Workflow design — the operator's force multiplier.

Module 1 gave you the load-bearing theory — archetypes, positioning, equity, voice/tone, semiotics. This module gives you the load-bearing workflows: the repeatable processes a brand assistant runs every week. We describe each workflow tool-agnostically, then in §2.4 we map the optional toolchains that can speed each step up.

By the end of Module 2 you should be able to:

1. Run a voice extraction on a brand's content corpus in five clearly-defined steps.
2. Conduct a brand-check audit against a rubric you can defend to a senior stakeholder.
3. Organize brand assets so a teammate can find the current canonical version in under sixty seconds.
4. Choose a toolchain (pen-and-paper, free tools, general LLM, Adytum apps, or hybrid) by matching the workflow stage to the tool's actual strengths — not by defaulting to whatever's familiar.

2.0 Why workflow discipline matters more than tool choice

Most brand-team failures are not tool failures. They are process discipline failures: missing steps in a familiar workflow, inconsistent ownership of artifacts, no clear definition-of-done for routine tasks. When a workflow is sound, swapping tools is a productivity adjustment. When the workflow itself is broken, no tool can rescue it.

This module rests on three observations from operations research and process design:

1. Variance, not average, kills throughput. Eliyahu Goldratt's Theory of Constraints (The Goal, 1984) and decades of subsequent operations research show that the cycle time of a process is dominated by its variance, not its average. A brand workflow that usually takes two hours but occasionally takes two days will frustrate stakeholders and erode trust faster than one that reliably takes three.
2. Atomic steps survive interruption. A workflow broken into atomic tasks — each one independently completable in a single sitting — can be resumed after interruption without state loss. Workflows specified as "do brand audit" cannot.
3. Clear ownership beats clever coordination. The single most-cited reason for process breakdown in brand teams is unclear ownership of the next step. Decision-making frameworks like DACI (Driver, Approver, Contributor, Informed) or RACI exist to make ownership unambiguous at every workflow stage.

The three workflows in this module are designed to satisfy all three principles: low variance, atomic steps, unambiguous ownership.

2.1 Voice extraction — the five-step workflow

Voice extraction is the foundational brand-assistant workflow: turn a corpus of existing brand content into a one-page voice document that any writer on the team can use. The output is portable across tools; the process is what matters.

Step 1 — Sample sourcing (30-60 minutes)

Gather 20-50 samples of the brand's existing content. Diversity across channels matters more than total volume — 30 samples spanning email, website, social, customer-service replies, and exec interviews will out-perform 100 samples drawn from a single blog. Aim for representation across:

• Channel — owned site, email, social platforms (each), customer service, product copy, exec communications
• Tone-context — celebratory, transactional, apologetic, instructional, persuasive, technical
• Time — recent (last 90 days) + older (12+ months) — drift detection requires temporal diversity

Store the samples in one location: a single document, a single spreadsheet, or a single folder. The single source of truth principle starts at the earliest workflow step — if samples are scattered, the extraction is already at risk.

Step 2 — Pattern identification (1-2 hours)

Read the samples twice. The first read is for impression — do not take notes. The second read is for pattern capture at three levels:

• Vocabulary — which words appear repeatedly across multiple samples? Which words you'd expect are conspicuously absent? Which sound off-brand even when grammatically correct? Maintain a two-column list: preferred / avoided.
• Syntax — what's the sentence-length distribution? Are paragraphs short (1-3 sentences) or long (5+)? Does the brand favor declarative or interrogative? Lists or prose? First person ("we") or second person ("you") or neither?
• Value-judgments — what does the brand treat as good (worth celebrating) vs bad (worth opposing)? What's important to it vs trivial? These judgments leak through word choice (e.g., a brand that calls competitors "alternatives" vs one that calls them "the old way" reveals its archetypal stance).

This step is cognitively demanding. Junior brand assistants often try to skip it by jumping straight to archetype labeling. Resist — the labels only have meaning if they emerge from the evidence.

Step 3 — Archetype matching (15-30 minutes)

Map the identified pattern against the eight voice archetypes from Module 1 §1.1.3. Pick the primary archetype that best fits the pattern. If a secondary archetype clearly contributes (e.g., primarily Calm Authority with a Warm Specialist undercurrent), name it. If no secondary emerges cleanly, do not invent one — over-specification is its own failure mode.

If no single archetype maps well, treat that as a finding rather than a flaw. A brand whose voice fragments across archetypes has a brand-strategy problem your voice extraction has surfaced. Escalate to a Practitioner-tier strategist; do not paper over the fragmentation with a synthetic label.

Step 4 — Voice document authoring (1-2 hours)

Write a one-page voice document with these sections:

1. Primary archetype (one sentence with rationale, two-sentence max)
2. Secondary archetype if applicable (one sentence)
3. Vocabulary — preferred words (10-15) and avoided words (10-15)
4. Syntax markers — typical sentence length, paragraph length, voice (active/passive), pronoun preferences
5. In-voice example sentences — three sentences that sound clearly like the brand
6. Out-of-voice example sentences — three sentences that sound clearly NOT like the brand, with a one-line note explaining why each fails

The contrast pair (in-voice vs out-of-voice examples) is the operational test for the document's clarity. If a new writer on day three cannot use the contrast pair to keep their first draft on-voice, the document is too abstract and needs rewriting.

