See Every Job's Progress at a Glance
The change everyone will notice first is a new Progress column on the Jobs and Items grids, on by default in both Standard and Codyt projects. Each job now shows its segments split into red, approved, translated, and untranslated, with a percentage-approved label right there in the list. Hovering reveals the totals and how recently the figures were updated, clicking opens the existing Counts and Activity view, and a small refresh icon recomputes a single job on demand. The column behaves like any other, so you can hide or reorder it, and jobs with nothing to track show an empty bar or N/A rather than a confusing zero. For anyone running delivery, it turns a page of job names into a live picture of where the work actually stands. The details are in the guide to the Jobs main page.
The same instinct, put the number where the decision is made, shows up in two more places. Completed Standard and Codyt jobs now carry an edit-distance figure in the word-count popup, a percentage that shows how much the translation was changed during the job; it looks only at the words and ignores formatting tags, so adjusting tags without touching text counts as no edit. The measure is explained under job edit distance. And when you create and assign work, the Standard project assignment wizard now includes a Counts column, so you can open the full word-count popup and review words, characters, segments, and pages without leaving the wizard, on a cleaner two-column layout. The steps are covered in create and assign jobs.
There is also a new lever for how status flows through a project. A project manager can now make one workflow step responsible for the project-level segment status, so every status change in a job built on that step, whether set by hand or by a QA check, saves straight to the global status without waiting for the workflow to finish. The option is off by default, set per project, and appears only when independent segment status per job is enabled, which pairs it naturally with the per-job progress view from earlier this month.
AI Workflows That Finish, and Explain Themselves When They Do Not
Automation is only useful when you can trust it to complete, so this release removes a quiet failure mode. Large documents processed through AI workflows, machine translation, automated post-editing, and quality checks, could previously fail partway through because the internal calls those workflows make to fetch translation memory and terminology matches counted against an hourly API limit. Those internal calls are now exempt, so large machine translation jobs complete reliably and several can run at once, while normal API usage stays protected by the same limits as before.
When something does go wrong, you now find out why. A failed AI workflow shows a readable, actionable message in the job trace, an authentication or rate-limit problem for example, along with identifiers that help support trace the issue, instead of cryptic technical output. Failures of post-editing workflows assigned to a supplier, which used to be silent, now surface as job messages. Reviewers get a smaller but welcome touch too: AI-raised issues in the current editor's QA panel now read as AI Addition, AI Omission, and AI Terminology rather than meaningless rule numbers, so it is clear at a glance what each quality finding is about.
More Control Over Access and Quality
Administrators gain real teeth on the security side. A new Enforce SSO access right, under a new Advanced Security group in the user-profile editor, lets you require that certain users sign in only through your single sign-on provider. Enable it on a profile and password sign-in is refused for everyone on that profile, so corporate employees can be held to SSO while freelancers on a different profile keep password access. The right is off by default, has no effect unless SSO is configured, and a locked-out user is always recoverable by reassigning their profile. This sits with the rest of the platform's security and access controls, and the setup is documented in the access-rights scenarios guide.
The Access Rights page itself is far easier to work with. A new search box and filter bar let administrators find rights by code, name, or description and narrow the list by license, enabled or disabled state, or which values differ from the profile default, and the table now shows a State indicator, a License type column, and a Dependencies column that spells out which rights a permission requires or relates to. Nothing about the permissions themselves changed; this is a presentation refresh, described in full on the access rights page. Quality control gains a new option as well: a built-in QA rule, "Punctuation - Brackets and quotes," flags translations where brackets, parentheses, or quotes are left unpaired or differ in number from the source, with a choice of which pair types to check and exception patterns for intended cases like numbered lists. It is off by default in existing profiles and enabled per QA profile, alongside the other QA categories.
Cleaner Files, Fewer Surprises
A run of fixes makes several formats behave the way you would expect. Marking an XLIFF version 1 file online no longer fails when the file's declared target language differs from the project target, matching how version 2 files already behave. In Word, Excel, and PowerPoint files set up to translate only text in specific colors, non-colored text sitting at the very start or end of a cell or paragraph is no longer dropped; it is kept as a protected tag and restored in the right place when the file is generated. Multilingual Excel files that use automatic language-column detection now download correctly even when a sheet without recognizable language columns sits before a valid one, and the "Show source text if translation is missing" option now works for multilingual Excel just as it does for Word, PowerPoint, and single-language Excel. Finally, segmentation rules that break a segment at a tab character now apply to Word tables of contents and bullet points and to InDesign content, where they were previously ignored; because this can shift segment boundaries, it takes effect for documents parsed under the new parser version, so existing translation memory matches on older files are left undisturbed.
Also in This Release
A handful of smaller refinements round out 2026.2.13.0. Creating an order through the API with many files now responds almost immediately and registers the files in the background, so high-volume orders no longer time out, with validation problems reported through the operation status the integration polls for. The Manager and Project Manager fields on the Order form configuration page now use a fast type-ahead search instead of loading the full user list. The Add resource popup opens with the "Project memories" filter unticked so it loads quickly on large accounts, with the filter still available on demand. An API call that retrieves order form settings no longer returns an internal error when a workflow template references a deleted or renamed task code. And the Find Projects page now loads correctly for users who are not allowed to see order IDs, keeping the order ID filter hidden while the target language filter stays available to everyone.
The full technical notes for 2026.2.13.0 live on help.wordbee.com, with configuration steps and screenshots for each item. If any of these changes will reshape how your team works in Wordbee, talk to our team before you roll them out. We would rather walk you through setup than let a release stumble in production.