Adalat AI makes India’s court system accessible to millions with Claude

Adalat AI is a legal-tech nonprofit building justice technology for courts across India and the Global South, supporting judges and court staff with real-time multilingual speech-to-text and  AI-powered translation with Adalat’s models. With Claude, Adalat AI accelerated its engineering capacity and launched India’s first AI-powered national case information helpline on WhatsApp—giving ordinary families the ability to understand their own court cases, in their own language, for the first time.

With Claude, Adalat AI accomplished:

• 50–200% productivity gains across engineering teams with Claude Code
• Production-ready products deployed to real users in under 30 days, built by a single engineer using Claude Code
• Claude API intelligence layer implemented in 2 days; first working prototype in 1 week
• Bug resolution dropped from half-day investigations to under two hours
• National WhatsApp helpline provides multilingual legal information across the country

The challenge: A justice system overwhelmed by delay

India has more than 50 million pending court cases. The average case takes more than 12 years to resolve. At the current pace, clearing the backlog could take centuries.

“India’s court backlog isn’t an abstract statistic, it shapes everyday life for millions of families,” said Adalat AI co-founder and CEO Utkarsh Saxena. “With over 50 million pending cases and average resolution times exceeding 12 years, justice often arrives too late to matter.”

The consequences are personal: people awaiting trial spending years in jail without conviction, farmers unable to reclaim seized land, and workers waiting for unpaid wages. With 76% of India’s prison population consisting of people awaiting trial, the burden falls hardest on the poor and marginalized. India has roughly 20 judges per million people, compared to 150–200 in developed countries. “Even saving 30 minutes a day per judge, multiplied across 4,000+ courtrooms, is equivalent to adding hundreds of judges to the system without hiring a single one,” said Saxena.

Beyond the courtrooms, litigants face a deep information access crisis. Most court portals have non-intuitive navigation, multiple CAPTCHAs, and content overwhelmingly in English—effectively unusable for low-income or non-English-speaking communities. Families often have no idea what is happening in their own cases.

Choosing Claude for privacy standards 

Adalat AI evaluated multiple models before selecting Claude Sonnet 4.5 and Haiku 4.5. “Claude consistently produced summaries that were both faithful to the source and genuinely readable for non-lawyers,” said Arghya Bhattacharya, Co-founder and CTO. “Other models either oversimplified, losing critical nuance, or reproduced the complexity without making it accessible.” Claude’s disciplined reasoning—staying within the document without speculating or providing legal advice—was noticeably stronger than alternatives. 

“In our world, institutional trust compounds slowly and collapses quickly,” Bhattacharya said. “We chose Claude in part because Anthropic’s approach to data handling and privacy aligned with the trust standards we need to maintain with our government and judicial partners.”

The solution: Claude Code for engineering speed, Claude’s API for accessible justice

Adalat AI turned to Claude on two fronts: Claude Code to accelerate internal engineering, and the Claude API to power a national WhatsApp helpline that puts court case information directly into litigants' hands.

A national helpline on WhatsApp. The WhatsApp chatbot, built on the Claude API, is India’s first AI-powered national case information helpline. Anyone in the country can search for their court case, check its status, review case history, see the next hearing date, and download orders—through a platform more than 500 million people in India already use daily.

Claude’s critical role is summarization and translation. Court orders are often lengthy, written in formal English legal jargon, and impenetrable to non-lawyers. A perfectly accurate English summary doesn’t help a Kannada-speaking farmer. Claude summarizes documents into clear, actionable takeaways and translates them into the user’s chosen language—currently English, Hindi, Kannada, and Malayalam.

“When you take a 15-page court order written in English legal jargon, and Claude summarizes it into three clear paragraphs in Kannada or Hindi explaining what was decided, what the litigant needs to do, and when their next hearing is, you’re transforming court records from inaccessible legal artifacts into actionable information,” Saxena said. “For the first time, a litigant can understand their own case without depending entirely on their lawyer or a paid intermediary.”

Accelerating engineering with Claude Code. The core engineering team uses Claude Code for architecture validation, feature development, rapid bug resolution, and an additional layer of code review. Adalat’s Zero-to-One Initiatives team—a rapid prototyping arm—staffs each initiative with a single engineer, with Claude Code as their force multiplier. The team runs 3 to 4 parallel initiatives simultaneously, each delivering fully functional products within 30 days. 

“Not mockups or wireframes, actual working products that we put in front of court users to get real feedback,” Bhattacharya said. “This means we skip the intermediate step of showing designs. Our users can play around with a living product and give us concrete, grounded feedback.”

The results: Faster engineering, accessible justice at national scale

With Claude Code, Adalat AI's engineering teams are seeing between 50 and 200% productivity gains. "For a philanthropy-funded nonprofit, that's an extraordinary multiplier on donor dollars,” Bhattacharya said.. 

The rapid prototyping team, which is shipping multiple new product lines simultaneously, is seeing gains of up to 200%, moving at the pace of a team three times its size. The core platform and engineering teams, which maintain and actively develop the product running in 20% of all courts in India every day, are seeing around 50% gains, consistently delivering output equivalent to a team 50% larger.

"Claude Code has been an absolute revolution internally,” says Bhattacharya. “As a nonprofit you're constrained in how many engineers you can hire, and the fact that you can boost productivity per engineer versus hiring more engineers is a huge resource optimization."

Combined with test-driven development practices, the result is a more robust codebase with fewer issues reaching production. Bug resolution dropped from half-day investigations to under two hours. The Claude API integration for the chatbot was implemented in 2 days, with the first working prototype operational within 1 week.

“The biggest implementation challenge wasn’t actually Claude,” Bhattacharya said. “It was the surrounding infrastructure: integrating with India’s fragmented court data systems, handling the WhatsApp Business API, managing multilingual flows, and building the case identification layer that sits upstream of the LLM. Claude was the part that ‘just worked.’”

Early user feedback has been strong. “They’re very surprised that this is accessible on WhatsApp,” said Saxena. “The idea that they can chat on WhatsApp and get the actual document and then also get it translated and summarized has been transformative.”

Looking ahead, Adalat AI plans to expand language support to 10 or more Indian languages and deepen integrations with state court systems. The team is already in conversations about expansion to Ghana, Kenya, and Zambia. “If the chatbot proves itself in India, the model is replicable across the Global South,” Saxena said. “The Anthropic partnership could scale with us into common-law jurisdictions serving over two billion people.”

‍