Jhana.ai platform
India
Justice and Governance
Implementing Organisation
Jhana.ai
India, Karnataka, Bengaluru
Implementing Point of Contact
Smita Gupta
Director, GovTech and Public Sector
Contributor of the Impact Story
Carnegie India
Year of implementation
2024
Problem statement
Legal risk accumulates silently across everyday life events such as housing, marriage, inheritance, taxation, and disputes, yet citizens lack any framework for legal wellness. Interaction with the Indian legal system is often opaque, slow, and expert-dependent, with high information asymmetry and limited accountability. Judicial institutions face a different but related challenge: massive scale, fragmented records, frequent roster changes, and loss of institutional memory. Courts process tens of thousands of pages weekly, yet judges and registries remain constrained by cognitive bandwidth rather than intent. Filing scrutiny, case preparation, research, dictation, and publishing remain manual and repetitive, leading to delays, errors, and inconsistent access to justice. Existing digitization efforts often replicate paperwork digitally rather than reducing effort or improving substance. The absence of AI-native legal infrastructure prevents courts from transforming raw filings into structured, searchable, and auditable institutional memory. Without such systems, both citizens and the judiciary continue to bear avoidable legal risk, administrative burden, and delays in justice delivery.
Submission Overview
Jhana is India’s leading legal AI lab, founded in 2022 at Harvard University, $1.7M seed-funded, and serving over 15,000 monthly active lawyers, judges, and principals. Courtroom by jhana is our pioneering judiciary practice that automates route work across the case lifecycle and brings powerful decision support tools for judges and listing registrars. AI handles filing scrutiny, metadata extraction, docket prep, and more, eliminating critical back office delays between filing, numbering, listing, and disposal. Courtroom is active in pilot or at scale at 5 courts including most notably the High Courts of Madras and Karnataka. PUBSEC by jhana is our public sector practice, built atop our operational fluency, understanding of sensitive government systems and discordant staff incentives, and expertise in change management. PUBSEC and Courtroom are designed as self-owned, composable API blocks where the state retains its data sovereignty while benefitting from online inference on jhana. Our APIs power or talk to a variety of MIS and DMS systems, and we have developed proficiency in the patching and scripting work that it can now rapidly take a legacy government system live.. jhana’s AI tools for researching case law and procedure have over 150 judges and 200 government staff among voluntary users who signed up and even subscribed on their own accord. Our proprietary National Legal Archive of over 16 million Indian legal documents is updated daily and structured to engineer tools that come with verifiable citations. jhana follows a strict responsible-AI framework emphasizing human-in-the-loop validation, no outcome prediction, court-owned data, and rigorous footnotes for explainable audit trails. Backed by leading technology investors and recognized across Asia for legal innovation, jhana combines deep legal expertise, advanced AI research, and on-ground operational fluency to improve access to justice, judicial efficiency, and legal certainty at population scale.
AI Technology Used
Generative AI
Key Outcomes
Efficiency & Productivity
Access & Reach
Accuracy & Quality Improvement
User Experience & Satisfaction
Knowledge & Skills Impact
Access to justice in India is often delayed due to fragmented records, and because filings, case indexing, and documentation consume significant time. Jhana.ai's legal infrastructure tools have compressed filing scrutiny from hours to minutes and indexing from days to a few minutes. The platform processes thousands of pages weekly for judicial note-taking and has been organically adopted by over 150 judges, which has led to improvements in workflow efficiencies and faster case resolution.
Impact Metrics
Average time taken to complete filing scrutiny checks on commercial e-filings
Baseline Value
1.5 hours was the turnaround time before the AI intervention
Post-Implementation
The turnaround time to complete filings came down to 3 minutes Minutes/ hours
Turn-around time from e-filing to indexing/processing completion
Baseline Value
2 -10 days turnaround time
Post-Implementation
Less than 15 minutes turnaround time Minutes
Time taken to convert published cause-list into processed, populated judge dashboards
Baseline Value
1 -10 days for conversion
Post-Implementation
Less than 15 minutes for conversion Minutes/ days
Volume of pages processed per week for judicial note-taking and preparation
Baseline Value
Workload used to be fragmented due to manual reading Pages per week
Post-Implementation
Over 10,000 pages per week processed Pages per week
Number of verified judge users adopting AI dictation tool organically.
Baseline Value
NA verified judge users
Post-Implementation
Over 150 verified judge users Number of users Reported Peiord - End: 26/11/2025
Net Promoter Score / satisfaction rating from pilot evaluation team benchmarking headnotes
Baseline Value
NA Score (out of 5) / NPS Percentage
Post-Implementation
100 % NPS (5 out of 5)
Number of verified legal professionals actively using AI legal research and drafting tool
Baseline Value
NA verified legal professionals
Post-Implementation
10 ,535 verified judge users
Implementation Context
Supreme Court of India, High Courts of Karnataka, Madras (Tamil Nadu), Telangana, Kerala, ITAT, district courts, arbitration bodies
Judges, registrars, court staff, government lawyers, arbitrators, mediators, lawyers and indirectly litigants and citizens across urban and semi-urban India, including underserved court users facing delays and information barriers
Key Partnerships
Government and Judicial Partnerships
Replicability & Adaptation
1. The use case demonstrates strong replication potential across Indian states and comparable common-law jurisdictions, provided procedural, linguistic, and institutional adaptations are undertaken 2. The core AI architecture is court-agnostic and composable, but judicial systems are rule-dense, locally governed, and operationally idiosyncratic, which necessitates structured adaptation rather than plug-and-play deployment 3. Replication is most feasible in environments where Courts already use CIS/DMS or e-filing systems (even if fragmented) 4. There is institutional willingness to pilot AI under controlled governance (LoI/ MoU / MLE) 5. Registries and judges face scale constraints similar to Indian High Courts (volume of filings, multilingual records, frequent roster changes) 6. The platform has already demonstrated transferability across multiple High Courts, tribunals, and ADR/ODR institutions, each with distinct procedural rules, validation standards, and stakeholder expectations—indicating that replication is operationally proven but not trivial 7. Replication requires structured localization, not model retraining in the abstract 8. Court-specific rules need to be encoded such as Filing scrutiny checklists, Court fee schedules, Jurisdictional requirements, Defect typologies and cure workflows 9. Alignment with local cause-listing, roster allocation, and lifecycle management practices is required 10. Support for local languages and scripts used in filings, orders, and dictation must be provided 11. Region-specific document formats, stamps, seals, and legacy scanning practices need to be handled 12. Custom KPI frameworks must be aligned with the court's priorities (speed, accuracy, adoption, neutrality) 13. Human-in-the-loop validation standards should reflect judicial comfort levels with AI assistance 14. Clear guardrails must exclude outcome prediction and preserve judicial discretion 15. Hands-on training programs for judges, registries, and clerks are essential 16. Gradual workflow augmentation should be implemented rather than forced replacement of existing processes 17. Continuous Monitoring, Learning, and Evaluation (MLE) cycles must be maintained to build institutional trust
Supporting Materials
* The data presented is self-reported by the respective organisations. Readers should consult the original sources for further details.