2025-03-05. Amar Ujala’s AI chatbot Sarathi, launched during for the Maha Kumbh (a major Hindu pilgrimage in India) to assist pilgrims, shows how AI experiments can create products that address audience needs and enhance engagement. The post How India’s Amar Ujala used an AI chatbot to boost user experience at Maha Kumbh appeared first on WAN-IFRA.
Amar Ujala, a leading publisher in India, launched the chatbot during the Maha Kumbh of 2025 to assist millions of pilgrims attending the grand religious event.
Jaideep Karnik, the Head of Digital at Amar Ujala, shared how the project helped address the challenges faced by attendees at the Kumbh Mela during our recent Bangalore AI Forum.
Launched as part of Amar Ujala’s broader strategy to cater to the diverse needs of the Maha Kumbh attendees, Sarathi served as a digital assistant, answering common queries about the event, such as transportation options, important dates, sanitation issues, and safety concerns.
Karnik said that Sarathi recognised the necessity of offering real-time information during such a massive religious congregation. The Maha Kumbh attracted more than 550 million pilgrims during the 37-day event.
“We were brainstorming and looking for innovative solutions to assist the pilgrims. Sarathi is a result of that exercise, developed to provide an answer to the questions that pilgrims had about the event,” Karnik stated.
Sarathi’s user count increased by over 200 times
Developed in just two months, Sarathi proved to be a massive success, exceeding all expectations in terms of user engagement and traffic, Karnik said.
Launched on 11 December 2024, and within one month of its debut, the chatbot’s user count increased by over 200 times, according to Karnik.
Amar Ujala has a total daily readership of 47 million and 70 million monthly users across its 22 editions in six states and two Union Territories. Digitally, it receives 361 million monthly page views.
“Sarathi’s impact was felt immediately as users engaged with it for information, navigating the vast Kumbh area and seeking updates on real-time events,” Karnik said.
The chatbot was built using OpenAI’s language models (LLMs) and integrated with the Retrieval-Augmented Generation (RAG) technology, which helped Sarathi handle both pre-existing data and real-time updates efficiently.
Karnik shared that a key part of the chatbot’s success was the use of government-provided data.
“We restricted the information to government sources and the content generated by our own team. This ensured that we provided accurate and trustworthy information,” he said.
A dedicated team was tasked with monitoring the queries, ensuring that the chatbot only provided data backed by reliable sources.
Initially, manual monitoring was required, but over time, the system evolved to automatically pull data and provide real-time updates.
Queries grew exponentially
Karnik said that the chatbot was designed to resolve issues in real-time.
“Our team was mindful of the overcrowding during the event and wanted to ensure that users could easily access the most pertinent information through the chatbot,” he said.
“Once it started, the number of queries we received grew exponentially, and we had to scale up our efforts to meet that demand,” he added.
Questions about religion, spirituality, and even marriage or job-related queries were received. In response, Sarathi’s team decided to keep the focus on the core event.
“We didn’t want to misguide people. For irrelevant queries, Sarathi would politely say, ‘Sorry, I do not have an answer for that right now,’” Karnik said, acknowledging that the chatbot still faced challenges in interpreting some complex or out-of-context questions.
“We didn’t monetise Sarathi with sponsored links or ads, as it was still an experiment, and we wanted to focus on providing accurate answers to millions of users without compromising the user experience,” he said.
After the Mauni Amavasya stampede on 29 January, the number of queries increased dramatically, and “we saw more return users, which was a positive outcome”.
Sarathi could be expanded to other languages
One of the challenges with Sarathi was the language barrier. While many AI-powered chatbots are primarily designed for English, Sarathi was tailored for Hindi users, given the demographic of the Kumbh pilgrims.
Sarathi was first deployed in Hindi and then in English. The focus on Hindi was crucial, as it allowed the team to cater to the largest segment of the audience.
“We used OpenAI’s LLMs and incorporated a variety of plugins to handle the translation and generate responses. The models worked well in Hindi, and we’ve been experimenting with expanding our language capabilities.”
Srishti: Amar Ujala’s AI anchor
In addition to Sarathi, Amar Ujala also introduced Srishti, an AI anchor launched in October 2022.
While not widely promoted, Srishti was an experiment aimed at creating content more efficiently. With the help of AI, the team generated video content quickly, increasing viewership while cutting down costs.
Karnik explained that Srishti was designed to produce videos that could be shared on various platforms.
Despite the challenge of differentiating AI-generated content from human-created content, Srishti garnered 80,000 views on a video created during its trial period, a number that exceeded expectations.
“The AI anchor helped us create content more efficiently. For example, when there were no producers available, just one producer could handle the task of creating videos, resulting in more content being produced,” Karnik said.
“Our reach on video content grew, and we saw good engagement, even with AI-generated videos,” he added.
‘AI isn’t about replacing jobs’
Both Sarathi and Srishti highlight how Amar Ujala leveraged AI to improve user experiences and streamline production processes.
The results of Sarathi and Srishti showcase how these tools are helping media organisations scale their operations, according to Karnik.
“AI isn’t about replacing jobs; it’s about improving efficiency and expanding what we can do. AI allows us to streamline processes and meet the demands of millions of people in a cost-effective way,” Karnik concluded.
The post How India’s Amar Ujala used an AI chatbot to boost user experience at Maha Kumbh appeared first on WAN-IFRA.
