Every segment in a wxrks project passes through four decisions: which engine translates it, which people work on it, what moves it forward automatically, and what stops it from being edited. This article explains how each decision is made and links to the article where you configure it. Use it to find out why a segment ended up where it did.

Who is this for? Account Admins and Project Managers who configure engines, workflows and automation. Linguists and reviewers (Vendors) meet the results of these settings in the Editor.

wxrks routes machine translation and AI separately. Each follows its own order of precedence, from the most specific setting to the least specific.

MT configurations are applied at the Account, Organization or Organizational Unit level, as a default or for a specific language pair. wxrks checks every language-pair setting first (Organizational Unit, then Organization, then Account), then every default in the same order, and finally the built-in wxrks (Powered by Microsoft) engine.

Set up in Connect a Machine Translation (MT) Provider to wxrks. For engines outside the provider list, see Connect your own MT engine (bring-your-own MT).

Each AI operation — Context Sensitive translation, Proofread, Smells and the rest — is routed on its own in Settings > Account Settings > Model Management. By default, all operations use wxrks's own default models, but any operation can be routed to a different model, which overrides the wxrks model for that operation. wxrks looks for the most specific row, in this order:

Set up in Configure Custom AI Providers with Model Management. For your own LLM, post-editing or quality-evaluation engine, see Connect your own AI engine for automatic post-editing (Context Sensitive) and quality evaluation (Smells).

ℹ️ Note: The two ladders are independent. An MT engine never powers an AI operation, and a Model Management row never changes which MT engine runs. They meet in Context Sensitive: the AI operation CS Translation receives the MT output as one of its inputs.

Which source fills a segment during pre-translation is set per project under Translation Settings > Auto Translate From: Context Sensitive, Translation Memory, Machine Translation, or a combination. When more than one is on, Context Sensitive takes priority, then Translation Memory, then Machine Translation — unless Prioritize TM matches over Context Sensitive translation is on and a TM match reaches its threshold. See Project's Translation Settings.

A workflow is an ordered list of workflow steps (for example Translation, then Review). Each file in a project follows one workflow, and each step becomes a task for someone to do.

See Managing Your Custom Workflows and Advanced Workflow Management.

A task goes to a person in one of two ways:

With Project Routing (Settings > Translation Settings > Project Routing), one submission creates several linked projects, one per Organizational Unit. Each language pair is routed to the unit, workflow and team configured for it. A file not assigned to any unit isn't translated, and wxrks warns you before you create the project. See How to use Project Routing in wxrks.

Pipelines (Settings > Automation Settings > Pipelines, Account Admin only) react to events — a project or task changing status, a new project, a new user or Organizational Unit — and run actions without anyone clicking. For example: create or update a workflow, check for manual edits or comments, check a project's work units, change a project's status, update a record, call an HTTP endpoint, or post to Slack. See Getting Started with Pipelines in wxrks.

The confidence gate is the other automatic mover: right after pre-translation, it can confirm segments whose Confidence Score reaches your minimum, so they skip a linguist. See the next section.

A locked segment can't be edited in the Editor until it's unlocked. There are three sources of locks.

In the Segments section of Translation Settings, Confirm segments by confidence score has auto-approve rules that offer two actions for segments at or above the minimum confidence:

See Auto-Confirming Segments by Confidence Score.

The Segments section of a project's or Organizational Unit's Translation Settings tab has:

These settings, and their Confirm segments with 100% / 101% match counterparts, are unavailable while Context Sensitive is on without Prioritize TM matches over Context Sensitive translation. In that setup Context Sensitive rewrites TM matches instead of keeping them.

Segments with nothing to translate (only numbers, or only special characters) are locked and confirmed automatically. On an Organizational Unit's Translation Settings tab, Unlock options can leave them open instead: Unlock segments with numbers, Unlock segments with special characters, or both. These options apply when a project is created.

Locking and unlocking in the Editor requires the Lock Segments permission ("Allow user to lock/unlock segments in Editor"), which Account Admins and Project Managers have by default and Vendors and Clients don't. Grant or remove it in People > Users > [User] > Permissions. The Editor's bulk selection menu offers Lock selected segments, Unlock segments and the non-translatable unlock options. A linguist who tries to type in a locked segment sees "Oops, you have to unlock the segment prior to changing it!".