Step 5 — Stakeholder review (30-60 minutes synchronous + iteration)

Show the voice document to people who work with the brand. Three categories of feedback indicate three different outcomes:

• "That's us" — the extraction succeeded. The voice document captures the brand the company actually is.
• "Interesting but not who we want to be" — the extraction picked up the brand's actual patterns, but those patterns are the brand the company has drifted into, not the brand it endorses. This is the most valuable feedback you can receive: it reveals the gap between strategy and execution.
• "That's not us at all" — either the sample was unrepresentative, or the pattern identification was wrong. Re-run with broader samples and a fresh reading.

2.2 Brand-check audit — the rubric-based workflow

Voice extraction produces a voice document. The brand-check audit is the recurring inspection that ensures live brand surfaces actually match that document. The workflow has four steps.

Step 1 — Establish the rubric (30 minutes, one-time per voice doc)

Pull from the voice document the five most-distinctive markers — the patterns most strongly diagnostic of voice fidelity. Examples:

• "Uses 'we' for inclusive framing, never 'you' for accusation"
• "Sentence length averages 14 words; no sentence exceeds 28"
• "Never uses superlatives ('best,' 'most innovative') without evidence in the same sentence"
• "Avoids consultant jargon ('synergy,' 'leverage,' 'unlock') across all surfaces"
• "Closes emails with action + next step, never 'thanks!' alone"

Five is the right number. Three is too few to surface meaningful drift; seven or more becomes a chore that auditors will short-circuit.

Step 2 — Audit page-by-page (5-15 minutes per page)

For each page, score against the five markers. Use a three-state scale:

• PASS — page hits the marker consistently
• PARTIAL — page hits the marker most of the time, with at least one specific violation noted with line reference
• FAIL — page contradicts the marker more often than it hits it

Record findings in a structured form (spreadsheet or markdown table). For each finding, capture: URL, marker, score, severity, line reference, and a one-sentence note describing the violation. The note matters more than the score — fixers cannot act on "FAIL" without specifics.

Step 3 — Severity-rank the findings (15-30 minutes)

Brand-check findings stack into four severity tiers:

• P0 — blocking. A high-traffic page (homepage, primary product page, paid-ad landing page) contradicts the voice. Customer first-impression is at risk; remediate within 48 hours.
• P1 — significant. A noticeable page (legal page, help-doc index, primary email template) drifts noticeably. Remediate within 2 weeks.
• P2 — minor. A lower-traffic page has one or two off-voice sentences. Batch into the next quarterly content review.
• P3 — cosmetic. Capitalization or punctuation inconsistencies that do not affect voice integrity. Document for the next style-guide refresh.

The P0/P1 vs P2/P3 distinction is the brand assistant's most consequential judgment call. Over-rating severity overwhelms the fixers and dilutes urgency. Under-rating it lets damage compound. When in doubt, ask: "If a customer landed on this page first, would they trust the brand?" Yes = P2 or P3. No = P0 or P1.

Step 4 — Hand off to remediation owners (15-30 minutes)

The auditor does not fix. The auditor produces a prioritized list and hands it to the people whose work it is — content writers, designers, product managers, legal copywriters. The handoff document includes:

• Prioritized list (P0 → P3)
• Per-item: URL, finding, severity, suggested fix (optional, brief)
• Acceptance criteria — what "fixed" means for each item
• Re-audit date — when the auditor will verify the fix

This separation of duties — auditor finds, owner fixes, auditor verifies — is itself a process-discipline principle borrowed from quality-assurance practice in software engineering and manufacturing. The same person finding and fixing creates conflicts of interest that compound over time.

2.3 Brand-asset organization — version control and single source of truth

The third workflow is the least glamorous and the most consequential. A brand whose assets are organized predictably is a brand that ships consistently. A brand whose assets are scattered across drives, decks, and Slack messages from 2022 produces drift by gravity.

The folder structure principle

One canonical folder structure, used universally, predictably navigable by a new hire on day one. A working pattern:

/brand
  /strategy
    voice-doc.md              ← single source of truth, dated
    positioning-statement.md
    tone-matrix.md
  /visual
    /logo
      primary.svg
      monogram.svg
      apple-touch-icon.png
      brand-marks-source.fig  ← editable source
    /color
      palette.json            ← machine-readable
      palette.pdf             ← human-readable
    /typography
      type-system.md
      webfonts/
    /imagery
      photography-guidelines.md
      illustration-guidelines.md
  /verbal
    naming-conventions.md
    tagline-current.md
    taglines-archive/
    glossary.md
  /audits
    brand-check-2026-Q2.md    ← dated
    brand-check-2026-Q1.md
  /governance
    request-intake-form.md
    escalation-paths.md
    approval-matrix.md

Three rules govern the structure:

1. One canonical version of each asset. If three "primary logo" files exist in three folders, none of them is canonical. Pick one, link the others as deprecated aliases.
2. Date the strategy artifacts. Voice docs evolve; brand audits accumulate. Embedding the date in the filename or document header prevents the most common drift cause: someone using a six-month-old voice doc as canonical.
3. Editable source files live alongside exported deliverables. A PDF of the brand guidelines is for consumption; the Figma or Notion source is for evolution. Both belong in the folder structure with clear naming.

