Babeltext expands translation platform to 249 languages

Babeltext has expanded its digital translation platform to support 249 languages and is using the milestone to challenge government policy on automated multilingual services.

The Sydney-based company argues that official thinking still treats all machine translation as if it were an uncontrolled public tool, rather than a supervised service system with human oversight. It says the distinction matters more as public agencies handle rising volumes of routine enquiries across digital channels.

Babeltext's platform is designed to let organisations keep their existing service teams while adding multilingual support across SMS, WhatsApp, web chat, WeChat, Messenger, Telegram, Instagram and Discord. Customers can send messages in their preferred language, with the system translating the exchange, routing the enquiry and passing routine questions to an approved AI agent before escalating more sensitive or complex matters to a human operator.

Staff continue to work in English while responses are translated back to customers in their chosen language, according to the company. Conversation history, reporting and audit logs remain available throughout the exchange within an AWS-hosted environment.

David Hayes, founder of Babeltext, said public sector policy had not kept pace with technical change. "Too much of government still treats every use of computer-driven translation as though it were an uncontrolled public translation website," he said.

He drew a distinction between broad online translation tools and systems built with defined boundaries and oversight. "That is not what Babeltext is. Babeltext is a governed customer-service platform that keeps people in control. The AI deals with approved routine enquiries, while the existing service team takes over whenever human judgement is required. That is the blind spot: Babeltext does not replace the service team. It makes the service team multilingual," Hayes said.

Earlier testing

Babeltext points to an independent review conducted during its early development with the Ethnic Communities' Council of NSW. The assessment examined usability, accessibility, and linguistic and cultural appropriiateness through structured sessions involving 22 bilingual speakers and community participants across seven languages.

Testing used six simulated service scenarios covering emergency response, tourism, hotel services, mental health support and Service NSW enquiries. In the final session, participants reported translation accuracy of more than 98 per cent and full comfort using text-based services such as SMS, web chat and WhatsApp, according to the company.

The review concluded that the service could complement rather than replace professional interpreters and translators when skilled language workers were not immediately available. The Ethnic Communities' Council of NSW described Babeltext as providing "instant and timely language support in the absence of expert language personnel."

Hayes said the product had changed considerably since that work was completed. "The testing demonstrated that an English-speaking operator could communicate effectively with people across multiple languages," he said.

"Since that review, Babeltext has progressed from a multilingual text-messaging service into an integrated platform supporting 249 languages, eight messaging channels, artificial intelligence, human escalation, APIs, reporting and auditability.

"The technology has continued to advance.

"Unfortunately, government thinking has not advanced at the same speed."

Policy debate

The company's argument centres on the gap between existing language services policy and newer forms of digital communication. Australian and New South Wales policy both set out the principle that public services should remain accessible regardless of linguistic background, but current guidance still places strong emphasis on the risks of automated translation in official settings.

That caution reflects long-standing concerns about errors, lack of context and the danger of using general-purpose tools for legal, medical or safety-critical material. Babeltext does not dispute those risks and says certified translators and interpreters remain necessary for court matters, clinical decisions, emergency instructions and other high-risk uses.

Instead, it argues that lower-risk interactions such as service navigation, bookings, complaints, community information and general enquiries should be treated differently. Procurement and policy settings, it says, should distinguish between a free public translation tool, a controlled translation environment, an AI-only service model and a workflow that combines automation with human handover.

It also points to broader shifts in the sector. More recent industry guidance has accepted that AI now forms part of the working toolkit for translators and interpreters, while maintaining that people must remain central to intercultural communication and that high-risk assignments still require expert supervision.

Government pilots

Babeltext is calling on departments, agencies and local councils to run controlled pilots in lower-risk, high-volume service areas. Suggested uses include appointment support, tourism information, community programmes, event information, complaints intake and referrals to human staff.

Such trials could be built around approved information sources, firm limits on what AI may answer, automatic or requested handover to staff, complete conversation records and measurable service outcomes, the company said. It proposes assessing pilots through metrics including multilingual enquiries captured, first-response time, resolution rate, human handover, customer satisfaction and cost per resolved conversation.

Hayes said the debate should move beyond a binary choice between automation and human expertise. "The question is no longer human or machine," he said.

"The question is how government can combine machine speed with human accountability.

"Nobody is suggesting that an AI system should replace a certified interpreter in a courtroom, a hospital or a critical safety situation. But people should not need to find an English-language document, make a telephone call, wait for assistance and repeat their story simply to ask an ordinary service question.

"People already communicate through SMS, WhatsApp and web chat. Government should be prepared to meet them there."

Australia presents a large potential market for multilingual digital services. Census data cited by Babeltext shows that 5.8 million Australians, or 22.8 per cent of the population, speak a language other than English at home.

Hayes framed the issue as one of access rather than replacement. "We are not asking the government to remove professional interpreters, translators or frontline staff," he said.

"We are asking the government to stop forcing too many multilingual interactions through the slowest possible service path.

"Keep the experts for the conversations where their skills are essential. Use governed technology to open the digital front door, and make sure a human is always available when the conversation requires one."