church tech

AI Should Be a Sheepdog, Not a Shepherd

By Michael Tribett · July 9, 2026 · 18 min read

The louder question is whether AI can do something in ministry. The better one is whether it should, and what it should never do.

Key takeaways

The frame is "can" versus "should." AI can write, summarize, automate, and even send messages if you let it. In the church, the decisive question is what must stay human, not what is technically possible.

ChMS tools map vertical relationships well and horizontal ones poorly. They track attendance, giving, and workflows between the institution and each member. They do not show who actually knows whom, or who is isolated while looking fine on paper.

Collie is a sheepdog, not a shepherd. The assistant in FlockConnect surfaces who may need attention and can draft a next step. It never sends a message, writes to a record, or changes anyone's care on its own. A person reviews and approves every action.

Some impressive features are off the table on purpose. Auto-intro texting, simulated pastoral care, and AI that acts without a human in the loop would trade the very relationships the product exists to protect.

AI can help a solo founder build, but it cannot pastor. Using AI to move faster on code is one thing. Outsourcing preaching, discernment, presence, or friendship is another, and the church should refuse that road.

Quick answer: Should AI replace pastors in church ministry?

I built Collie in FlockConnect to surface who may need attention and draft a next step, never to pastor on its own. AI can summarize, automate, and draft, but it should not send messages, write to care records, or replace the presence only a human image-bearer can offer. The right question is not whether AI can do ministry tasks, but which tasks must stay human, and I believe the church should draw that line before adoption, not after regret.

Should AI be a shepherd in the church?

No. AI in church ministry should help a pastor see who is drifting and clear the busywork that hides them. It should not replace shepherding, pastoral presence, spiritual discernment, or the slow human work of friendship. The healthy posture is a sheepdog that alerts the shepherd, not a shepherd that acts on its own. FlockConnect was built around that line: Collie surfaces and drafts; a person decides.

The question that has shaped how I build

Over the last year, building FlockConnect, I have landed on a question that matters more than whether something can be built: whether it should be built, and what it should never do.

That boundary has shaped most of my product decisions. AI can write, summarize, generate, automate, send messages, create content, and even make decisions if you let it. Capability is not permission. In the church, the gap between those two is enormous.

When I started building FlockConnect, I was not trying to build another Church Management System. There are already plenty of ChMS platforms out there, and many of them are good at what they do. They help churches track attendance, giving, child check-in, serving schedules, forms, workflows, and communication. All of that helps. What I kept seeing in my own church and in church tech, though, is that most ChMS systems are really good at vertical communication: staff to ministry leaders, leaders to volunteers, church to congregation.

What they do not really help churches understand is horizontal relationships: who knows whom, who feels like they belong, who would get a call if they were struggling, and who shows up every week but still feels isolated. That is a different question than, "Did they check in on Sunday?" For more on why that horizontal layer matters to pastoral care, see church relational health matters.

When a close friend looked connected but was not

That question became personal for me because of something that happened at my own church. A close friend was really struggling with feeling connected. The hard thing was that, from the outside, everything looked fine. He was in a small group. He was volunteering. He was around people. He was over at our house all the time. So I assumed he felt connected. Our pastors assumed he felt connected. Other people probably assumed the same thing.

But then one day he told me he was thinking about checking out other churches because he did not really feel connected. And that hit me pretty hard. Because on paper, everything looked fine. He was attending. He was serving. He was in community, at least technically. But he did not feel known.

That is why I built FlockConnect. I want to help pastors see who is drifting before they are gone. If that could happen with one of my closest friends, someone I saw all the time, then how many people are in churches every week who look connected in the system but are actually isolated? How many are in a group but do not feel known, serving without real friendships, slowly drifting while nobody notices?

What the research suggests (without reducing people to numbers)

A friend also shared research with me that helped put language around some of this. The basic idea, carried to pastors by Win and Charles Arn from assimilation work Flavil Yeakley first measured, is that members who form real friendships in their first months tend to put down roots and stay. Members who move through programs and events but never actually connect to anyone tend to drift, often without a word. I do not think relationships can be reduced to a number. People are not data points. But I do think that kind of signal gives pastors something important to pay attention to.

Are people actually forming relationships? Do they have people they can go to? Do they feel known? Are they connected to the body, or are they just attending the event? Those are the kinds of questions FlockConnect is trying to help churches ask. For a fuller look at the research without turning friendship into a scorecard, see the seven friend threshold.

The church is a people, not only an event

Theologically, the church is more than a Sunday morning event. The sermon, worship, and gathering all matter, and I am not minimizing any of that. The church is also a people: a body where members know one another, pray for one another, bear burdens, and walk together.

When people show up every week but are not known, when they sit in the room but stay isolated, when they serve until they burn out and nobody notices, pastors feel it before any dashboard does. Most already know someone is slipping through the cracks. FlockConnect is not inventing a new pastoral worry. It is trying to make visible what leaders already care about.

