240 hours a year: what AI gives India's lawyers, and what it can't

Legal professionals expect AI to free up nearly 240 hours a year each, up from 200 hours a year earlier, worth about $19,000 per person (Thomson Reuters, 2025). That is six working weeks. But in Indian courts, AI has also produced something no lawyer wants: judgments that don't exist.

This post covers both sides. What is happening in Indian law, which roles gain, the skills to build, and the risk every advocate now has to take seriously.

What is happening in Indian law

The courts themselves use AI. SUVAS translates Supreme Court judgments into 18 Indian languages, and the e-Courts project has an outlay of ₹7,210 crore (PIB, 2026). The government's position is clear: AI "assists but never replaces judicial decision-making".

Large firms are moving. Cyril Amarchand Mangaldas uses AI for due diligence, fact cross-verification, litigation support, first drafts and legal research (Conventus Law, 2025). It piloted its generative-AI tool with 380 lawyers before rolling it out firm-wide (Legora, 2025). A pilot of that size is a sign of care, not hesitation.

Strategy matters. Thomson Reuters found organisations with an AI strategy are twice as likely to see revenue growth as those without one. Using AI casually is not the same as using it well.

For most Indian lawyers, though, the question is not firm strategy. It is personal practice. A solo advocate or a small chamber has no AI team, no approved-tools list and no one to check the output. The habits each lawyer builds now will decide whether those 240 hours are real.

Which roles gain, and how

The examples below are illustrations of a working week, not survey findings.

Litigation associates. An associate in a Delhi chamber has a bail hearing tomorrow and three hundred pages of chargesheet. She asks an assistant to build a timeline of events with page references. She then opens every referenced page and checks it. The timeline took an hour instead of a day, and the checking is what makes it usable.

Transactional and corporate lawyers. In due diligence, an assistant flags change-of-control clauses across two hundred contracts. The lawyer reviews each flagged clause and the ones the tool marked as clean, because missing one is the expensive mistake.

In-house counsel. A company's legal head gets a first draft of a vendor agreement from the company's standard templates, then negotiates the parts that matter. Routine NDAs stop eating her week.

Compliance officers. A compliance team summarises a new regulator's circular into an action list for each department. Someone still reads the original circular, line by line.

Paralegals and law clerks. Organising documents, preparing indexes and drafting cover letters get faster. The clerk who learns to check AI work carefully becomes more valuable, not less.

Litigants and small-town practitioners. SUVAS translations mean a lawyer in a district court can read a Supreme Court judgment in her own language. That widens access to the law itself.

The skills to build

• Verifying every citation. Open every case the tool cites. Read the paragraph it relies on. If you cannot find it, it does not exist.
• Giving precise instructions. Name the statute, the jurisdiction, the facts and the question. Vague prompts get vague, and sometimes invented, answers.
• Protecting client confidentiality. Know which tools your firm has approved. Client documents must never go into a public AI assistant.
• Knowing where AI is weak. In a Harvard and BCG study, consultants were right 84% of the time on a task designed for AI to fail, but only 60–70% of the time with AI (HBS AI Institute). Legal reasoning on unusual facts is exactly that kind of task.

The risk: citations that don't exist

This is not a hypothetical. A Bengaluru bench of the Income Tax Appellate Tribunal withdrew an order that cited non-existent judgments (Taxscan, 2025). On 27 February 2026, the Supreme Court of India said that relying on fake AI-generated judgments "would be a misconduct" (MediaNama, 2026).

It is happening everywhere. A public database had recorded 2,079 such cases worldwide as of 25 September 2026 (AI Hallucination Cases database). And professional firms are not immune: Deloitte Australia refunded over A$97,000 on a A$440,000 government report that contained a fabricated court quote and non-existent papers (CFO Dive, 2025).

The cause is usually not the tool alone. It is the habit of trusting it. 66% of employees rely on AI output without evaluating it (KPMG & University of Melbourne, 2025). In law, that habit can end a career.

What to do next

1. Adopt one rule for yourself today: no citation goes into a draft until you have read it in the original.
2. Try an approved assistant on one low-risk task, such as a chronology or a document index, and check every reference.
3. Ask your firm or chamber which tools are approved for client documents, and which are not.
4. Read the Supreme Court's February 2026 remarks on fake AI judgments, and share them with juniors.
5. Keep a short log of errors you catch. It will teach you where the tool is weakest in your area of practice.

If you want to build these habits with others, AI Fluency is taught live in small cohorts, with a project you keep each month.

Keep reading