Copy in the brand voice — not the average of the internet

An unconfigured model writes toward the statistical middle: polished, agreeable, abstract, and familiar. Asking it to “sound like us” does not provide enough evidence. Brand voice becomes operational when examples, exclusions, audience choices, factual rules, and review criteria are encoded as a skill that the team can test.

Build a brand kit the role can execute

Collect approved examples across formats, but annotate why each works. Define sentence rhythm, vocabulary, point of view, level of technical detail, claims policy, humor boundary, and preferred calls to action. Add negative examples: empty superlatives, inflated certainty, clichés, fake customer intimacy, and phrases the company never uses.

Separate durable voice from campaign context. Voice may remain direct and evidence-led while audience, offer, channel, and length change. Give the role source material for every factual claim. If a fact is absent, it should ask or leave a marked placeholder rather than improvise.

Encode voice in five to seven days

Day one gathers examples and reviewer disagreements. Days two and three produce the first skill and rubric. Days four and five run a test set spanning landing copy, email, product update, social draft, and long-form explanation. Days six and seven resolve failure patterns and establish an owner.

This timeline encodes a usable first version; it does not justify auto-posting. The marketing role drafts into a sandbox and links its sources. A reviewer checks factual accuracy, fit, legal constraints, and taste. Publishing and campaign-system writes default to Ask.

Measure consistency without flattening creativity

Use rubric dimensions such as recognizability, clarity, unsupported claims, prohibited language, audience fit, and edit distance. Have multiple reviewers score some blind samples to expose disagreement. Track recurring edits and commit useful corrections back to the skill. Do not optimize a single “brand score” until every sentence becomes mechanical.

The role should create more good starting points and make standards reusable across the team. Humans still choose the idea, take responsibility for public claims, and decide when an intentional break from the house style is the right creative move.

Run it as a role, not a prompt

The durable implementation is a role inside the company operating system. Its instructions, checklists, examples, and connector definitions live in git, so every change has an author, review, and rollback path. Each run gets a sandbox and an auditable record. The team can use the best model for each step instead of tying the workflow to one vendor. Models can change; the role, tests, permissions, and history remain.

Connections are brokered server-side. CRM, document, mail, and accounting credentials never sit in a prompt or a browser extension. Read access is scoped to the records needed for the task. Writes default to Ask: the agent prepares the proposed update, message, or file, then an accountable person approves it. Self-hosting is available when policy, residency, or network boundaries require it.

Make the first pilot measurable

Choose one queue with enough volume to observe within two weeks. Record the current cycle time, rework rate, backlog, and escalation rate before the first run. Test historical cases, including awkward and incomplete ones, before touching live work. During the pilot, compare accepted outputs, corrected outputs, false escalations, and time returned to the team. A useful role becomes more reliable because corrections are committed back to its skill and evaluation set.

Keep the boundary explicit. The agent can collect evidence, apply a checklist, draft, route, and update systems after approval. A named employee owns exceptions and consequences. That division is what turns model capability into dependable operations without pretending that probability is judgment.

This work runs on Neuro OS. To scope a first role, get started.

Explore Neuro OS →