Level
1
General chat
Part 2: Using AI well
AI for
Ministry:
4 Levels
Practical guidance and live demos for pastors and administrators
Eric Louw, EC3, Southern Union Evangelism Conference
From expectations
to a working practice.
What AI promises and what it delivers
Three promises sell almost
every efficiency tool.
Save ten hours a week.
Improve the bottom line by 56%.
Stay ahead of everyone else.
Saved hours rarely become leisure. Costs cut in one area usually get assigned to another.
The promise
A 20-hour workweek while AI bots complete our responsibilities.
Human nature, competition, and capitalism undermine it.
The likelier result
A raised standard from extended capability and output.
Higher expectations quickly become the new competitive norm. AI can still save time. Plan for it to extend performance and skills first.
Activity
List every responsibility you carry.
Write out and describe all of your responsibilities, especially the most repetitive ones, without asking whether AI could or should do them.
Cross off everything that would be inappropriate to outsource. What remains is your list for the end of this session.
Privacy, copyright, and your role
Know what is entrusted to you before you type.
Where private things resurface.
Conversations thought private can be requested from AI companies by law enforcement.
Uploading, adapting, or imitating a person’s photo, likeness, or copyrighted work carries legal and ethical risk.
Never connect AI to financial institutions. If you connect email, use an account that logs into nothing else important.
A question for your context
What data must be kept separate from any possibility of AI access?
Approach AI as a coach or resource.
Replacement
Outsource the skills you were hired for, and your employment may be at risk.
Assistant
Enhance what you do and who you are becoming. Prioritize growth, not only a more efficient outcome.
Limit outsourcing to outcomes with no spiritual requirements and limited operational liability.
Five ways to keep AI in the assistant role.
Four levels of AI use, with live demos
AI is built to answer, even when there is no answer.
Ask for a date and timeline for each book of the Bible and it will answer, even if most of the dates are fictional.
Thinking time and model choice
change the result.
Free chat on a website
Seconds of compute
Limited time, cheapest models, rapid search. The average Google AI query gets a few seconds at best.
Paid tier or installed app
An hour or two of work
More capable models and higher limits, especially when the goal is clearly and carefully defined.
Each level adds capability and risk.
Agentic use
Claude Desktop, ChatGPT Work / Codex, Grok Bot, Google Antigravity
Learning enhancement
Google Notebook, Superwhisper, AdventAI.app, Devonthink MCP
Simple searches
Google, Perplexity, Copilot
General chat
ChatGPT, Gemini, Claude, Grok, Meta / WhatsApp
More capability and autonomy at each level; privacy and verification matter more at every step.
Prompt intentionally: context, task, format.
Iterate with a scientist’s mindset; good questions are often open ended.
Framing biases the answer: “Do not think of a pink elephant” puts one in the picture.
Priority: Explain your goal and what a desired outcome should look like.
When facts matter, ask for sources.
Risks at this level: fabricated sources, accidental plagiarism, and answers shaped by thin compute and cheap models.
Specialized tools for study and nuanced finding.
Risk: overdependence weakens memory, creativity, and critical thinking. Verify against the primary text.
One month free if you sign up today at EC3
(no credit card required)
AdventAI.app
BibleWorld.ai
Agents work toward an outcome, not a single prompt.
Usage limits reset in five-hour or weekly segments, and models differ in cost. An installed agent can reach whatever is stored on the device.
Treat conversations like
members of a team.
Separate roles. Long chats drift as the context window fills. Give separate tasks and roles to separate conversations.
Handoff files. Each role keeps an updated text or markdown file so another agent can take over if something breaks.
Skills. Turn a process you coached into a text file any chat or agent reads to repeat it exactly.
Count the cost and work the checklist
Count the cost before you adopt AI.
Subscribing to an enterprise solution and learning it costs both.
The cost cascades to training other workers too.
A church or ministry may need to change its context and operations.
Success goes beyond good prompting: state objectives and outcomes, and adjust procedures.
Before you adopt: decide and prepare.
As you adopt: plan, assign, and test.
Back to your list
1Take the responsibilities you did not cross off.
2Ask an AI to outline areas where it could help.
3Place each item on a level, then run the checklist.
Which responsibilities remain, and where could AI extend you?
Questions
Join 5,500+ Adventist AI users.
Scan the code with your camera app or visit AdventAILab.com and tap “AdventAILab” to join the Facebook group.
Eric Louw, AdventAILab.com
