AI Schliemann Launches “Multi-Translation Agent” to Automatically Select Multiple Flagship AI Engines by Text and Language

Sto m Inc. (headquartered in Koto-ku, Tokyo; CEO: Akira Toyama), the developer and provider of the website multilingualization SaaS “AI Schliemann,” announced today the release of its “Multi-Translation Agent” feature. The new function automatically switches among multiple flagship AI engines depending on the content to be translated and the target language. The feature combines three latest flagship AI engines, GPT-5.5, Claude Opus 4.7, and Gemi i 3.1 P o, and automatically selects the optimal translation process based on the characteristics of the text, including IR, legal, technical documents, brand messaging, recruitment pages, product descriptions, and web content containing HTML. This allows AI Schliemann to move beyond reliance on a single translation engine and deliver more stable website translation quality by leveraging each AI model’s strengths. In corporate website multilingualization, the required level of translation quality is becoming more sophisticated every year. Translation is no longer just a matter of replacing Japanese with English or Chinese. It must also preserve brand tone, technical terminology, industry context, IR expressions, the natural feel expected of recruitment sites, and HTML structure. At the same time, websites often contain many different types of text. A single corporate site may include IR news and earnings presentation materials where accuracy is the top priority, CEO messages and brand stories where tone and readability matter, product descriptions and recruitment pages that require natural and accessible wording, technical specifications and help pages where terminology and logical structure are important, and web-specific content containing HTML tags, links, and button labels. It is difficult for a single AI model to translate all of these optimally at all times. The Multi-Translation Agent automatically selects the most suitable AI engine from GPT-5.5, Claude Opus 4.7, and Gemi i 3.1 P o based on the type of text, writing style, level of specialization, density of proper nouns, target language, and HTML structure. Its process includes analyzing the source text, determining text type and expertise, checking compatibility with the target language, selecting the AI engine, executing translation, and verifying terminology, numbers, and HTML structure. GPT-5.5 is used for texts requiring accuracy, such as IR, legal, and technical documents, including IR news, earnings presentation materials, integrated reports, corporate histories, contracts, legal documents, technical specifications, SaaS help pages, and security-related materials. Claude Opus 4.7 is used for content that requires natural readability and expressive tone, such as brand copy, CEO messages, sustainability pages, narrative sections of integrated reports, tourism and cultural introductions, customer stories, PR copy, and literary writing. Gemi i 3.1 P o is used for short to medium-length web content such as corporate site descriptions, product and service introductions, recruitment sites, company profiles, UI copy, general web content, and marketing-oriented descriptions, helping restructure heavy Japanese syntax into clear and readable target-language text. The feature also supports engine selection based on the target language. In addition to Japanese-to-English translation, it considers the characteristics of Chinese, Korean, European languages, and Southeast Asian languages. It is designed to better handle fine-grained language differences, such as Taiwan- and Hong Kong-oriented expressions in Traditional Chinese, honorifics and corporate style in Korean, and politeness levels and loanword handling in Southeast Asian languages. AI Schliemann states that translation data processed by the service will continue not to be used for training any LLMs. Although the new Multi-Translation Agent uses multiple AI engines, customer website content and text submitted for translation will not be used by each LLM for training purposes. Akira Toyama, CEO of Sto m Inc., said that while the translation quality of generative AI has improved significantly, practical website translation cannot be solved simply by choosing one high-performance AI. IR documents, brand messages, recruitment pages, technical documents, and web content containing HTML each require different translation qualities. The newly released Multi-Translation Agent automatically uses three flagship AI engines according to the text and language, aiming to further stabilize translation quality for corporate website multilingualization. The Multi-Translation Agent feature is available in AI Schliemann’s new translation function. Details on eligible plans, implementation methods, and pricing are available via the company’s official inquiry page. Going forward, AI Schliemann plans to further improve the precision of engine selection by text type and target language, strengthen integration with company-specific terminology, brand tone, and translation history, and build a higher-quality multilingual website infrastructure for practical business use.