Companion to Shrewd as serpents. For church leaders. Revision 1. ~2,900 words.
Shrewdness is a word about means, and a claim about means that stays abstract is not a claim about anything. So here is exactly what it consists of, for a church, this year. Not principles. The actual content of the competence.
Before any of it: I build these systems and I sell this training. Weigh the whole piece accordingly, and note that most of what follows costs nothing and does not involve me.
I will start with what a shrewd leader knows, because most of the mistakes are knowledge failures rather than courage failures. Then what a shrewd church does. Then, at some length, what it refuses, because refusal is part of the competence and not a limit on it.
What a shrewd leader knows
That the tool is worst at the thing everyone asks about first
Every conversation I have with a pastor arrives at sermons within four minutes.
Here is the honest answer. These systems can produce a competent, structured, theologically unobjectionable sermon on almost any text in about ninety seconds. It will be usable. It will also be assembled from the average of everything ever written about that passage, which means it will contain no encounter, no risk, and nothing that came from you sitting with the text until it did something to you.
A shrewd leader understands why that matters, and it is not squeamishness. Preaching is not the transmission of content about a text. If it were, the machine would be fine at it. Preaching is a person who has been changed by a passage standing in front of people they know and telling them what they found. The congregation is not receiving information. They are watching someone who has been somewhere.
The average of the internet has not been anywhere.
So the sermon question is the wrong question, and the fact that it comes first is itself the tell. It reveals that the leader is thinking of this as a writing tool, which is the least interesting and most dangerous frame available.
That the real gain is memory, not efficiency
Here is what I think most churches have never noticed about themselves.
You are an institution whose entire theology is continuity. You baptize people and expect to bury them. You tell a story that spans generations. And you have catastrophic amnesia.
A pastor takes a call. Within eighteen months, most of what the previous pastor knew is simply gone. Not the records. The knowledge. Which family is still hurting from a split fifteen years ago. Why the trustees will not discuss the parking lot. Which of the older members quietly funds most of the benevolence. What was tried in 2011 and why it failed. Who has been asked to serve too many times and who has never been asked once.
A long-tenured member dies and takes forty years of congregational history with her, and nobody thought to ask.
You have twenty years of sermons in a folder, thirty years of newsletters in a box, session minutes going back further than that, mission trip reports, annual reports, prayer chains, funeral homilies. All of it is a record of what this congregation has believed and done and suffered, and none of it is reachable. Nobody can ask it anything.
The shrewd leader sees that as the asset. Not because it makes anything faster, but because the thing this technology is actually good at is making a large, scattered body of material answerable. That is not an efficiency play. It is an antidote to the specific way churches lose themselves.
That something is already answering questions about your church
Somebody in your community asked a machine about your church this month. Someone considering visiting, someone whose kid is thinking about your youth group, a reporter, a family looking for a funeral.
It answered. It did not say it was unsure. And it answered from whatever it could reach, which is your website, your Google listing, a couple of reviews, and anything anyone else has said about you. Not your actual theology. Not what you have learned in twenty years of being in that neighborhood. Not the sermon series that would have told them exactly who you are.
You do not get to opt out. The only decision available is whether the thing answering has anything real to work with.
That the cost curve moved and nobody told you
The kind of system that used to require a budget most churches will never have can now be built for a small fraction of it. That is not marketing. It is the actual reason I have a company, and it is also the reason I can publish the method free and still have one. [FIGURES: the before and after numbers, with what each actually bought, or this paragraph stays qualitative. I would rather it stayed qualitative than carried a number I cannot source.]
Practically, this means capabilities that were previously available only to churches with a communications department are now available to a congregation of ninety. That window will not stay open in this form. Right now the tools are cheap, the stakes are low, and being wrong is survivable. Both of those conditions are temporary in opposite directions.
That someone in your congregation is already fluent
There is a twenty-six-year-old in your pews using this six hours a day at work. There is a nurse in your congregation whose hospital has already run the whole governance conversation you are dreading. There is probably a schoolteacher living through the most intense version of this debate happening anywhere.
