Skip to content
Article WAN-IFRA Mar 2026

WAN-IFRA: how The Hindu applied AI to election data and heat stress reporting

What the article is about

This WAN-IFRA case study covers The Hindu’s experience using AI tools for data-intensive journalism. The primary voice is Srinivasan Ramani, the newspaper’s Deputy National Editor for data, who describes how the team used large language models to handle structured, verifiable tasks across two significant reporting projects.

Context: the projects

The first project involved Indian election data at scale: 22 million voter records spread across three states. The team used AI for optical character recognition on scanned documents, for generating SQL queries to interrogate the dataset, and for scaffolding the code needed to build an interactive web application for readers. Ramani describes the result as an interactive election tool assembled without traditional manual coding — the AI generated functional code from natural language descriptions of what the application needed to do.

The second project was a heat stress study. The team applied similar methods — AI-assisted data extraction and analysis — to climate and health data, again focusing on tasks where the AI’s output could be checked against known values or structured sources.

Key takeaway

Ramani’s framing is precise: he describes LLMs as a “sophisticated intern” — capable of handling discrete, bounded tasks when given clear instructions, but requiring editorial oversight on every output before it appears in print. The cases he describes are ones where the AI’s work is verifiable: a SQL query either returns the right records or it doesn’t, and OCR output can be checked against source documents. He does not describe using AI for interpretive or analytical writing.

Who it is useful for

Data journalists and editorial teams in newsrooms that are evaluating AI for structured, verifiable tasks such as document digitization, database querying, and interactive application development. The case is also relevant for editors and technology leads designing AI policies: the conditional framing — AI for the verifiable layer, journalists for editorial judgment — provides a concrete model for scoping AI use without foreclosing it.