Most church software is built by people who have never sat with a grieving family at a kitchen table. That is not a criticism of engineers. I was one for fifteen years before I ever preached a sermon.
But picture the call that comes at two in the morning, the one where a man says his father is dying and he does not know who else to phone. Somebody wrote the code that answers that call. Somebody decided, months earlier and hundreds of miles away, whether those words count as "urgent." Somebody decided whether the form he fills out afterward is a row of data or a person in the worst hour of his life. Those decisions were made by people who have never been on the receiving end of that phone call. That is not malice. It is simply a fact about who usually writes the code, and it shapes every decision inside the product.
If you remember one sentence from this article, make it this one: the most important feature in church AI is knowing when to stop talking and hand the person to a human.
Who I am and why I am building this
I spent fifteen years as a software engineer. Then I spent fifteen years pastoring in Reformed churches in Ontario.
Which meant hospital hallways. Funeral planning. Budget meetings that ran long. And a voicemail box, blinking in an empty office, full of messages nobody had time to return. I knew the names attached to most of those messages. That is its own particular loneliness.
Now I have returned to engineering to marry the two halves of my life and build the tool I wish I had owned as a pastor. ChurchWiseAI is a set of AI tools for churches (a phone answering agent, a website chatbot, and a simple church website) built by a pastor and engineer so that every conversation ends with a real human, not with the machine.
I do not think that makes me a better engineer than the people at large software companies. It gives me a different set of instincts about what a church's front door is for. When I read the words "who builds church AI software," my honest answer is that it is mostly built by very capable people who have never taken the calls their software will now answer.
I also believe in a God who places people where they are needed. Esther's cousin told her that perhaps she had come to her royal position for such a time as this. I am no Esther. But I have spent thirty years being prepared, half in code and half in congregations, for a moment when churches would need someone who could translate between the two. So I am building.
What dual experience actually changes
It is easy to say "built by a pastor" on a landing page. It is harder to explain what that changes in the software. Here is what it changes for me.
I know what a pastoral emergency sounds like. It rarely arrives with the word "emergency." It arrives at 11pm, in a flat voice, as "I was just wondering if someone could call me back sometime." An engineer building a generic answering bot hears a low-priority callback request and files it accordingly. A pastor hears the flatness in the voice and starts driving. So the system is built to treat anything that could be a crisis as a crisis, and to route it to a real person right away rather than to a queue.
I know that a form submission is a person. Someone sat at a kitchen table and typed out the worst thing in their life into a little box on a church website. They read it over. They almost closed the tab. Then they pressed send. That is not a row in a database. It changes how the confirmation reads, how quickly a human is notified, and what the AI is allowed to say in reply. It is never allowed to counsel. It is allowed to acknowledge, to comfort briefly, and to promise that a person will follow up. Then a person must actually follow up.
I know that silence or a referral can be the right feature, not a missing one. Vendors tend to measure success by how many questions the AI resolved without a human. In a church, that metric is backwards for the things that matter most. When someone asks whether God is punishing them, or types that they do not want to be here anymore, the correct behaviour is to stop being helpful in the software sense and become helpful in the human sense: give a crisis line, wake the pastor, step aside. The AI is a bridge to real people, never the destination.
Three decisions that came from the pulpit, not the vendor playbook
The AI is never the last voice. Every product in ChurchWiseAI is designed so that the conversation ends with a human being notified and the visitor told who that human is. A typical vendor treats "deflected to self-service" as a win. I treat it as a failure when the topic is grief, faith, or fear.
The system does not keep secrets from the church. The AI does not hold pastoral conversations in a private memory that the pastor never sees. Everything a visitor shares that needs care is surfaced to a real person on the church's team. AI should not become a confidant that quietly replaces the shepherd.
Pastoral boundaries are written into the prompts, not left to chance. The agent is explicitly told it is not a counsellor, not a pastor, not a doctor, not a lawyer, and not a financial advisor. When a conversation drifts toward any of those, it is instructed to name a real human who can help and to make the handoff. That is not a limitation we apologise for. It is the design.
How this compares with the usual approach
Picture two teams building the same church phone agent, in two different rooms.
The typical vendor approach: maximise resolved calls, minimise human involvement, treat every message as a ticket, measure success by deflection rate, and let the AI improvise on sensitive topics because the model "handles it well."
The pastor-built approach: minimise the distance between a hurting person and a real human, treat every message as a person, measure success by whether the right human heard about the right call in time, and refuse to let the AI improvise on grief, crisis, theology, or medical questions.
Both rooms are using similar technology. The difference is in the judgment about what the technology is for. That judgment comes from having sat on the other side of the phone.
What this does not claim
I am not going to tell you that hundreds of churches have transformed their ministry with this. ChurchWiseAI is young, with a small number of real customers, and I would rather you hear that from me than discover it later. What I can tell you is what the tools are built to do, who built them, and the convictions behind every escalation rule. Those convictions are not going to change when the customer list grows.
Frequently asked questions
Is church AI software built by pastors actually different from regular AI tools?
The underlying models are often the same. What differs is the design judgment: how emergencies are recognised, what the AI is forbidden to say, and how quickly a human is brought in. A pastor who has taken the late-night calls builds those rules from experience rather than from a generic support-desk template.
Can an AI answering service replace a pastor or church secretary?
No, and it should not try. A well-built church AI answers the routine questions (service times, directions, who to talk to about baptism) and routes everything that touches a person's life to a real human. Anyone selling you AI as a replacement for pastoral care is selling you something a church should not want.
How do I evaluate a church AI vendor before trusting them with our front door?
Ask three things. First, what happens when someone in crisis calls or messages, step by step. Second, does the pastor see everything a visitor shares, or does the AI keep its own private record. Third, who built it and have they ever done the pastoral work the software now touches. If the answers are vague, keep looking.
What does "AI as a bridge" mean in practice?
It means the AI's job is to get a person to the right human faster, with less friction, and with dignity. It gathers the information a pastor needs, offers immediate comfort where appropriate, and then steps aside. It does not counsel, diagnose, or give legal or financial advice.
Is it safe to let an AI talk to grieving or vulnerable people at all?
It is safe only when the boundaries are explicit and enforced in the system, not hoped for. That is why the escalation rules and the "I am not a counsellor" instructions are written directly into the agent and tested. Even then, the AI is a first responder that dials the human, not the human itself.
A human next step
If you are a pastor or a board member sitting with this decision, weighing whether to trust an AI vendor with your congregation's front door, I would rather talk with you than have you read another page. Visit churchwiseai.com, try the demo, and then book a call with me. I will tell you plainly where the tools fit your church and where they do not. And if the honest answer is that you do not need this yet, I will say so.
John Moelker is an engineer (15 years) and pastor (15 years), founder of ChurchWiseAI and WiseAI Agency.