Workflow

Review and publish translations through a structured approval process.

Overview

Every translation in Bosca — whether it is a string, a plural form, or a full document — goes through a structured workflow before it reaches your users. This workflow enforces quality by requiring review and approval before anything goes live, and it creates an audit trail so you always know who changed what and when.

Translation States

Each translation is always in one of the following states:

StateWhat It Means
DraftThe translation is being written or revised. This is the starting state for new or reworked translations.
AI GeneratedThe translation was created by AI. It needs a human to review it before it can move forward.
In ReviewThe translation has been submitted for review by a translator or manager.
ApprovedA reviewer has approved the translation. It is ready to be published.
RejectedA reviewer found issues and sent it back for revision. The translator should address the feedback and resubmit.
PublishedThe translation is live and available in production. Users can see it in your app or website.
ArchivedThe translation has been retired. It is kept for historical records but is no longer active or included in exports.

How Translations Move Through the Workflow

Translations do not jump directly from Draft to Published. They follow a defined path to ensure every translation is reviewed before it goes live. Here are the allowed transitions:

The Main Path

The typical journey of a translation:

  1. Draft — A translator writes or revises the translation
  2. In Review — The translator submits it for review
  3. Approved — A reviewer approves it
  4. Published — A manager publishes it to production

The AI Translation Path

When a translation is generated by AI:

  1. AI Generated — The system creates a machine translation
  2. In Review — A human submits the AI translation for review (possibly after editing)
  3. Approved — A reviewer confirms the translation is accurate
  4. Published — A manager publishes it to production
AI-generated translations cannot be published directly. They must always go through human review first. This is a deliberate safeguard to ensure machine translations are accurate and appropriate before your users see them.

When a Translation Is Rejected

If a reviewer finds issues with a translation:

  1. In Review — The reviewer rejects it with feedback
  2. Rejected — The translation returns to the translator
  3. Draft — The translator revises the text based on feedback
  4. The translation then goes through the review cycle again

Other Transitions

  • Approved back to Draft — If you need to make changes to an approved translation before publishing, it can be sent back to Draft for revision.
  • Published to Draft — A published translation can be sent back to Draft if it needs to be updated (for example, after a product change).
  • Published to Archived — When a translation is no longer needed, it can be archived to keep it out of exports while preserving the history.
  • Archived to Draft — An archived translation can be brought back if it is needed again.

Bulk Transitions

When you have many translations to move at once, bulk transitions save time. Instead of updating translations one by one, you can move all translations for a specific language from one state to another in a single action. Common bulk operations include:

  • Publish all approved translations — After a review cycle, push everything that has been approved to production at once
  • Submit all drafts for review — Send an entire language's worth of draft translations to the review queue
  • Archive all published translations — When deprecating a language, archive everything in one step

The bulk operation only affects translations that are in the specified starting state and where the transition is valid. It safely skips any translations that do not qualify.

Audit History

Every change to a translation is recorded automatically. This creates a complete history that you can review at any time. Each history entry captures:

  • Who made the change
  • When the change was made
  • What changed — the previous state and new state
  • Text changes — the previous text and the new text, so you can see exactly what was edited
  • Origin — whether the change was made by a human, AI, import, or sync
Audit history is essential for compliance and quality tracking. If a translation issue is reported in production, you can trace back through the history to see every version of the text, who approved it, and when it was published. This is especially valuable for regulated industries where translation accuracy is a legal requirement.

Why the Workflow Matters

A structured translation workflow provides several important benefits:

  • Quality assurance — Every translation is reviewed before users see it, catching errors and inconsistencies early
  • AI safety — Machine translations get human oversight, preventing inaccurate or inappropriate text from reaching production
  • Accountability — The audit trail shows exactly who is responsible for each translation decision
  • Compliance — For industries with regulatory requirements (healthcare, finance, legal), the full history provides the documentation you need
  • Rollback — If a published translation has an issue, you can see the previous version and restore it by reverting to Draft