For church leaders. Revision 1. ~2,700 words. A cluster spoke of There is no such thing as AI safety.
There is a line in Luke that I have not been able to get out of my head for about two years now.
Jesus tells a story about a manager who is about to be fired. Before the books close, the man goes to everyone who owes his master money and reduces what they owe, so that when he is out of a job there will be people willing to take him in. And then the master commends him. Not for his honesty. Luke goes further than withholding that word and calls him dishonest outright. He is commended for his shrewdness.
Then Jesus says the thing. The children of this age are shrewder in dealing with their own generation than the children of light. Luke 16, verses 1 to 8.
I have been in and around ministry for most of my adult life, and I have spent the last three and a half years watching what churches and Christian organizations do when a new technology arrives. I do not think that sentence has ever been more plainly true than it is right now.
Before I go any further you should know that I build these systems for a living and I sell exactly the kind of training this piece argues you need. That is a real conflict and it runs through every paragraph below. I say it here rather than at the end because an argument like this one is easy to abuse, and you should be able to weigh it from the beginning rather than discovering the interest after you have been persuaded. I say more about it further down, including the parts that cut against me.
What I have actually watched
Since December 2022 I have shown this technology to church leaders, denominational staff, nonprofit teams, seminary people, and about every kind of ministry organization there is. Most of it for free, most of it alongside a full time job.
The pattern almost never varies. People are impressed. And then nothing happens.
Sometimes nothing happens because the leader has decided this is a threat, and that waiting is the faithful posture. Sometimes nothing happens because the leader is enthusiastic but has no idea what the next step would be. And sometimes something does happen, but it happens sideways: nobody decides anything at the top, so the youth director and the communications person and the executive assistant each quietly figure it out on their own accounts, at night, using it for the tasks that are hardest to write, which are exactly the tasks that involve real people.
Two failures, opposite in appearance. One organization is paralyzed. The other is unsupervised. Neither has deliberated.
What strikes me is that Scripture has a word for each of them, and it is not the word most of us reach for.
Both halves of the command
When Jesus sends the twelve out, he tells them they are sheep among wolves, and then gives them a two-part instruction that most of us can quote and few of us hold together.
Be shrewd as serpents. Be unmixed as doves. Matthew 10, verse 16.
Start with the first half, because it should be more shocking than it is. The serpent is not a neutral image in the Hebrew Scriptures. The word used of the serpent in Genesis 3 is the same word Proverbs uses over and over for the prudent person, the one who sees trouble coming and takes cover while the naive walk straight into it. One word, condemned in Eden and commended in Proverbs.
The tradition never resolved that by getting rid of the faculty. It resolved it by locating the moral weight in what the craft is aimed at. Cunning aimed at deceiving is serpentine. The identical cunning aimed at protecting the people in your care is wisdom.
So when Jesus says be shrewd as serpents, he is reaching for the most negatively charged picture of cleverness available in his own Scriptures and telling his people to match it.
The second half is where most of us have quietly changed the meaning. The Greek word behind "innocent" does not mean naive. It means unmixed, unalloyed, the way metal is described when nothing has been added to it. It is a word about purity of composition, not about being uninformed.
Which means the command is not be clever but also a little bit stupid. It is: be maximally competent in your means and completely unadulterated in your motives.
And notice where it is said. Sheep among wolves. This is not general life counsel. It is what you tell people walking into danger. The reason for shrewdness is that the situation is dangerous. Naivety in a dangerous situation is not innocence. It is negligence that has learned to describe itself as innocence.
Both halves are commanded. Shrewd without unmixed is the serpent in the garden. Unmixed without shrewd is a sheep in a wolf pack. We have spent a long time treating the second one as the safe error. It is not. It is just the more respectable one.
The buried talent
The parable that will not leave the fearful alone is the one about the talents, and specifically the third servant. Matthew 25, verses 14 to 30.
He does not steal. He does not squander. He buries what he was given and returns it intact, and he gives his reason plainly: he was afraid. He knew the master to be a hard man, so he protected what he had.
