The AI question has reached every believing boardroom and most believing shop counters, and it usually arrives unstructured: staff already using chatbots nobody sanctioned, a vendor pitching automation nobody evaluated, a competitor moving faster and a conscience unsure what faithfulness requires. The unhelpful responses are the reflexes: prohibition (which drives use underground), enthusiasm (which adopts whatever ships), and the values statement without teeth this series has already buried. The helpful response is a policy: one page, three lists, ban, require, audit, adoptable by a five-person firm on Monday and expandable as the tools evolve. This essay writes that artifact for the faith-driven company, with the reasoning attached, because a policy whose logic staff understand gets kept, and one absorbed as arbitrary rules gets routed around. The governance pillars, truth, judgment, service, accountability, order the whole page.
Key Takeaways
- Companies need an AI policy before they think they do: staff adoption precedes management decision in most firms, so the real choice is governed use versus ungoverned use.
- The BAN list draws bright lines: impersonating humans without disclosure, fabricating evidence of any kind, delegating firing and discipline to machines, manipulative persuasion, and spiritual counterfeit.
- The REQUIRE list builds the floor: disclosure when AI touches customers, human review of consequential outputs, verification of AI claims before they ship, data boundaries for what enters public tools, and skill-building rather than silent replacement.
- The AUDIT list keeps both honest: a quarterly use inventory, error and harm review, disclosure spot-checks, and one standing question about who was quietly displaced.
- The policy’s theology is practical: truth-telling (no fabrication, full disclosure), judgment kept human (consequential calls), service (AI toward customers’ good), accountability (audits with names).
- One page, read aloud, revised annually: the policy is a discipleship document as much as a compliance one, teaching a company what its confession means at the keyboard.
What belongs on the BAN list, and why?
Five bright lines, each traceable to a command older than the technology.
No undisclosed impersonation of humans. AI may draft, but customers and staff always know when they are talking to a machine: the chatbot is labeled, the AI-written message is signed as assisted where it matters. The line is truth-telling in its simplest commercial form, the disclosure posture as company rule, and it forbids the small daily counterfeits, fake “personal” outreach, synthetic testimonials, invented authorship, that current tools make effortless.
No fabricated evidence, ever. No AI-generated reviews, credentials, case studies, imagery of events that did not happen, or citations that do not exist, in any document, internal or external. Fabrication is the ninth commandment with a GPU, the deepfake essays’ line drawn through ordinary marketing.
No machine-decided discipline or dismissal. AI may summarize and surface; it may never decide a human being’s employment, discipline, or accusation. Firing is shepherding, the most judgment-laden act a company performs, and delegating it to a system is abandonment dressed as efficiency.
No manipulative persuasion engineering. No AI-optimized exploitation of fear, compulsion, or cognitive weakness: dark-pattern generation, addiction-tuned messaging, the sales-motion line enforced at the tool level.
No spiritual counterfeit. No AI voicing God, generating “prophecy,” or impersonating spiritual authority, in products or in the company’s own devotional life. Machines may serve the faith’s work; they may not speak in its Name.
What belongs on the REQUIRE list?
Five floors that make ordinary use trustworthy.
Disclosure where AI touches people. Customer-facing AI is labeled; significant AI assistance in deliverables is acknowledged; the company’s disclosure norm is written, so individuals are not improvising ethics alone.
Human review of consequential outputs. Anything that commits money, promises customers, touches safety, or represents the company publicly passes human judgment before it ships, the escalation principle as workflow: AI drafts, people decide.
Verification before publication. AI-supplied facts, figures, and citations are checked against sources before use, and the checker is named. The tools hallucinate; a truth-telling company budgets the verification time, the same source-discipline this corpus applies to itself.
Data boundaries. A written list of what never enters public AI tools: customer personal data, staff records, unpublished financials, everything the crown-jewels essay fences, with sanctioned alternatives provided so the boundary is keepable.
Skill-building over silent replacement. When AI absorbs tasks, the company invests in transitioning the person, new skills, redesigned roles, honest conversations, the people-development posture applied to automation: efficiency is welcome, quiet discarding of people is not.
What does the AUDIT list check, and who checks it?
Quarterly, one hour, inside the review cadence the firm already runs, four items with a named owner. The inventory: what AI tools are actually in use, by whom, for what, surfacing the shadow adoption every firm has. The harm review: errors, complaints, near-misses involving AI outputs, logged and examined without blame theater. The disclosure spot-check: sample customer touchpoints and deliverables against the labeling rules. And the displacement question, asked out loud: whose work changed this quarter because of these tools, and did we handle it as shepherds or as spreadsheets? Findings feed the annual policy revision, because the tools will not stop moving and a policy last touched in the year of its writing is already folklore.
The audit is where the policy becomes accountability rather than aspiration, and its deepest function is formative. A company that reads its one page aloud at onboarding, keeps its bans visibly even when costly, and audits itself quarterly is teaching its people, more durably than any retreat, that the confession governs the keyboard too. That is the policy’s real product: not compliance, but a workforce that has watched its employer mean it. Write the page this week. Read it aloud on Monday. And let the machines meet, in your company, the older Word they will be serving.
FAQ
Why does a small firm need an AI policy?
Because staff adoption precedes management decision: chatbots and tools are already in use ungoverned. The choice is not whether AI is used but whether its use has lines, floors, and checks.
What are the five bans?
Undisclosed impersonation of humans, fabricated evidence of any kind, machine-decided discipline or dismissal, manipulative persuasion engineering, and spiritual counterfeit (AI voicing God or spiritual authority).
What are the five requirements?
Disclosure wherever AI touches people, human review of consequential outputs, verification of AI claims before publication, written data boundaries with sanctioned alternatives, and skill-building for anyone whose tasks AI absorbs.
What does the quarterly audit cover?
A use inventory (including shadow adoption), an error and harm review, disclosure spot-checks, and the displacement question: whose work changed, and was it handled pastorally? Findings drive the annual revision.
Is this policy anti-AI?
No: it is pro-use with integrity. The bans forbid counterfeits, not capability; the requirements make everyday use trustworthy; the audits keep the whole arrangement honest as tools evolve.
Related Reading
- A Governance Model for AI in the Kingdom Economy
- Redemptive AI Requires Product Decisions, Not Press Releases
- Chatbot Jesus: Why AI Impersonation of Christ Crosses a Line
- Hiring, Firing, and Shepherding as a Christian Founder
Sources and Evidence
- Praxis Journal, “A Redemptive Thesis for Artificial Intelligence”: the redemptive frame applied to AI adoption.
- Exodus 20:16, ESV: the false-witness command underlying the fabrication and impersonation bans. See also Proverbs 11:1.
- Barna Group, “What Faith Looks Like in the Workplace”: the workplace-integration context the policy serves.