How we use AI

We use the tools we sell. On ourselves, first.

We do not ask organizations to do anything we have not done. Every guardrail we ask a leader or a board to ratify, we have built into our own work first. This page is where we name what we refuse, how we sort use cases, and what we will not build, so anyone reading the site knows where we stand.

Most public AI discourse splits into “AI is a savior” or “AI is a threat.” We reject both. AI surfaces and amplifies what humans have already produced, and how a person reacts to it often reveals their underlying view of human nature. Our work starts from a simpler operational rule: if AI mirrors the humans it is around, govern the humans, the corpus, and the relationships first.

Our commitments

What we will not do, ever.

These are not preferences. They are the line. Breaking one would cost the exact trust we exist to protect.

No manufactured urgency or scarcity. We will never manufacture a deadline or a fear of missing out to move you.
No faked authorship. We help your people tell their own story. We never write under a human's name as if it were them.
AI gathers and drafts. Humans discern, care, and speak. AI does the gathering and the drafting. It never does the discernment, the shepherding, the relationships, or the care.
We claim only what is earned. We do not promise outcomes the work has not produced yet. If a number appears on this site, it has a footnote.
Extra care wherever vulnerable people are involved. When you serve minors or families, we are more cautious, not less.

What we refuse

Three lines we do not cross.

These refusals are essential, not ethics bolted onto a product after the fact. They shape every line of code we ship and every document we help produce.

Replacement of relationship

We do not use AI to simulate pastoral presence, mentoring, or care. Formation happens in relationship. AI can support the work around that relationship; it cannot be the relationship.

Formation without presence

We do not hand over the work that forms us. Preaching, teaching, and writing that carries our convictions go to the machine only with our full engagement. If it goes out under our name and is meant to form people, we have read it, weighed it, and owned it.

Amplification without credibility

We do not use AI to look bigger than the work and network actually warrant. Not more output, more polish, or more reach when we have not done the work or do not have the people to back it up. Looking more credible than you are fails when it outruns the credibility you actually have.

Green / yellow / red

How we sort use cases.

Not all AI use is equivalent. This is the same framework we use in Safety and Sandbox engagements, and how we evaluate our own use internally.

Green light

Clear human benefit, no meaningful negative consequence.

Structure, findability, admin, and translation. These are tasks that do not require your voice or your judgment. First drafts you will edit and own before anything goes out under your name. AI handles the supporting work; you stay in the loop where the output represents you.

Making existing work searchable
Metadata, headings, internal linking
Translating work you have already written
Formatting and administrative repetition

Yellow light

Real benefit, but only with guardrails.

Context-dependent. A quick internal summary is not the same as a public-facing piece. A first draft you will heavily edit is not the same as a post you publish with light changes. When in doubt, slow down. When the output will form people or represent you publicly, the human in the loop is you, reading, weighing, owning.

Drafts you review before sending
Research where citation and verification are required
Facilitated conversations where AI never substitutes for participants

Red light

Harm, regardless of how carefully it is deployed.

The lines we do not cross, and the lines we help organizations name in Safety. AI used to author content under a human's name without disclosure. AI impersonating humans in pastoral or counseling contexts. AI making eligibility decisions that affect people's access to services. We do not deploy these. We help the organizations we work with publicly commit to refusing them as well.

Undisclosed authorship under a human name
Pastoral or counseling impersonation
Decisions about access to care or services

You stay in charge

AI supports your credibility, it does not replace you.

The sustainable pattern: you handle creation and authority, what you teach, what you write, what you put your name on. AI handles discoverability and circulation: structure, metadata, translation, formatting, linking. Your work becomes easier to find and connect without the machine becoming you.

AI is trained on the broad output of human writing, speech, and behavior. When you interact with it, it mirrors and amplifies what it has been trained on, and it mirrors and amplifies you specifically. That is why the most important operational decision is not which model you pick. It is who you deliberately gather around the work: the humans whose corpus, judgment, and vouching the system will reflect back.

AI can help your work be legible and connected. It never becomes you. Only people can vouch for you. Only you can mean what you say before God and stand answerable for it as true.

You out front. AI in the background.

The scenius refusals

What we will not build into the network.

The extractive version of this business is obvious, and we are building against it. That includes product choices, not only policy statements.

No ranked leaderboards
No vanity metrics
No engagement-optimized feeds
No parasocial substitutes where reading replaces relationship

If a feature would feel extractive to us as members, we don't ship it.

Ask your AI about this

Tap a suggestion to open a page. Type here to talk it through.