Churches consistently reach for an outside consultant while ignoring three people in the fourth row who know more than the consultant does about how this actually goes in an institution.
Shrewdness is noticing what you already have.
What a shrewd church does
Writes down its lines before anyone needs them
Not a policy document. Half a page, in your own words, that a volunteer could read and follow.
What never goes into these tools: anything a person told you in confidence, anything about a minor, anything from a pastoral conversation, anything from a counseling session, giving records tied to names, personnel matters.
What always has a human being on it: anything sent under the pastor's name, anything going to a family in grief, anything that will be read from the front, anything involving a child.
The reason to do this before rather than after is not legal. It is that your staff are currently deciding it individually, in private, at nine at night, on their own accounts, with the best intentions. They are pasting in the hardest things to write, which are exactly the things involving actual people, because those are the tasks where help is most wanted. Nobody is being unreasonable. But information about people in your care is leaving your church, nobody chose that, and it cannot be recalled.
Picks a real sandbox and staffs it
A bounded piece of real work, with real stakes, where being wrong is recoverable and someone is paying attention.
Concretely, the good candidates in a church are the ones with volume and low pastoral risk. The newsletter. The bulletin. Event descriptions. Grant applications. Meeting minutes. First drafts of routine correspondence. The website copy nobody has had time to rewrite since 2019. Making twenty years of sermons searchable.
The bad candidates are the ones people reach for first: anything pastoral, anything for a family in crisis, anything that involves speaking to a person about their own life.
What makes it a sandbox rather than just usage is that somebody is watching and writing down what happens. Which tasks got better. Which got worse. Where somebody caught something wrong. Where somebody nearly did not.
Forms people at a scale that feels excessive
This is the part I would defend hardest, because I have three years of evidence and it all points the same way.
I have shown this technology to hundreds of people, most of them free, most of them people who trusted me and wanted it to work. They were astonished almost every time. Almost none of them adopted anything.
A close friend, for whom I spent five days learning his profession so the demonstration would sit inside his actual work rather than next to it, was impressed and did not touch it again for two years.
Demonstration does not form anybody. You already know this. It is why you do not disciple people with a really good sermon. It is why your best volunteers were formed over years by somebody who stayed with them. Nothing about this technology suspends the way human beings actually learn.
So the honest number of hours is much larger than anyone wants it to be, it has to be with people rather than a video, and it has to be on their own real work rather than on examples. The engagement I am in the middle of right now is a hundred hours across every team in the organization. Not a pilot group. Not the technically inclined. Everyone. [The organization is named, with numbers, once they have agreed to it and the work is finished. Until then it stays anonymous, because a progress report is not a case study.]
A church can do a smaller version. But if the plan is a two-hour workshop, the plan is to have had a workshop.
Puts its own house in order first
Before any tool, ask who knows what around here, and what happens when they leave.
Most of the fragmentation in a church is not technological. It is that the person who knows why the Hendersons give is not the person writing the appeal. The person who was on the mission trip is not the person telling the story. The person who remembers the 2011 attempt is retired.
A tool does not fix that. People deciding to write things down fixes that, and then a tool makes it reachable. In the wrong order you get an expensive system that indexes nothing.
Talks to the people who already know
The nurse. The teacher. The twenty-six-year-old. Ask them what they have watched go well and badly, and what they wish their institution had decided earlier.
You will get better counsel in ninety minutes than most churches get from a consultant, and you will have done something more important than gathering information, which is to signal that this is a congregational conversation rather than an office decision.
Says out loud what it is doing
If your newsletter is drafted with AI assistance and edited by a person, say so once, plainly, somewhere findable.
There is a version of this that is fear management and a version that is shrewdness. The shrewd version understands that trust is the only asset a church has that cannot be rebuilt quickly, and that the way you lose it is not by using a tool. It is by having used it quietly and being found out. Disclosure is cheap now and priceless later.
What a shrewd church refuses
I want to be careful not to present refusal as the safety half of this. Knowing what a tool is for includes knowing what it is not for, and getting that wrong is a failure of competence, not just of nerve.