John Piper draws a related line in Brothers, We Are Not Professionals. His warning is that ministry drifts toward a managerial self-understanding, where competence at running things stands in for the care of souls. The professional manages programs and measures output. The shepherd knows the flock by name and notices when one is missing. Software that only sees records will always favor the first instinct. The second one is what pastors are actually called to.

AI should be a sheepdog, not a shepherd

I explain my AI philosophy with a sheepdog metaphor, which is why FlockConnect's assistant is named Collie. I had a border collie once. The image fits: a sheepdog alerts the shepherd. It does not own the flock, make the final call, or love the sheep the way a shepherd must.

Collie can flag that someone may need attention, surface an attendance pattern, notice a group dropout, organize a care report, or point out relational disconnection. Collie should not pastor the person.

Tim Keller makes a kindred point about pastoral presence in Every Good Endeavor: faithful work in ministry is not only about output, but about showing up for real people in real situations. A tool can prepare the ground for that presence. It cannot supply the presence itself.

What I will not build, even if it demos well

There are features I could build that I am not going to build. And some of them would probably look impressive in a demo. For example, someone could say, "If Collie knows two people might connect well, why not just have Collie text both of them?" Let's say Ryan and John both like rock climbing, and they both have Tuesday mornings open. Technically, a system could send a text and introduce them.

But I do not want to build that. Auto-intro texting would remove the very thing I am trying to protect: real connection, real pastoral care, and the appearance of care when only a bot sent a message.

What I would rather happen is Collie tells the pastor, "Hey, Ryan and John may be a good connection." Then the pastor sees Ryan on Sunday and says, "Hey, have you met John yet? He is also into rock climbing. I think you two would really get along." An automated text is not the same work. A pastor making the introduction in person is. Collie surfaces the connection; the pastor stewards it. For the working definition of the connection I am trying to protect, what is a church connection is the place to start.

How Collie actually works

Pastors cannot see everything, especially as a church grows. Someone stops attending. Someone disconnects from group. Someone's serving rhythm changes. Someone has a care note from a leader. Someone says they feel disconnected. Someone has been at the church for months but does not seem to have real relationships yet. That information can be scattered all over the place, and a pastor may not see it until much later.

So if AI can surface scattered signals before a pastor would have seen them, that can help. The ministry still belongs to the pastor. AI can bark. The shepherd still has to shepherd.

Collie surfaces who looks isolated and can draft a note or a suggested next step. Then it stops and waits. It does not send the message. It does not write to a record. It does not change anyone's care on its own. A person reviews and approves every action. That boundary is the product, not a limitation we are apologizing for. A fuller theology of why presence stays human is in when the robot writes the reminder.

I am not anti-AI. I am pro-boundaries.

I want to be clear that I am not anti-AI. I use AI every day. I am building FlockConnect with the help of AI. I would not be able to build this the way I am building it without AI. I am a solo founder. I am a full-time dad. I do not have a team of engineers. I do not have outside funding. AI has helped me build something that, a few years ago, I probably would not have been able to build.

But AI can slide from tool to replacement fast. Cleaner code from AI is one category. Organizing information or surfacing a pattern is another. Replacing pastoral presence, discernment, creativity, preaching, care, or friendship is something else entirely, and churches should refuse that road.

That is where I would slow down. A companion piece on the two ways churches get AI wrong, rejecting tools or expecting software to do relational work, is the AI question churches are getting wrong.

Just because we can does not mean we should

Michael Crichton nailed this in Jurassic Park. He put it this way:

Scientists are actually preoccupied with accomplishment. So they are focused on whether they can do something. They never stop to ask if they should do something.

Michael Crichton, Jurassic Park

I think that line lands hard right now, especially with AI in the church. We can build AI pastors, sermon writers, counselors, prayer bots, and systems that message, follow up, disciple, and generate spiritual content on autopilot. "Can we?" is the wrong stopping point. "Should we?" is the real question.

A lot of the time, the answer is no. Or at least, "Not like that."

Why I am concerned about AI-written sermons

Sermons are where I draw the hardest line. I do not think pastors should use AI to write them. A sermon is a pastoral act, not a blog post or a talk about a passage.

A pastor needs to know the text, but he also needs to know the people. He needs to know what his congregation is walking through. He needs to know their struggles, questions, temptations, suffering, immaturity, hopes, and needs. He needs to faithfully bring the Word of God to the people God has entrusted to him. AI cannot do that for him.

A pastor might use AI for narrow prep tasks: research, tightening wordy sentences, brainstorming an analogy, surfacing another view on a passage. Outsourcing the sermon itself is a different matter. He should not ask AI to write it and preach the output unchanged.

At that point, something has gone wrong in how we understand preaching, pastoring, and shepherding people through the Word.

Creativity, the image of God, and what we might give away

Theologically, humans were created to create, cultivate, steward, and build. When generative AI does the creating for us, we need to ask what we are giving away: using a tool to help us create, or letting the tool create instead of us.

Those are not the same thing, and the line blurs in practice. I use AI to build, edit, organize thoughts, and move faster. The question still matters: is AI helping me be more faithful to what I am called to do, or replacing something I am called to do?