The verdict is severe, and it is worth sitting with why.
First, his fear is named as the cause, and it is not accepted as a defense. It is treated as the offense itself.
Second, and this is the part I find most convicting, the fear had produced a theology to justify itself. "I knew you to be a hard man." He had an account of the master that made his paralysis reasonable. That is almost always how this works. Inaction rarely arrives announcing itself as cowardice. It arrives as a considered position about the nature of the risk, and it is usually articulate.
Third, the master's response closes the last exit. At minimum, he says, you could have put it with the bankers. Even the most cautious available option involved handing it to somebody else and letting it work. Burial was not the low-risk choice. It was the only guaranteed loss.
I want to be careful here, because this parable has been used to bless a great deal of ordinary ambition by dressing it in Scripture, and church leaders are right to be wary of that. I am not saying God will judge your technology decisions. I am saying the parable makes a specific claim about the structure of fear, and that claim is portable: doing nothing is not the neutral position, and sincerity about your reasons does not make them true.
If your organization has decided to wait, the question worth asking is not whether the decision is defensible. It probably is. The question is whether it is a decision at all, or a fear that has found language.
The oil that could not be lent
Then there are the ten young women waiting for a wedding party that is late. Matthew 25, verses 1 to 13.
Five are called wise and five foolish, and the difference between them is one thing. The wise brought extra oil against a delay they could not predict. It is worth noticing that all ten fall asleep. The wise are not more vigilant. They are better provisioned.
The line that lands hardest comes at midnight, when the foolish ask to borrow. The answer is no, and it is not cruelty. There will not be enough for both. Go and buy your own.
That refusal is a statement about what kind of thing oil is. Some capacities cannot be transferred at the moment of crisis, and cannot be acquired at the moment they are needed.
This is the part of the current conversation that I think church leaders most consistently get wrong, and I say that having watched it up close for three years.
The assumption is that when this technology matters enough, the organization will learn it then. There will be a workshop. Somebody will figure it out. We will catch up when catching up becomes necessary.
Formation does not work that way. It never has, in any domain your church already understands. You cannot disciple someone at the moment they face the crisis. You cannot build a culture of discernment during the week you need discernment. The reason your church invests years in people before anything visible happens is that you already know this about every other kind of formation. It is equally true here.
The capacity to use a powerful tool wisely is built slowly, in low-stakes conditions, before the stakes arrive. When it arrives first and the capacity is missing, you do not get a slower version of the same outcome. You get a capable instrument in the hands of an organization with no formed judgment about it, which is worse than having nothing at all, because now the mistakes are fast and confident and they carry your name.
Reading the moment
There is a rebuke in Luke 12, verses 54 to 57, that I find harder than the others because it is aimed at competent people.
Jesus tells the crowd they can look at the sky and tell you what the weather is doing, and then asks why they cannot interpret the present moment. And then, sharper: why do you not judge for yourselves what is right.
The charge is not ignorance. These are people who read their world expertly in ordinary matters. The charge is a selective competence, sharp about everything except the thing that carries the most consequence.
And the second sentence is the one I would put in front of every leader I know. Judge for yourselves. Not wait for a denominational statement. Not wait until a consensus forms among your peers. Not wait for someone with more technical credibility to tell you what moment this is. The waiting for permission is itself the failure being named.
Count the cost
I want to put one guardrail on all of this, because everything I have written so far pushes in one direction and the Scriptures do not leave it there.
Jesus also tells two short stories about estimation, in Luke 14, verses 28 to 32. A man begins a tower, runs out of money, and cannot finish it, and everyone who walks past mocks him. A king considers whether ten thousand can meet twenty thousand.
Notice what these are not. They are not arguments against building. Both assume the project. The failure is beginning without honest estimation, and the specific damage is public humiliation in front of everyone who watched you start.
