Navigating Language Barriers with AI: A Revolution in Delivery Services

Vineet Sawant, a delivery driver in Mumbai, shares his experience of overcoming language challenges through AI-enabled solutions, highlighting the need for better multilingual support across various sectors in India.

Vineet Sawant, a delivery driver in Mumbai, has faced significant challenges over the past two years—not from traffic, but from language barriers. Navigating through the bustling city with limited English proficiency, Sawant relies on translation technology to enhance his efficiency with Zepto, an online grocery delivery service. "Being on the road is always stressful, but language complications add another layer of difficulty. My first language is Marathi. At first, everything was in English, and while I could grasp some of it, reading was a challenge," he explains.

To address this issue, Zepto partnered with Reverie Language Technologies and introduced an AI translation service that allows drivers to interact in six languages through the app. "Now, if I see 'ring bell', I get it in Marathi. No more guessing or confusion," he adds, vastly improving his speed and accuracy on deliveries.

India’s rich linguistic diversity—22 official languages and numerous dialects—complicates the mainstream application of AI in the country. Professor Pushpak Bhattacharyya from IIT Mumbai emphasizes that access to high-quality language data is crucial: "Millions are excluded from technological advancements due to the absence of accurate translation tech that can accommodate these languages."

While popular languages like Hindi and English are easier to support due to available training data, regional dialects and tribal languages remain underserved. Vivekananda Pani, co-founder of Reverie, warns that less common dialects risk being marginalized as AI translation tools proliferate.

To combat this, Professor Bhattacharyya is involved in the Bhashini initiative, which aims to develop comprehensive language datasets and translation services for various Indian languages. Launched in 2022, Bhashini has made significant strides, creating 350 AI models that have accomplished over a billion tasks. These tools facilitate multilingual chatbots for public services and translate essential government information into local languages.

Amitabh Nag, CEO of the Digital India division overseeing Bhashini, hopes this technology will empower rural citizens with language-specific access to government services in the near future. Meanwhile, research is underway on AI programs, such as one designed to help smokers quit, focusing on language adaptability to provide tailored advice at different intervention stages.

Currently, initial trials of this health-focused AI are being conducted in English and Hindi, but plans are in place to accommodate all 22 languages as the project advances. As for Vineet Sawant, his productivity has skyrocketed from ten to around thirty deliveries a day, thanks largely to the language features integrated into the Zepto app. "When the app speaks our language, we feel more confident, and we work better," he concludes, embodying the transformative potential of AI for diverse communities in India.