Newslaundry’s AI chatbot is doing more than answering queries

"Ask Birubala," Newslaundry's in-house AI chatbot, was designed to manage subscriber queries, troubleshooting and documentation.

Its deeper impact lies in upskilling staff: by automating repetitive tasks, it has enabled the support team to move into more complex work such as data analysis and campaign management, creating greater long-term value for the subscription-first organisation.

For Chitranshu Tewari - who led the project and, until recently served as Newslaundry's Director of Product and Revenue - the tool is both a statement of intent and a practical solution. Though he has since moved on, it reflects his commitment to building products that serve real user needs rather than chasing trends or ad revenue.

Over its 13-year journey, Newslaundry has built itself as a fully ad-free, subscription-funded organisation with a global community of thousands of paying readers.

Tewari led Newslaundry's product and revenue team for nine years, overseeing the design and scale of products ranging from custom podcast players and bundle subscriptions to AI-powered tools like Ask Birubala.

A reader-first approach to product

"Our journey has been slow, but it has entirely been powered by subscription," said Tewari. "We are confident about it, and the conviction runs across the team, whether in editorial, product, or business."

The company's philosophy rejects vanity metrics. Instead of optimising for clicks or impressions, its dashboards measure engagement, loyalty, and which stories subscribers spend the most time on.

Privacy is also a guiding principle.

The Newslaundry app allows users to control notifications and manage or cancel subscriptions with ease.

Tewari pointed out that unlike traditional media companies, Newslaundry avoids conflicts of interest between editorial and sales.

"Everyone has the same objective, which is to increase the number of paying subscribers," he said.

That clarity shapes product decisions, from simple subscription management tools to long-term investments in AI.

Mobile-first, user-focused products

For Tewari, product success is about solving reader pain points, not flashy design.

"Newsrooms make the worst product decisions," he said. "Often, energy is spent on animations or fancy layouts that don't even load on mobile. But 95 percent of our traffic is mobile. A reader who wants to finish a 15,000-word story doesn't care about fancy graphics - what matters is a smooth reading experience."

Newslaundry's custom podcast player is one example.

When its podcasts moved behind the paywall, listening became a poor experience on existing platforms. The in-house player restored user control with queue management, playback speed, and show notes, encouraging subscribers to stay engaged.

Similarly, its mobile app prioritises user choice, from subscription upgrades to notification settings, reinforcing the company's belief that good products build trust.

Ask Birubala: an AI assistant

Unlike some AI tools in newsrooms focused on experimentation, Ask Birubala addresses practical operational challenges.

"Too often, AI tools in newsrooms are gimmicks. They don't solve real problems," Tewari said.

Ask Birubala, by contrast, is a retrieval-augmented generation (RAG) tool that addresses the everyday challenges of a lean subscription-tech team.

The chatbot functions as a troubleshooting assistant for subscriber issues, a knowledge assistant for onboarding staff, and a documentation shortcut.

By training it on historical support emails, FAQs, and developer notes, Newslaundry has created a single source of truth for subscription workflows and technical processes.

Paste a subscriber's email into the tool, and Ask Birubala generates a clear, context-rich reply explaining what went wrong and what steps are required.

It can also answer internal questions such as how to map bulk subscriptions or troubleshoot failed payments. The tool is continuously refined with live knowledge base edits and user feedback.

Building and iterating Ask Birubala

Ask Birubala was built using Next.js, React, Node.js, Langchain, OpenAI, and Pinecone. Internal documentation is created with Scribehow, which turns screen recordings into PDFs.

These are split into chunks, vectorised, and stored in Pinecone with metadata tags such as topic or platform. When a query is entered, the system retrieves the most relevant chunks and generates a cited response.

Tewari admitted the process was not easy. "We didn't have detailed documentation, and many workflows were undocumented or manual. We had to learn AI concepts from scratch and keep updating as our tech stack evolved."

Iterations included adding metadata classification, enabling edits from within the interface, and introducing a feedback mechanism to rate answers.

Measurable impact on productivity

The results have been tangible. Ask Birubala now saves more than 10 hours a week for senior developers and product managers who previously handled repetitive queries. Support staff have shifted towards data analysis and campaign management, improving productivity and job satisfaction.

"This was about solving real problems, not chasing hype," Tewari said. "Our goal was to reduce the load on our team and give consistent, accurate answers. It's about building tools that fit the culture and needs of our newsroom."

The team behind Ask Birubala was a small cross-functional group of three people.

Neha Gupta

Multimedia Journalist

[email protected]