Security, sustainability, scalability. Risk gated on your rules.
wxrks is the translation management system enterprises run their AI on. Content enters from your systems, is translated on the models you choose from your own memories and terminology, gated per segment against thresholds you set, and returned with a record. wxrks supplies the framework. The configuration is yours.
Four pillars, one platform, your configuration
Security your review team can sign off. A configuration that survives every model change. Behaviour that holds from one page to a hundred million words. Quality treated as risk, gated on thresholds you set.
Security
Built for security review. SOC 2 Type 2 held continuously since 2020. AWS in the US or Ireland at your election, single-tenant available. SSO, SCIM, MFA, role-based access to the individual action. Stateless models, no training on your content, written into the DPA.
Sustainability
A configuration that outlives any model. Your memories, terminology, instructions and decisions live in your assets, not in a fine-tuned model, so switching provider or model family changes one setting and loses nothing. A different model for each operation.
Scalability
Predictable at one page and at a hundred million words. Account, Organization and Organizational Unit hold brands, regions, business units and suppliers apart under one roof. Unlimited users and units. No fixed throughput limit.
Risk
Quality as gated risk management. Every segment scored with a named finding, a severity and an explanation. Thresholds per content tier set by you. Output above threshold passes and locks. Below it is fixed automatically where the fix is deterministic and routed to a named person where it is not.
Risk is gated while translation happens, not after
In a conventional workflow every review round is a human pass, so a file gets one review and whatever it misses ships. Here evaluation is a machine operation. It runs as many times as your rules require, at no added cost or turnaround, and you set the threshold at every decision point.
Ingest and assemble context
Content arrives through a connector, the API, the CLI or the portal. The source is analysed for subject, tone and style, and the relevant context is retrieved semantically from your memories, glossaries, reference material and prior decisions.
Translate in your voice
Context Sensitive Translate produces one suggestion per segment with your terminology, your brand voice and your precedent already inside it, on the model you assigned. A Sous-Chef agent carries the register and audience instruction for that brand or unit. Nothing is trained.
Score every segment
Smells evaluates each segment semantically and returns a named finding with a severity and a short explanation. Omission, undue addition, register mismatch, terminology inconsistency, numeric error, tag misplacement and more. Severity becomes a confidence score per segment and an aggregate per file.
Apply your rules
Per content tier you decide what happens next. Deliver, fix deterministic issues and deliver, or fix what is deterministic and route the rest to named reviewers. A reviewer sees the twelve segments that need judgement, with each issue named, rather than four thousand words to read against source.
Re-evaluate, deliver, learn
The completed file is scored again against your delivery thresholds. If it fails, it goes back. Translations return to the source system with provenance on every segment, machine, memory or human, and a reviewer's correction on one segment changes the suggestion on the next, with no retraining cycle.
专为机构和企业团队打造
What your security team receives, and what they find
SOC 2 Type 2 report, penetration test summary, policy set and CAIQ-Lite under NDA. Customer-requested penetration tests permitted. Security review is a precondition, not a phase, so it runs in parallel with the pilot.
Assurance
SOC 2 Type 2 examined annually by an independent CPA firm, held continuously since 2020. Continuous control monitoring. Report available under NDA.
Hosting and isolation
AWS only, us-east-1 or eu-west-1 at your election, across multiple availability zones. Row-level security bound to the account, organizational units as a second boundary inside it. Dedicated single-tenant environments on the same release train.
Identity
SSO with Okta, OneLogin and Microsoft Entra ID. SCIM provisioning and deprovisioning from your directory. MFA. Four roles, permission groups across seventeen categories, scoped to the unit.
AI and content
Every model runs stateless. No customer content trains any model, by wxrks or by any provider. Provider list published with purpose and processing location, 30 days' notice of change. Written into the DPA.
Contract
DPA with Standard Contractual Clauses and UK addendum. 99.9 percent monthly availability with service credits. Export in open formats at any time. Deletion within 30 days of a verified request, to NIST SP 800-88.
Availability and recovery
Recovery time and recovery point objectives of four hours. Continuous backup with point-in-time recovery, restore tested annually. Threat detection and audit trails across every account and region. Public status page.
Measured on production programs
One published study and the results named customers report on their own configurations. What your program sees depends on your content, languages and baseline, which is what the proof of concept measures.
wxrks study, 4.3M segments
39%
More effective than traditional machine translation post-editing, measured as translation edit rate against a neural MT baseline. Average edit distance fell from 18.23% to 11.17%.
Coca-Cola Europacific Partners
95%
Share of content running through the agentic framework with no human review step, on CCEP's own thresholds. 41,000 employees, 31 countries, 155 languages.
Harley-Davidson
70%
Turnaround reduction across 42 languages for owner and service documentation, with cost down more than half year on year since 2018.
Zendesk
70%
Lower translation technology spend across 30 markets, with a 40% efficiency gain, through the WordPress connector and automated project creation.
Automation Anywhere
50%
More than half off translation cost, with requests raised from Microsoft Teams by people who are not translators.
Scaled Agile
60%
Reduction in time to market on training content.
Study period 22 June to 23 December 2023, English source, 20 language pairs. Context-sensitive output outperformed in every language pair tested. Customer figures are as reported by each program.
From the teams who run their work on it
It doesn’t ask us to supervise a machine — it gives our linguists a semantic exoskeleton. It allows us to move from a defensive stance of fixing mistakes to a progressive stance of amplifying creativity.
Rik Grant, Global Head of Language & Cultural Services, Tag (a dentsu company)
When localisation is framed primarily as a clean-up step for machine generated output, linguists risk being positioned as fixers rather than experts. The future isn’t a compromise between human and machine but rather, a synthesis.
Tag, a dentsu company, announcing the wxrks partnership, May 2026
A named team, a dated plan, a pilot that covers every language
A 90-day program with a named owner on each side. The first live project typically runs inside the first two to four weeks.
01
Discovery
Working sessions on your memory and glossary estate, users and roles, content types and volumes by language, workflows and review model, engines, systems in scope, identity and residency, reporting. Output is a written inventory and a gap list with an owner against every item.
02
Build
Unit design, engine routing, workflow steps and review placement, connectors and identity federation. Memories, glossaries and termbases imported at any size and tuned until match behaviour lines up with what you see today.
03
Pilot
Every language and content type in scope, end to end, each checked by its own reviewer or agency. Findings logged with category, root cause, owner and target date and closed before cutover.
04
Cutover and cadence
Hypercare with daily contact, then a weekly check-in with a written status report, quarterly business review, annual review, and a roadmap cadence with the people who ship the product.
One annual subscription, priced on processed volume
Plans are sized by processed volume and the depth of work the platform performs on it. Leverage from your memory or from your own engines is never penalised.
Not per seat
Your team, your requesters and your suppliers are not licensed per seat. Unlimited users and organizational units in one instance.
Implementation inside the subscription
In enterprise plans, implementation, onboarding, training, asset import and export on exit sit inside the subscription. No first-year spike and no separate implementation invoice.
Support with response targets
Dedicated customer success manager at enterprise tiers. Response targets under two hours at Quantum and Pulsar and under one hour at Quasar.
持续发货
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.avif)
A 30-day proof of concept
on your content
Full platform, your real content, your target locales, up to three of your existing suppliers working inside the instance as they would in production. Four measures agreed at the start. Edit distance against your current baseline, time from intake to delivery, the working experience for your team and suppliers, and whether the platform fits your process rather than the reverse.
Book a demo