What building with AI has taught me

When I first started, I used tools like Lovable and Bolt to prototype and get something on screen. Sometimes you need to move instead of waiting for the perfect team, funding, timing, or plan.

I did that. I also learned that building with AI creates a mess fast if you do not know where you are going. A vague prompt can produce something impressive and still wrong.

The first version of FlockConnect was too complicated. I kept meeting with pastors, getting ideas, adding features, changing things, and trying to make it work. And after a while, the architecture was messy. Even when I moved to stronger tools, the agents had a hard time understanding the codebase. They would fix one thing and break another thing.

So eventually I decided to rebuild it. Not because the idea changed, but because I had more clarity. I had talked to pastors. I had gotten feedback. I had a better sense of what the product needed to be. I had a better sense of what it should not be. And I had a better sense of the architecture.

That has changed how I build. Now I am much more focused on specs, structure, boundaries, and clarity before I ask AI to build. AI can help you move fast, but if you are not clear, it can help you make a mess fast too.

Talk to pastors before you build

Anyone building church tech should talk to pastors and leaders before writing code: does the church need this, does it already exist, does it solve a real problem, where does it overstep, and where does it make ministry less human?

Because you might have a great idea in your head, but if you never bring it into real conversations, you might be building something nobody actually needs. And with AI, that gets even more dangerous because you can build so quickly. You can move fast in the wrong direction.

How FlockConnect holds the line

The goal of FlockConnect is not to make the church more automated. It is not to replace pastors. It is not to replace small group leaders. It is not to replace the normal, slow, human work of relationships. The goal is to strengthen the human element of the church.

FlockConnect is a Church Relationship Manager, the relational layer a ChMS was never built to hold. It works alongside the church management system a church already runs. A church on Planning Center connects through the official two-way Planning Center integration. A church on anything else brings its people in by CSV. For why the category exists at all, see why FlockConnect is the world's first ChRM.

I want to help pastors see who is drifting before they are gone, help churches notice people who might otherwise be missed, and help leaders care with more clarity so people are known by pastors, leaders, and the body of Christ, not merely tracked by a system.

I talk to AI more than I talk to most people these days. It does not spiritually feed me, truly know me, or challenge me the way my wife, my pastor, or a real friend can. Useful, yes. An echo chamber that tells you what you want to hear, also yes. That is one more reason the church needs to become more human, not less.

AI should be a sheepdog, not a shepherd. It should help the shepherd see. But the shepherd still has to shepherd.

About the author

Michael Tribett is the founder of FlockConnect, a Church Relationship Manager built to help pastors see who is connected and who is drifting. He holds a Master of Divinity in Christian Ministry from Southeastern Baptist Theological Seminary, where he focused on missions and discipleship, and he serves as a small group leader at his church in the Raleigh, North Carolina area. FlockConnect is an official Planning Center partner.

Frequently asked questions

What is the sheepdog metaphor for AI in the church?

The frame is "can" versus "should." AI can write, summarize, automate, and even send messages if you let it. In the church, the decisive question is what must stay human, not what is technically possible.

Can Collie send messages to church members on its own?

Why should churches be careful with AI sermon tools?

What should AI never do in a church?

Over the last year, building FlockConnect, I have landed on a question that matters more than whether something can be built: whether it should be built, and what it should never do.

How does FlockConnect use AI without replacing pastoral care?

Should AI replace pastors in church ministry?

No. AI can clear busywork, surface patterns, and draft a next step, but shepherding requires presence, discernment, and love that software cannot supply. The healthy posture is assistive: a sheepdog that alerts the shepherd, not a shepherd that acts alone.

What does "sheepdog, not shepherd" mean for Collie?

Collie is the advisory assistant inside FlockConnect. It surfaces who may need attention and can draft a note or suggested next step. It never sends a message, writes to a record, or changes anyone's care on its own. A person reviews and approves every action.

Will FlockConnect auto-introduce members by text?

No, and I do not plan to build that. Introducing two people is a pastoral, relational act. Collie may flag that two members could connect well, but a pastor or leader still makes the introduction in person.

Can pastors use AI for sermon preparation?

Small uses around research or tightening prose may be fine. Outsourcing the sermon itself is a different matter. Preaching is a pastoral act that requires knowing both the text and the people God has entrusted to the preacher.

How does FlockConnect connect to a church's existing systems?

FlockConnect adds the relational layer on top of the ChMS a church already runs. Planning Center is the one native, two-way integration. Every other system connects by CSV import.

Is it wrong for a church to use AI at all?

No. The question is where the line falls. Using AI to organize information and surface who looks isolated can free a pastor for presence. Letting AI send care, make pastoral decisions, or replace friendship crosses the line.

Why did you rebuild FlockConnect after prototyping with AI tools?

The first prototype moved fast but grew messy. Pastor conversations gave me clarity on what the product should and should not do. Rebuilding with clearer specs and boundaries let AI help on a foundation that would not break every time we fixed something.

What is FlockConnect priced at?

FlockConnect is priced by church size, not by staff or volunteer seats, with a free trial. The people who serve a church are never the line item.