One prompt doesn’t fit all: how RWS improved LLM translation quality for a global tech company
A multinational technology company delivering enterprise software, cloud services and digital tools to organizations worldwide needed to ensure high-quality, domain-specific outputs from large language models (LLMs).
The client needed a partner capable of optimizing LLM outputs across multiple content domains, including IT, finance, marketing and legal, while meeting the high standards of technical accuracy and contextual relevance required by enterprise and professional users.
High-quality, domain-specific outputs
Reduced manual post-editing
Scalable, multilingual workflows
High-quality, domain-specific outputs
Reduced manual post-editing
Scalable, multilingual workflows
Key benefits
Higher-quality, domain-accurate outputs across IT, financial, marketing and legal content
Significant reduction in manual post-editing effort
Increased consistency and scalability in multilingual workflows
Faster turnaround and adaptability for new tasks and content types
Cost-effective, high-value output without requiring in-house AI infrastructure
The challenge: meeting complex LLM needs
The client needed to produce high-quality LLM outputs across diverse content domains, each with distinct requirements for terminology, tone, accuracy and compliance. IT content demanded technical precision and alignment with product-specific glossaries. Financial content required clarity and adherence to regulatory standards. Marketing needed creativity and audience engagement, while legal demanded rigorous accuracy.
When generic or mismatched prompts were used, the results were either too rigid or misaligned with the intended audience. This created a burdensome need for manual post-editing, which was unrealistic at large volumes.
“The client introduced new categories for translation, one of which involved using LLMs and AI to assist and support human post-editing,” explains Program Director Ivan Lukavsky. “However, the prompts we developed for IT content didn't work well for content with different syntax and requirements. That's when the team started exploring ways to improve output for non-IT content.”
The client also had tight deadlines and high-volume content demands. They needed scalable solutions that could deliver consistent, reliable output across multiple languages while balancing technical precision, linguistic quality and creative flexibility. The goal was to ensure every content type met business and user expectations without slowing production.
The project required:
Domain-specific prompt engineering to ensure accurate outputs
Improved post-editing rate across content types
Consistency and scalability in multilingual workflows
The right balance of creativity and precision for each content domain
The solution: tailored prompts and expert workflows
For this project, the RWS account team – with the support of Linguistic AI Services (LAIS) – designed prompts tailored to each content type and task, ensuring outputs were contextually appropriate and aligned with the client's standards. To support scale and quality, RWS mobilized its global network of language specialists to staff an initial team and rapidly retrained them in AI quality control and prompt engineering to enable fast, cost-effective and high-quality delivery.
This included human-in-the-loop requirements. RWS delivered a flexible, human-centered system designed to evolve alongside the project. Key aspects of the solution included:
Domain and task-specific prompt engineering: Prompts were tailored not only to content types but also to specific tasks. For example, marketing prompts encouraged creativity, while IT prompts emphasized technical accuracy.
Integration of linguistic expertise: Human reviewers validated every output, applying style guides and contextual insights to maintain accuracy, consistency and relevance across all domains.
Rapid, scalable processes: The system was built to adapt quickly, allowing new features, prompts and workflows to be added without disrupting production. Monthly innovation cycles and continuous monitoring ensured high-quality, reliable results.
Cost-efficiency: By leveraging existing networks and building an AI sandbox from scratch, RWS minimized manual post-editing while delivering production-quality outputs at scale, worldwide
Challenges
Ensuring high-quality LLM outputs across diverse content domains
Reducing the burden of manual post-editing at scale
Achieving consistency and scalability in multilingual AI workflows
Business Outcomes
Produced domain-accurate, high-quality LLM outputs
Reduced manual post-editing and faster delivery
Created scalable, repeatable multilingual AI workflows
Results
High-quality, domain-specific LLM outputs
Reduced post-editing effort
Scalable multilingual content workflows
Faster, more reliable content production
Increased value without in-house AI infrastructure
Technical expertise: AI and domain knowledge in action
RWS brought a level of technical expertise uniquely suited to this project. The team quickly aligned with the client's pioneering standards, mastering AI-supported workflows, prompt engineering and production-quality post-editing in real time. We built custom systems to manage multiple tasks and content domains, ensuring outputs consistently met the client's rigorous expectations.
