# Services > Human in the loop, development, consultancy and language data: four services that take AI localization from first prompt to governed production. Source: https://www.morphologic-localisation.eu/services/ --- ![](/_astro/services.BhdzCBDH_ZsAb0N.webp) Services # From first prompt to governed production. Four services for what AI localization needs at scale: linguists in the loop, tools built for your systems, processes that hold, and language data you can trust. [Human in the loop](#human-in-the-loop) [Development](#development) [Consultancy](#consultancy) [Language Data](#language-data) Human in the loop ## Linguists are your quality gatekeepers. Automated processes need quality gates. Where they sit depends on the quality you expect and the risk you can accept: a product launch needs more checkpoints than an internal knowledge base. At each gate, our linguists step in where they add the most value, from post-editing to in-context review, so the output meets your quality goals before it moves on. ### What's included - Risk-based gate design — We decide with you where human review pays off in your process chain, based on content type, audience and risk. - Post-editing — Linguists correct AI output to the agreed level, from light to full post-editing. - In-context review — Translations checked in the running product or final layout, where truncations, ambiguities and context errors show up. - Quality evaluation and feedback — Sampling and error scoring against a widely accepted standard such as MQM, with findings fed back into prompts, terminology and data, so the same error doesn’t come back. Development ## From a single skill to a full platform. Every project and every process needs different tools. Sometimes a single skill or agent added to the process is enough; sometimes it takes a set of agentic workflows connected through MCP servers. And where you want it ready to run, [AILT](https://ailt.cloud/), our language transformation platform, delivers AI-native localization out of the box. ### What's included - Prompt engineering — Prompts designed and tested on your content, so results are consistent and repeatable. - Skills — Packaged instructions and resources that teach an AI assistant a task, such as applying your style guide. - Agents — AI agents that carry out a localization step on their own, such as terminology checks or QA, and hand over at the quality gates. - MCP servers — Connections through the Model Context Protocol that give AI agents secure access to your systems: TMS, terminology database, repositories. - [AILT](https://ailt.cloud/) — Our AI-native platform: context-aware AI translation and QA with your own terminology and translation memory, in your own cloud tenant. Consultancy ## You know your processes. We bring the language. Your teams know the processes in their own domain inside out. What is often missing is the language perspective: how terminology, markets and AI output behave once content leaves the source language. We bring that perspective, and our experience from years of localization projects, into your processes, so language is designed in from the start rather than fixed at the end. ### What's included - Process design — Localization workflows with AI built into your existing processes, with quality gates and handovers defined from the start. - Tool recommendations — Advice on translation management systems, AI services and QA tools that fit your systems, content and budget. - Risk assessment — Where AI output can go out as is and where it needs review, assessed by content type, market and regulatory exposure. - Language-specific requirements — What each target language needs, from scripts and text expansion to formality and legal and cultural conventions, and how AI handles it. Language Data ## The fuel that runs your LLM. Even a Formula 1 engine won’t perform on regular fuel. The same goes for AI: the best model only delivers when the terminology, reference texts and examples it draws on are right. We help you create, curate and use state-of-the-art language data for your AI language automation. ### What's included - Terminology work — Term extraction, definition and approval, managed in a termbase your AI tools can query. - Language data creation — Reference translations, example sets and test data built for your domain and languages. - Language data QA — Cleanup of translation memories and corpora: duplicates, inconsistencies and outdated translations removed before AI draws on them. - Data annotation — Labeled data for evaluating and fine-tuning AI models, from error annotation to quality ratings. - Style guides — Language-specific rules for tone, formality and conventions, written so linguists and AI can both apply them. How we work ## Start with a free consultation. Every engagement starts with a free initial consultation about your automation potential. From there, we decide with you, step by step, how your AI localization should work. 1 ### Free initial consultation Your content, languages and goals: where AI can take over, and what that saves. 2 ### Workflows and processes How your workflows change, where the quality gates go, and who hands over to whom. 3 ### Skills, agents and tools Which skills, agents and tools fit each step, from a single prompt to AILT. 4 ### Risk assessment Data protection, confidentiality, quality risks and reliance on vendors and models, each with a plan to manage it. Questions & answers ## What our customers ask first. Short answers to the questions that come up in almost every first conversation. ### AI and quality Can AI replace human translators? For some content, largely; for some, not at all. AI handles the volume, and linguists sit at the quality gates where the risk calls for them. Where that line runs depends on your content, your markets and your quality goals, and finding it is part of our work. How do you measure the quality of AI output? With sampling and error scoring against a widely accepted standard such as MQM, measured against thresholds we agree with you. The findings go back into prompts, terminology and data, so quality improves over time. What happens to our data? Data protection and confidentiality are part of every risk assessment. With AILT, processing runs in your own cloud tenant, so your content and language data stay with you. ### Working with us Do we have to use AILT? No. We work with the tools you already have and recommend what fits. AILT is one option, for when you want an AI-native solution that runs out of the box. What does the free consultation cover? Your content, languages and goals, and where AI can take over in your process and what that saves. It is a first conversation, with no obligation. Can you work with our existing vendors and teams? Yes. We fit into your setup: your translation management system, your language service providers and your in-house reviewers. We add the governance around them. Contact us ## Where does your AI localization stand? Tell us about your content, your languages and your AI setup. We’ll show you where governance makes the difference. [Talk to us](/contact/) ## Where does your AI localization stand? Tell us about your content, your languages and your AI setup. We’ll show you where governance makes the difference. [Talk to us](/contact/)