A church decision worksheet for using AI
AI Faith Monitor · 2026-09-27

Work through one real use of AI, identify the people affected and record a decision your church can explain.

Start with a concrete task
Choose one activity: preparing a newsletter, translating a notice, researching sermon background or answering a pastoral question. Do not approve “AI” as one undifferentiated category. Write the task, the proposed benefit and the person responsible. Compare it with the current way of doing the work. A tool that saves an administrator time may be useful even if it is unsuitable for confidential care.

Task: __________. Person responsible: __________. Benefit we will test: __________. People affected: __________.

Map what enters and leaves the system
List the information the tool receives and who can read the result. Use an invented example for testing. Names, prayer requests, private counseling details and children’s information do not belong in an experiment merely because entering them is convenient. Ask the supplier about retention, access and deletion, and keep a record of the answer. If the answer is unavailable, mark it unknown rather than assume that a paid account makes the use appropriate.

Information needed: __________. Information excluded: __________. Supplier explanation and date: __________.

Test the output against an original
Give the tool a task with an answer your team can check. Open every cited source. For a biblical passage, check the translation, surrounding argument and the difference between quotation and interpretation. Record an error as a concrete example: an invented reference, a misleading omission or a claim the source does not support. Fluency, warmth and agreement with the prompt are not measures of reliability.

Test input: __________. Claimed source: __________. What we checked: __________. What failed: __________.

Apply a Christian account of responsibility
James 3:13–18 describes wisdom through conduct, humility and peace. Applied here, that directs attention to how a church treats people, not simply whether its output sounds wise. A leader remains responsible for an adopted message; attributing its language to software does not discharge that responsibility. Love of neighbor also asks whose burden a convenience shifts. Does the new process save staff time by making a vulnerable person harder to reach?

Read James 3:13–18 in context: https://www.biblegateway.com/passage/?search=James+3%3A13-18&version=NIV

Our reason for this use: __________. Human relationship we must preserve: __________.

Make a bounded decision
Choose trial, revise or decline. For a trial, define the permitted task, information limits, named reviewer, date of reassessment and a stopping condition. For example, a church could trial help organizing public event notices for two weeks, with a staff member checking every date and no pastoral information entered. This is an illustration, not a report about a particular church.

Decision: __________. Trial ends: __________. Reviewer: __________. Stop if: __________. How we will explain the use to affected people: __________.

Review what actually happened
At the review meeting, compare saved time with time spent checking and fixing errors. Ask affected people what became easier or harder. Record both benefits and failures; a successful newsletter trial does not authorize AI counseling. Keep the decision with the test examples so a future leader can understand its limits.

Observed benefit: __________. Observed cost: __________. Feedback received: __________. Continue, narrow or stop: __________.