Multilingual Intake Translator

The problem this kills

A ticket lands in Portuguese. The next one is in German, then Polish, then Quebec French. Your team speaks two of those languages on a good day. So tickets get parked, machine-translated in a browser tab, copy-pasted around, and answered slowly - or worse, answered wrong because someone mistranslated a product name or a SKU.

Hiring a native speaker for every market is not realistic for a small team. But losing customers because you can't read their first message is not acceptable either. You need a way to read every ticket, work it in the language you actually think in, and reply in the customer's language - without guessing and without sending a machine translation no human ever checked.

What you'll build

An internal, login-only web app that turns your small team into a multilingual support desk:

• A ticket comes in (from a form or a CSV import) in whatever language the customer wrote.
• The tool detects the language and translates the message into your agent's working language, showing the original right next to the translation so nothing is hidden.
• Your agent works the ticket normally and writes (or asks the AI to draft) a reply.
• The AI translates the reply into the customer's language, and shows your agent a back-translation (the customer's-language reply translated back to your language) so the agent can confirm it actually says what they meant.
• Your brand names, product names, and SKUs are protected by a glossary so they are never translated or mangled.
• The agent approves, and only then is the reply sent via Resend - with both languages stored on the ticket forever.

What's inside the Implementation Plan

• It starts by interviewing you about your business. Before a single line is built, the plan has the AI agent ask you about your markets and languages, the systems and spreadsheets you use today, the exact shape and naming of your ticket data, your glossary of do-not-translate terms, your volumes, and your approval rules. It reflects back a short tailored spec and waits for your thumbs-up. You get a tool shaped around how you actually work - not a generic template.
• A step-by-step build, where each step ends with a ready-to-paste prompt for your AI coding agent.
• A clear data model for tickets that stores original text, detected language, translated text, drafted reply, back-translation, and approval state.
• The glossary feature that locks your product names and SKUs out of translation.
• The side-by-side review screen so your agent stays accountable for every word sent.
• A verification checklist so you know it works, plus a no-integration CSV fallback path.

The governance it includes (this is the point)

This isn't a toy translator. The plan builds in the controls a real support operation needs:

• Login so only your team can open the tool.
• Row-level security so each organization only ever sees its own tickets.
• A complete audit trail - who translated, who edited, who approved, and exactly when.
• A hard human-in-the-loop approval gate - the AI drafts and translates, a person reviews both the incoming translation and the back-translated reply, and nothing reaches the customer until a human approves.
• Duplicate guards keyed on the ticket ID so the same ticket can't be processed or answered twice.

Who it's for

Support teams serving multilingual customers without a native speaker on every shift. If you run ops or BPM for a small support desk that covers several markets, this is for you - you don't need to be a developer.

You've got this - paste the first prompt and let the plan interview you.