Plug & play models
Connect the provider that fits each language or market. Change it later without rebuilding the workflow. No lock-in to one AI vendor.
FEATURE · AI WORKFLOWS2026
AI where it helps. Your people where it matters.
PIM Gate lets you plug in the language model that fits each language and market — OpenAI, Anthropic, DeepL, Azure OpenAI or an EU-hosted or local model. AI output enters your workflow as a proposal, is reviewed by your people, and is marked as automatic in the audit log. You stay in charge of provider, region and result.
MODEL PER LANGUAGE OR MARKETHUMAN IN THE LOOPSOURCE: AUTO IN AUDIT
01 · What it does
Plug in a model. Put it to work. Keep control.
Governed workflows
AI runs inside PIM Gate's workflows — with your terminology, your rules and a review step — not in a side tool someone copies from.
Predictable usage
AI credits are included in every version. Beyond that, usage is billed transparently per 1,000 assets.
02 · Models
The right model for each language — your choice, your region.
No single model is best at every language, and not every market accepts every provider. PIM Gate treats the model as configuration: assign a provider per language, and switch when a better option appears. If your policy requires it, route everything through EU-hosted or locally operated models. Your terminology and translation memory travel with the request, so quality doesn't depend on the provider alone.
- Connect OpenAI, Anthropic, DeepL, Azure OpenAI or EU/local models2026
- Provider per language
- Region choice, including EU-hosted and local models
- Provider switch without changing the workflow
Models
The right model for each language — your choice, your region.
03 · Translation
Translate only what changed — in your terminology, reviewed by your people.
Translation is where AI earns its keep in product data today. PIM Gate detects which segments are new or changed, reuses your translation memory, checks your approved terminology, and sends only the remainder to the model you chose for that language. Every proposal lands in a review queue per language. Prefer human translators? Export a small delta package, re-import it, done.
- Delta-only: unchanged segments are never re-translated2026
- Translation memory (TMX) and terminology (TBX) reuse2026
- Review queue per language, with approve / request change2026
- Export/import packages (XLIFF, XLSX, TMX) for external translators2026
- CMS content translation for Ibexa DXP and DrupalCOMING SOON
Translation
Translate only what changed — in your terminology, reviewed by your people.
See the translation workflow in detail →
04 · AI credits
AI usage included — and priced plainly beyond that.
Every PIM Gate version includes a monthly allowance of AI credits for translation and other AI steps. When a large catalog launch or a new market needs more, additional usage is billed at €8 — no separate AI contract, no surprise invoice from a model provider you never signed with. Credits are counted per tenant and visible in the admin console.
- Monthly AI credits included in every version
- Additional usage: €8
- One invoice — AI usage runs through your PIM Gate subscription
- Usage visible per tenant
| Version | AI credits / month |
|---|---|
| S | 1,000 |
| M | 5,000 |
| L | 20,000 |
| XL | 50,000 |
| Beyond included | €8 |
05 · Principles
Three rules we don't bend.
- Human in the loop AI output is a proposal. Nothing is published to a portal, feed or API without the review step you configure for that language, market or attribute.
- You choose provider and region Model, provider and hosting region are your decision, per language or market — including EU-hosted and local models. PIM Gate itself runs in Germany, with Switzerland on request.
- Full transparency Every AI change is marked source: AUTO in the append-only audit log. You can always tell what a machine wrote and who approved it.
Principles
Three rules we don't bend.
06 · On the roadmap
Next: AI that helps beyond translation.
- Asset tagging Suggested tags and alt texts for product images, reviewed before they count.ROADMAP
- Product descriptions Draft descriptions generated from structured attributes and your style guide, per language and channel.ROADMAP
- Anomaly detection Flags values that look wrong against the rest of the class — a 10 bar gauge listed at 100 bar, a price off by a factor of ten.ROADMAP
- Configurable AI scripts per tenant Your own AI steps — prompts, rules and target attributes — defined per tenant and run inside the same governed workflow.ROADMAP
On the roadmap
Next: AI that helps beyond translation.
See the roadmap →
07 · Specifications
AI workflows at a glance.
| Spec | Value | Status |
|---|---|---|
| Model providers | OPENAI · ANTHROPIC · DEEPL · AZURE OPENAI · EU / LOCAL | 2026 |
| Assignment | PER LANGUAGE · PER MARKET | 2026 |
| Product-data translation | DELTA · TMX · TBX · REVIEW QUEUE | 2026 |
| Languages in production | de-DE · de-CH · en-GB | — |
| CMS content translation | IBEXA DXP · DRUPAL | COMING SOON |
| Review | CONFIGURABLE PER LANGUAGE / MARKET / ATTRIBUTE | 2026 |
| Audit marking | source: AUTO | 2026 |
| AI credits / month | S 1,000 · M 5,000 · L 20,000 · XL 50,000 | — |
| Additional AI usage | €8 | — |
| Asset tagging · descriptions · anomaly detection | — | ROADMAP |
| Configurable AI scripts per tenant | — | ROADMAP |
08 · Works with
Models and formats.
OpenAIAnthropicDeepLAzure OpenAIEU-hosted modelslocal modelsTMXTBXXLIFFXLSXIbexa DXP · COMING SOONDrupal · COMING SOON
Provider names are trademarks of their owners; no partnership is implied.
09 · FAQ
FAQ
- Only if you decide so. MT/LLM output is a proposal; nothing is published without the review step you configure. You can also work fully with human translators via export/import packages.
- PIM Gate runs in Germany, with Switzerland on request. Which AI provider and region process a request is your choice per language or market — including EU-hosted and local models.
- Bring them. PIM Gate imports TMX and TBX and uses them before any model is called, so existing translations and approved terms are reused.
- Yes. The model is configuration, not code. Switching a language to another provider does not change the workflow or the review step.