I have been on the wrong side of that one. In 2025 I told a leadership team a build would take three months. I had spent two years training them to expect things much faster than that, and I was confident. Then I hit a part of the work I did not understand, and a week in I gave a demo that did not work. I diagnosed the problem quickly and I delivered the whole thing in about six weeks, which was well inside what I had promised. It did not matter. The room had decided in week one, and I never got it back. [ONE CHECKABLE DETAIL: what the build was, or what it did once it shipped. The story is currently the only claim in this piece a reader has to take on trust, and it is the one about my own competence.]
The lesson was not that I moved too early. It was that I announced a tower I could not yet see the top of, in front of people who were watching. Shrewdness includes knowing the difference.
So: build. But size it to be finished, and say out loud what it will actually cost.
What I am not saying
Three things, plainly, because this kind of argument is easy to abuse and you should be able to see me refusing to.
I am not saying the moment Jesus tells you to read is this technology. It is not. In Luke 12 it is his own presence, and reading a market window into that is a category error. What transfers is a claim about how fear and competence and timing relate to each other in decisions people actually have to make. That structure is portable. The urgency of the kingdom is not on loan to anybody's product.
I am not saying that faithfulness requires adopting this. It requires deciding about it. Those are different, and a church that deliberates carefully and concludes it will use very little of this has done the thing I am asking for. What I am against is drift, in either direction.
And on the interest I declared at the top, one thing that cuts against me. Almost everything in the sequence below can be done by a church alone, this quarter, for nothing. The half page of lines, the conversation with the people in your congregation who already know, the bounded piece of real work. If you did only that you would be ahead of nearly every congregation I have talked to in three years, and you would not have paid me anything. I have written the whole practice out in detail, free, for the same reason. Outside help earns its keep later, when a large body of material has to be made reachable, which is a real technical problem that does not solve itself. Not at the start.
What shrewdness actually looks like this year
If I had to reduce all of it to a sequence, it would be this, and the order matters more than any single part.
Decide what this is allowed to touch, before anyone uses it for anything that matters. In your own words, in language your staff would actually use, written down. Not a policy nobody reads. A short set of lines about what does not go into these tools, particularly anything about the vulnerable people in your care, and what always requires a human being.
This is the innocence half, and there is real urgency in it. Every week you do not decide, your staff are deciding individually, in private, with the best intentions. Nobody is being unreasonable. They are trying to do their jobs and a tool showed up that helps. But information about real people is leaving your organization, nobody chose that, and there is no way to walk it back.
Then make somewhere for people to actually learn. A bounded place, real work, real stakes, where being wrong is survivable and someone is paying attention to what happens. This is the oil. Buy it before midnight.
Then form people, at a scale that feels excessive. Not a workshop. Enough hours that a practice forms. I have watched three years of astonishing demonstrations produce almost no change in behavior, in people I know well, who wanted it to work. Demonstration does not form anybody. It never has, in any part of your ministry.
Then, and only then, build. By that point you will know what to build, because the people who will use it will have been working with it for months and can tell you.
Everyone wants to start at the end. The tools are the interesting part and the part with a price attached. Starting there is the tower with no foundation.
That sequence is the argument of this piece compressed into four moves. The long version, with the actual content of each, is in the companion piece.
The sentence I cannot get past
The children of this age are shrewder in dealing with their own generation than the children of light.
That is not a compliment to the children of this age. It is a rebuke to us, and it is observably accurate right now. The commercial world is deliberating about this with enormous seriousness. Meanwhile the institutions whose entire work runs on trust, which is the thing this technology most directly threatens, are largely doing neither. Not deciding what is permitted. Not building anything. Not shrewd, and not particularly innocent either. Just late.
We have a command with two halves. We have spent a long time treating one of them as optional and calling that faithfulness.
It is not the safe error. It is just the one that looks better while it costs you.
Scripture references were checked against the text for this revision. The reading of the Greek and Hebrew terms is mine and is contestable, and any pastor with the languages should feel free to tell me I have it wrong.
Joshua Shepherd builds AI systems for organizations whose work depends on trust, and trains the people who use them. [Contact or newsletter line.]