This expertise allowed RWS to handle complex LLM workflows across IT, financial, marketing and legal content, translating highly technical guidance into actionable prompts while maintaining contextual accuracy. The team also implemented monitoring systems to track every output. Continuous oversight and feedback safeguarded quality and consistency as the project evolved.
Tailored, domain-specific prompt engineering
Human-in-the-loop quality assurance
Scalable, flexible AI-supported workflows
Exceptional talent: skilled teams driving results
RWS drew on its longstanding network of language specialists, rapidly retraining them to take on AI-focused roles. “We had to start from scratch and learn along the way. We developed our own system and process to be able to meet the client needs,” says Valerie Dokic Nemeckova, Project Manager.
Linguistic experts transitioned into AI workflow specialists, while other team members expanded into technical and engineering responsibilities. This flexibility enabled RWS to deliver high-quality results from the start, without the need to recruit new staff.
Every output undergoes review by a human expert before delivery, a critical human-in-the-loop safeguard. The team's adaptability and skill made it possible to implement new prompts, agents and features as the project evolved. This exceptional talent gives the client confidence that complex, high-stakes content can be produced reliably and efficiently, supporting ongoing innovation and protecting their global reputation.
Results: delivering high-quality, cost-effective and domain-specific outputs
RWS's approach provided a reliable, cost-effective solution for producing high-volume, domain-specific LLM outputs. By combining task-specific prompt engineering, human review and AI-supported workflows, RWS consistently met the client's rigorous standards across multiple content domains while minimizing the need for manual post-editing.
The system also allowed the team to adapt quickly to new tasks and content types, ensuring the client could continue innovating without delays.
“We’ve developed a cost-effective, flexible solution, easy to localize for different services and deploy new functionalities,” says Solution Architect Michal Sedivy.
Key results included:
Outputs consistently matched the terminology, tone and quality expectations for IT, financial, marketing and legal content.
Team members retrained quickly to handle new tasks and content types without slowing delivery.
Prompts generated engaging, contextually relevant content while maintaining clarity.
A repeatable framework supported ongoing content production, new functionalities and multilingual workflows.
High-quality outputs were achieved without requiring in-house AI infrastructure, delivering substantial value for the client.
RWS continues to implement new prompts, features and AI agents, with every output vetted by human experts before release, supporting the client's ongoing innovation needs. This ensures reliability for a client with a global reputation to maintain.
Even more importantly, RWS continues to serve as a trusted partner in innovation and development. This long-term relationship has evolved alongside changing technology and customer expectations, enabling us to help this client deliver high-quality results across a wide range of outputs over the years
Every AI project must include human review. A human checks every output to catch issues—including hallucinations—before it reaches end users.
Ivan Lukavsky
Program Director at RWS
A trusted partner achieving accuracy, creativity and efficiency at scale
RWS's expertise in linguistic services and AI-assisted workflows enabled the client to produce high-quality, domain-specific LLM outputs across IT, financial, marketing and legal content. By combining tailored prompt engineering, rapid team retraining, human review and scalable processes, RWS delivered accurate, contextually relevant and cost-effective results that reduced post-editing and supported high-volume content production.
The team's technical expertise and exceptional talent ensured that every output met the client's rigorous standards. At the same time, flexible workflows allowed the project to evolve with new prompts, tasks and AI features. Today, the client continues to rely on RWS for ongoing innovation, confident that each deliverable will maintain quality, efficiency and global reliability.
Discover what RWS can do to help your organization achieve the same. Our experts enable you to optimize LLM outputs and streamline multilingual content workflows at scale.
IndustryTechnology
Founded1975
The client: A Fortune 100 technology leader specialized in enterprise technology services