Anything that is constituted by presence. A hospital visit. Sitting with someone after a diagnosis. The conversation in the parking lot after the meeting. These are not tasks with outputs. The presence is the substance, and a more efficient version of it is not a version of it.
Anything a person said to you in confidence. No exceptions and no clever workarounds. Not summarized, not anonymized, not "just to help me think it through." The moment a congregant's disclosure becomes input, you have changed what confession means in your church, and you will not be able to change it back.
Discernment. A machine can lay out considerations, and that is useful. It cannot weigh them, because weighing requires knowing what this congregation has already been through and what it can bear. If a decision requires knowing your people, the tool is not a participant.
Prayer. I am not going to argue this one.†
Anything where the person on the other end thinks they are getting you. A note to a grieving family. A response to someone in crisis. If they believe a person attended to them and no person did, you have not saved time. You have taken something from them without telling them.
And anything involving a child, until you have thought about it much harder than you want to. Photos, stories, safety records, anything a minor said. Churches have learned expensive lessons about child safety in the last thirty years and every one of those lessons applies here in a new form.
Notice that these refusals are not arbitrary and they are not about risk tolerance. They mark the places where the work is constituted by a human being doing it. That is a knowable line, and knowing where it falls is precisely the competence I am arguing for.
The first ninety days, concretely
If you did nothing else this year, this would be enough, and it is achievable for a church of any size.
Weeks one and two. Three or four leaders, separately, answer the same three questions on paper. What should this never touch here. Where would it obviously help. What are we afraid of. Then compare. The disagreements are the whole exercise, and there will be more of them than you expect.
Weeks three to six. Draft half a page of lines. Read it aloud to your staff and let them argue with it. Revise. The read-aloud is not a formality; anything that cannot be read aloud without qualification is not written clearly enough to be followed.
Weeks seven to ten. Pick two tasks with volume and low pastoral risk. Give them to two people. Ask them to keep a running note of what actually happened, including the near misses.
Weeks eleven and twelve. Sit with the notes. Decide what graduates into normal practice, what stops, and what still needs watching. Then talk to your congregation about what you have decided and why, before anyone asks.
Ninety days, no budget, no vendor. And at the end you will have something almost no church has, which is a decision.
The part that cuts against me
I said at the top that I sell this. Here is the part that costs me something to say.
Most of what I have listed above does not require me or anyone like me. The half page, the sandbox, the conversation with the nurse and the teacher, the ninety days. a church can do all of it alone, this quarter, for nothing. If you do only that, you will be ahead of nearly every congregation I have talked to in three years.
The place outside help earns its keep is later, when you want to make a large body of material actually reachable, which is a real technical problem that does not solve itself. Not at the start.
Why this is shrewdness and not caution
Read back over the list and notice what is not in it. There is nothing about protecting yourself, nothing about liability, nothing about waiting to see how it shakes out.
Every item is about competence. Knowing what the tool is good at. Knowing what you already have. Knowing who in the room already knows. Knowing where the line falls and why it falls there. Building the capacity to judge before the moment when judgment is needed, because that capacity cannot be borrowed at midnight.
That is what practical wisdom has always consisted of, in Proverbs and in Luke and in every ordinary decision your elders have ever made well. It is the intelligence of a person who has to act, in a situation that will not wait, with responsibility for people who did not choose the situation either.
The children of this age are shrewder in dealing with their own generation. They are, at the moment, and it is not because they are smarter than you. It is because they decided it was their business to be competent about this, and we decided it was somebody else's business.
It is not somebody else's business. There is a twenty-six-year-old in your pews who has already formed her habits, and nobody in your church has said one word to her about it.
† No source. This is a judgment, not a fact, and it is the one line in this piece I have declined to argue for. I think a machine praying on your behalf is a category error rather than a risk, and I am aware that is a theological position and not a finding.
Joshua Shepherd builds AI systems for organizations whose work depends on trust, and trains the people who use them. [Contact or newsletter line.]