom are formed.
Growth of the preacher
For preaching, this distinction has theological as well as professional significance. The demanding practices of reading, interpreting, thinking, praying, listening, imagining, writing and revising do more than produce sermons. Over time, they contribute to the formation of the preacher. Research, argument, imagination, composition and revision similarly form the speechwriter and public communicator. They are not merely obstacles between the communicator and the finished product; they are constitutive elements of the craft.
The goal should therefore be disciplined augmentation: using generative AI to extend research, sharpen criticism, widen imaginative possibilities, remove unnecessary barriers and improve communication while deliberately preserving the intellectual, creative, relational, ethical and spiritual practices that technology cannot responsibly replace. The decisive distinction is not between human beings and machines, or even between using AI and refusing to use it. It is between AI that extends human formation and AI that circumvents it.
AI should help us to practise our craft more intelligently, critically and creatively. It should not relieve us of the need to acquire, exercise and preserve that craft. For preachers, the sermon is not the only thing being formed in sermon preparation. The preacher too is being formed. Generative AI will never grow in holiness and godliness, but you have the opportunity to do so, in part through your preaching.
Rev Dr Rod Benson is General Secretary of the NSW Ecumenical Council and a minister of the Uniting Church in Australia serving at North Rocks Community Church in Sydney.
Image source: author
All Told 