Bengaluru-based Blue Machines AI has launched FLOE, a new evaluation framework designed to measure how effectively voice AI systems detect and respond when users switch between languages during conversations.
The framework targets code-switching, a common feature of multilingual conversations in markets such as India, where speakers may move between languages within the same interaction. Blue Machines AI said FLOE is intended to provide a structured way to assess whether voice systems can recognise such changes accurately and respond in the appropriate language.
The development addresses a technical challenge for conversational AI platforms as enterprises deploy voice agents across customer service, financial services and other high-volume communication workflows. A voice system that fails to recognise a language change can misinterpret user intent or continue responding in the wrong language.
Blue Machines AI already positions multilingual interaction and real-time language switching as part of its enterprise voice AI offering. Its platform combines speech-to-text, text-to-speech and large language models to handle conversations and execute tasks within enterprise workflows.
FLOE adds an evaluation layer to this approach by focusing specifically on how models handle language transitions rather than assessing only general speech recognition performance.
The launch is relevant for the Indian market, where conversations can involve English alongside Hindi and regional languages, sometimes within the same sentence or exchange. This makes multilingual performance dependent not only on whether an AI system supports multiple languages, but also on how quickly and accurately it identifies changes while a conversation is underway.
Blue Machines AI's existing enterprise platform is designed for multilingual voice interactions across sectors including lending, healthcare, insurance, recruitment, mutual funds and education. The company says its technology can deliver sub-300 millisecond latency while incorporating governance mechanisms for privacy, compliance and security. These performance figures are company-reported.
The company has also been deploying conversational AI within regulated industries. Earlier this year, Blue Machines AI announced a deployment with Aditya Birla Capital across businesses including lending, health insurance and housing finance. The implementation involved voice agents being connected with operational systems for customer engagement and workflow automation.
As enterprise voice AI adoption expands, evaluation is becoming an important part of deployment. Systems operating in customer-facing environments must account for accents, interruptions, colloquial expressions and multilingual speech alongside requirements around latency and accuracy.
FLOE represents Blue Machines AI's attempt to create a more targeted benchmark for one of those challenges. The framework focuses on whether AI systems can follow the linguistic patterns of real conversations, particularly when users move naturally between languages rather than remaining within a single-language interaction.
For enterprises deploying conversational AI in multilingual markets, such evaluation could provide another measure for assessing voice agents before and during production use.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.