**Note that this could be done after the 1.0 release.** Fable 5 pass Here's a documentation-ready set of use cases, grouped by consumer and framed for the post-[#3586346](https://git.drupalcode.org/project/ai_context/-/work_items/3586346) world where anything — agents, CKEditor, translation, automators — can be a context consumer. Each one names the CCC features it exercises so readers connect the scenario to the mechanics. --- ## Writing assistance (AI CKEditor) **1. Brand voice everywhere, enforced once.** The marketing team maintains one "Voice and tone" context item — reading level, banned phrases, Oxford comma policy, how to refer to the company. It's scoped **Global**, so every AI-assisted edit in CKEditor follows it, on every content type, without any editor having to remember it exists. When the brand refresh lands, one edit updates every future AI interaction site-wide. **2. Content-type writing rules.** Press releases need the legal boilerplate and inverted-pyramid structure; product pages need spec tables in a fixed order; blog posts allow first person. Each ruleset is a context item scoped to its **Entity Type** (all Press Release nodes, all Product nodes), so the AI gets different instructions depending on what the editor is writing — automatically. **3. Canonical facts for high-stakes pages.** The "About Us" page has exact founding dates, executive names, and headcount that AI must never improvise. A context item scoped to that **Specific Entity** injects the canonical facts whenever AI touches that page — and only that page. **4. Campaign windows.** Holiday promotion messaging ("lead with the free-shipping offer, never mention the January price change") is a context item **scheduled** to publish November 15 and unpublish December 26. AI writing help is on-message during the campaign and clean after it, with no manual cleanup. ## Translation (AI Translate) **5. Glossaries and do-not-translate terms.** Product names, trademarks, and technical terms that must stay in English live in a context item scoped to the **Language** scope. Per-language items add locale rules: German uses formal *Sie*, Japanese follows the company's katakana conventions. Every AI translation request gets the right glossary for its target language. **6. Market-specific legal requirements.** The French site legally requires certain consumer-rights phrasing; the Canadian site needs bilingual disclaimers. Language- and section-scoped context items make AI translation and localization respect regulatory differences without translator training. ## Automators (bulk and background AI) **7. Meta descriptions and SEO at scale.** An automator generates meta descriptions for 5,000 products. The SEO team's context item (character limits, keyword placement, no clickbait) is scoped to the **use case**, so the whole batch follows the house SEO policy — and when the policy changes, regeneration picks it up instantly. **8. Consistent auto-tagging.** An automator classifies incoming content into taxonomy. Context items define what each category *actually means* on this site ("'Insights' is thought leadership, not news; news goes in 'Newsroom'"), so AI classification matches the editorial team's mental model instead of generic dictionary definitions. **9. Alt text that meets your accessibility bar.** Automated alt-text generation consumes the accessibility team's context item: describe function over appearance, no "image of", 125-character target, brand-name conventions. Accessibility standards become enforceable configuration instead of a wiki page nobody reads. ## Agents, assistants, and chatbots **10. Support assistant with governed answers.** The support chatbot's refund policy, escalation rules, and tier definitions are context items subscribed by **use case**. Support leads — not developers — edit the policy, and it goes through **editorial workflow** so a manager approves changes before the bot starts saying them. **11. Department-scoped intranet assistant.** One assistant serves the whole intranet, but HR policies inject only in the HR **site section**, engineering runbooks only in engineering. Tag and section scoping keeps each conversation relevant and prevents cross-department leakage. **12. Hard guardrails as content.** "Never provide medical advice. Never quote prices — link to the pricing page. Always disclose you're an AI." These are global, **moderated** context items owned by legal. The compliance team controls AI behavior through a workflow they already know, with **revisions** as the audit trail of who changed what, when. ## Governance and compliance (cross-cutting) **13. Regulated industries.** A healthcare or financial site requires that every instruction reaching an AI system be reviewed. CCC context items are content entities with **workflow states, revisions, and scheduled publishing** — so the existing compliance apparatus (draft → legal review → published) governs AI behavior, and **usage tracking** shows exactly which approved context was in play for any given AI interaction. **14. Trustworthy civic and electoral information.** A government or election site injects authoritative facts — polling dates, eligibility rules, official contacts — so AI features answer from the canonical source instead of hallucinating. Specific-entity and use-case scoping keeps the authoritative context attached to the pages and tasks where it matters. **15. Multi-brand and multisite consistency.** An organization running several brands scopes voice, terminology, and legal context per **site section** (or per site), so one platform serves all brands with AI that never mixes them up. ## Site building and operations **16. Your content model, explained to the AI.** A "site architecture" context item describes how this site actually works — "events use the Event content type with a required venue reference; never create locations as plain text." Agents doing content operations build things the way *your* site expects, not the way generic Drupal training data suggests. **17. One token budget, many consumers.** All of the above compete for limited prompt space. CCC's **priority ordering and token budgeting** decide what wins when space runs out — guardrails and legal text first, style preferences last — consistently across every consumer, instead of each integration inventing its own truncation. --- The through-line worth stating explicitly in the docs: **CCC turns AI behavior into governed content.** Everything editors already know — workflow, revisions, scheduling, translations, permissions — now applies to what AI systems are told, and (after #3586346) applies uniformly whether the consumer is an agent, a CKEditor button, a translation job, or a bulk automator.