The asset lifecycle

Every brand asset moves through five lifecycle states:

1. Draft — being authored; not yet approved for use
2. Review — submitted for stakeholder approval
3. Active — canonical version, in use across the brand
4. Deprecated — superseded by a newer version; kept available for reference but flagged as old
5. Archived — historical record; not for use

State should be visible in the asset name or metadata. A folder of files where everyone has to ask "is this current?" is a folder that produces drift.

Ownership and update cadence

Every brand asset has a named owner (a person, not a team) and a defined update cadence (when it's reviewed for drift). The voice doc is reviewed annually unless a material rebrand event occurs. Visual identity files are reviewed when components change or every 18 months, whichever comes first. Audit artifacts are produced quarterly and aged out after a year. A brand-asset registry tracks owner + cadence + last-update date for every artifact.

2.4 Tool-options matrix — what to use when

Different workflow stages reward different tools. A useful matrix:

Read this matrix as a default-suggestion, not a prescription. A team with no LLM access can complete every workflow with free general tools. A team with no software budget can complete most workflows with pen-and-paper plus a spreadsheet. The credential is in the discipline; the tools are negotiable.

2.5 Sample artifacts (invented brands)

What does the output of each workflow actually look like? Below are three artifacts for an invented brand: a fictional small-business productivity SaaS called "Quartz" (no relationship to any real product).

QUARTZ — Voice Document v1
Authored 2026-05-01 · Owner: brand assistant · Review: annually

PRIMARY ARCHETYPE
Crisp Modernist. Quartz speaks like a senior IC engineer
giving a teammate a quick answer: minimum words, no hedging,
no metaphor.

SECONDARY ARCHETYPE
Straight Shooter (light). Quartz acknowledges what the
product doesn't do as readily as what it does.

VOCABULARY
Preferred: ship, fix, work, change, simple, fast, exact,
plain, real
Avoided: leverage, unlock, journey, ecosystem, seamless,
empower, transform

SYNTAX MARKERS
Sentence length: avg 11 words; cap 20.
Paragraph length: 1-2 sentences typical.
Voice: active. Always.
Pronouns: "you" for the user; "we" only when speaking as
a team to the user.

IN VOICE
- "Quartz tracks the time you spend in each app. That's it."
- "If our import breaks, we'll fix it within a business day."
- "We don't do dashboards. Plenty of tools do dashboards."

OUT OF VOICE (and why)
- "Quartz empowers you to unlock productivity." → empty
  verbs, no concrete claim
- "On your journey to better focus..." → metaphor we avoid
- "Our innovative platform leverages..." → consultant jargon

QUARTZ — Brand-Check Rubric v1
Authored 2026-05-01 · Owner: brand assistant

Score each page PASS / PARTIAL / FAIL against:

M1: Sentence length averages ≤14 words; no sentence ≥20
M2: No words from the avoided list (leverage, unlock, journey,
    ecosystem, seamless, empower, transform)
M3: All verbs active voice
M4: Pronouns: "you" for user, "we" only as team-to-user;
    never institutional ("our team is excited to...")
M5: Honest about what Quartz doesn't do — at least one
    explicit boundary statement per primary marketing page

QUARTZ BRAND CHECK — 2026-Q2 — Findings
Audited by: brand assistant · 2026-05-15

P0 — BLOCKING
- quartz.example/pricing (line 14): "Quartz empowers teams
  to unlock their productivity potential"
  M2 FAIL (3 avoided words), M3 PARTIAL
  Fix: rewrite as "Quartz tracks time. Teams use it to find
  the work that pays off."
  Owner: content writer · Due: 2026-05-18

P1 — SIGNIFICANT
- quartz.example/about (line 7): "Our innovative platform
  leverages cutting-edge..."
  M2 FAIL (3 avoided words)
  Fix: rewrite per voice doc
  Owner: content writer · Due: 2026-05-30

- quartz.example/blog/welcome (paragraph 2): metaphor
  ("the journey to focused work")
  M1 PASS, M2 PARTIAL
  Owner: content writer · Due: 2026-05-30

P2 — MINOR
- quartz.example/help/import-csv: sentence-length avg 17
  (above 14 floor)
  M1 PARTIAL
  Batch for Q3 content review

P3 — COSMETIC
- footer: inconsistent capitalization of "Quartz" vs
  "QUARTZ" across 4 pages
  Roll into next style-guide refresh