Strategy · July 30, 2026 · 11 min

Move aside, AI Policy. Have You Considered an AI Covenant?

One is a list of rules nobody reads. The other is a conversation your team actually has — and it changes how AI gets used before you write a single guideline.

Sarah Romorini · Governance, without the legalese

Sigh. An AI policy is usually a list of rules in legalese that nobody reads. An AI covenant is a conversation your team actually has — and it changes how AI gets used before you write a single guideline.

Small teams adopt AI the way they adopt everything else: quietly, in the cracks of the day. Someone uses it for an estimate. Someone else drafts a client email. A third person tries it on social and never mentions it. There is no shared agreement. Default silence becomes a confusing policy later.

A covenant is not anti-AI. It’s the opposite. It’s a one-page, plain-language understanding of how you use intelligence here — built around four questions your team can actually answer.

Four questions, one page

Your why. Specific reasons — faster client answers, more human judgment on the work that matters, fewer nights spent rewriting the same paragraph.

What you use AI for. Honest current tasks: drafts, summaries, checklists, first-pass research. Not a fantasy roadmap.

What you protect. Strategy, approvals, client secrets, anything under NDA. The human-first list, written down.

How you handle disagreement. Early discussion, good intent, updates when the work changes. A norm, not a tribunal.

Try this now

Take twenty minutes with whoever is in the work. Answer the four questions out loud before anyone opens a document. Then ask an AI to draft a covenant under 300 words in a direct tone. Post it where people will actually see it — not in a policy binder.

The goal is not restriction. It’s alignment: this is how we use AI here. This is why.

Keep the Moonbird Intel newsletter coming