/ for enterprise /
Security, sustainability, scalability. Risk gated on your rules.
The translation management system enterprises run their AI on. Your systems in, your models, your thresholds, a record on every segment.
- SOC 2 Type 2 since 2020
- Forrester Wave TMS Q3 2025
- AWS US or EU
- G2 4.6 / 5
/ 300+ companies run translation on wxrks /
/ the structure /
Four pillars, one platform, your configuration
Security
SOC 2 Type 2 since 2020. AWS in the US or Ireland. Stateless models, nothing trained on your content.
Security and dataSustainability
Your assets carry the configuration, not a fine-tuned model. Switching provider is an edit to one row.
SustainabilityScalability
Brands, regions and suppliers apart under one account. Unlimited users and units. No fixed throughput limit.
ScalabilityRisk
Every segment scored and named. Above your threshold it passes and locks. Below it, fixed or routed to a person.
How the gate runs/ security /
Built for security review
The documents your security team will ask for are ready under NDA. Review runs in parallel with the pilot.
Security and data position- Assurance
- SOC 2 Type 2, independent CPA firm, held continuously since 2020.
- Hosting
- AWS us-east-1 or eu-west-1, multiple availability zones, single-tenant available.
- Identity
- Okta, OneLogin, Microsoft Entra ID. SCIM. MFA. Permissions scoped to the unit.
- Encryption
- TLS 1.2+ in transit, AES-256 at rest, keys in AWS KMS.
- AI and content
- Every model stateless. No training on your content, by anyone. In the DPA.
- Availability
- 99.9 percent monthly with credits. RTO and RPO of four hours.
- Deletion
- Export any time. Deletion within 30 days, full purge within 60, to NIST 800-88.
Nothing is trained,
so nothing is lost.
/ sustainability /
Models change every quarter. The configuration does not.
Memories, terminology and reviewer decisions are applied at request time. The model behind an operation is one row you can change.
the model is one row · the assets stay
One row per decision
Scope, operation, provider, model, credentials.
A model per operation
Translate, Proofread, Smells, Fix Tags, each where you choose.
Tokens where you decide
Matches never reach a model. Repeats propagate once.
- OpenAI
- Azure OpenAI
- Anthropic
- Google Gemini
- Vertex AI
- Groq
- OpenRouter
- Cerebras
- Any base URL
- DeepL
- Amazon
- Microsoft
- Intento
- TransPerfect
- ODIN
/ scalability /
One account. Brands, regions and suppliers kept apart.
Account, Organization, Organizational Unit. Settings cascade down. Each unit sees only its own work.
projects live in units · supplier A never sees supplier B
Unlimited users and units
Provisioned from your directory through SCIM.
No fixed throughput limit
Files up to 2 GB. Work splits across suppliers and reassembles.
Arrival without a request
Connectors create projects. Rules route, quote and assign.
- Adobe Experience Manager
- Contentful
- Drupal
- WordPress
- Webflow
- Optimizely
- Kentico
- Marketo
- Eloqua
- Knak
- Zendesk
- ServiceNow
- Pendo
- Figma
- GitHub
- GitLab
- Bitbucket
- Phrase
- QuickBooks
- PayPal
- REST API
- Webhooks
- CLI
- MCP server
Quality is a risk to gate,
not a score to report.
/ risk /
Risk is gated while translation happens, not after
Evaluation is a machine operation. It runs as many times as your rules require, at no added cost or turnaround.
- 1Content arrives from your systems with its context.
- 2Translated in your voice, on the model you assigned.
- 3Smells names each finding and its severity.
- 4Your thresholds decide: pass, fix, or route.
- 5A reviewer sees twelve segments, not four thousand words.
- 6Scored again before release. Fails go back.
- 7Delivered with provenance on every segment.
/ who holds the controls /
wxrks supplies the framework. The configuration is yours.
Models
Which providers, for which operations and units.
Assets
Which memories and glossaries apply, in what order.
Thresholds
Confidence per content tier, and what happens either side.
Workflow
The steps, their order, and where a person sits.
Structure
Brands, regions and suppliers as units.
Environment
Region, identity provider, retention, tenancy.
/ evidence /
Measured on production programs
Less editing than neural MT post-editing, by translation edit rate.
95%
Of content with no human review step. 41,000 employees, 155 languages.
70%
Faster turnaround across 42 languages.
70%
Lower translation technology spend across 30 markets.
50%Automation AnywhereOff translation cost. Requests raised from Microsoft Teams.
60%
Less time to market on training content.
Improvement over neural MT post-editing, by language
Translation edit rate, 20 language pairs, English source, June to December 2023.
It doesn’t ask us to supervise a machine — it gives our linguists a semantic exoskeleton.
The future isn’t a compromise between human and machine but rather, a synthesis.
No workflows, no audit trail, no TMs or glossaries, no consistency. And frighteningly frequently: no expert human oversight.
/ commercial model /
One annual subscription, priced on volume
Not per seat, not per supplier. Implementation and onboarding inside the subscription.
See plans- Discovery
Inventory and gap list, an owner on every item.
- Build
Units, routing, workflows, connectors, identity.
- Pilot
Every language in scope, checked by its own reviewer.
- Cutover
Hypercare, then weekly check-ins and quarterly reviews.
First live project inside the first two to four weeks. Response under two hours at Quantum and Pulsar, under one at Quasar.
/ from enterprise evaluations /
Asked in RFPs, answered here
Can we use our own models or MT engines?
Yes. Eight AI providers, eight MT engines, or any endpoint, with your keys, per operation and per unit.
Is our content used to train anything?
No. Every model runs stateless. Written into the DPA.
Do we have to replace our translation vendors?
No. They work inside your instance as vendors, on your configuration, without per-seat licensing.
Where is data hosted?
AWS, us-east-1 or eu-west-1, at your election. Single-tenant available.
How do we keep brands or business units apart?
As Organizational Units, each with its own assets, models, prices, members and audit trail.
How do we get off our current TMS?
Import memories and glossaries at any size. Export in TMX, TBX, XLIFF, XLSX and CSV at any time.
Where does human review happen?
Wherever you place it. Segments below your threshold route to a named reviewer.
/ next step /
A 30-day proof of concept on your content, at no cost
- 01Edit distance against your baseline
- 02Time from intake to delivery
- 03How your team and suppliers experience it
- 04Whether it fits your process, not the reverse
