# Legal AI for solo practitioners in India: a working stack **TL;DR:** A solo advocate has no library, no junior, and no research budget, so the AI question is not which enterprise suite to buy but which free tool to open first thing tomorrow morning. Start with eCourts, the National Judicial Data Grid and Indian Kanoon, all free, before spending a rupee. Pay only once a specific task has cost you real time on a real matter and you can name what you are buying, and treat every AI-generated citation as unverified until you have opened the primary source yourself. --- ## On this page - [Who this is for](#who-this-is-for) - [The four bottlenecks that eat a solo's day](#the-four-bottlenecks-that-eat-a-solos-day) - [The first week](#the-first-week) - [The free-plus-cheap stack before you pay for anything](#the-free-plus-cheap-stack-before-you-pay-for-anything) - [What a solo should automate and what stays manual](#what-a-solo-should-automate-and-what-stays-manual) - [When paying is justified, and when it is not](#when-paying-is-justified-and-when-it-is-not) - [The confidentiality problem when there is no firm around you](#the-confidentiality-problem-when-there-is-no-firm-around-you) - [The verification habit that keeps you out of trouble](#the-verification-habit-that-keeps-you-out-of-trouble) - [A week-one adoption sequence](#a-week-one-adoption-sequence) - [Frequently asked questions](#frequently-asked-questions) --- ## Who this is for This is written for one person: an advocate who practises alone, has no associate to hand research to, no clerk who maintains a diary across three courts, and no partner to check a draft before it goes out. If you share a chamber with two or three other independent advocates who each carry their own files, this still applies to you, because sharing rent does not mean sharing a research budget or a junior. This is not the situation covered in [legal AI for a small Indian law firm](/blog/legal-ai-small-law-firm-india), which assumes three to fifteen lawyers, a partner who signs off on associates' work, and a firm bank account that can absorb a subscription that fails. A solo has none of that structure to lean on. Every decision on this page, from which tool to open first to who reads a draft before it is filed, is made by exactly one person, and that person is also the one who appears in court and answers to the client and the Bar Council if something goes wrong. The constraint that shapes everything else is time, not money. A solo advocate with five matters listed on a Tuesday does not have three hours to read a 140-page judgment before an 11 am hearing. The question this page answers is what actually saves that person time on Tuesday morning, at no cost or low cost, without creating a bigger problem than the one it solves. --- ## The four bottlenecks that eat a solo's day Four problems recur across solo practice regardless of what area of law you handle. Naming them precisely matters more than naming a tool, because the tool that solves one does nothing for the other three. **Finding authority fast before a listing.** A matter gets listed, or a client walks in with an order passed against them, and you have hours, not days, to find the authorities that support your position. In a firm, a junior does a first pass and a partner narrows it. Alone, you do both passes yourself, usually late at night before a morning hearing. The cost of a slow search here is not academic. It is standing up in court with a weaker citation than the one that exists on the record. **Drafting from scratch.** A legal notice, a reply, a bail application, a writ petition. Every solo advocate has a personal library of precedent drafts on their laptop, built matter by matter over years, and every one of those drafts has to be adapted, not copied, because facts differ. Building the first draft of a new document from a blank page is the single biggest time cost in a solo's week that does not involve standing in a courtroom. [Drafting a bail application](/blog/drafting-bail-application-guide), [drafting a legal notice](/blog/how-to-draft-legal-notice) and [drafting a writ petition](/blog/ai-draft-writ-petition) each have their own structure, and getting that structure right on the first pass is what separates a draft that needs one round of edits from one that needs five. **Translating vernacular orders.** A trial court order from a district in Uttar Pradesh, Bihar or Madhya Pradesh often comes in Hindi. An order from a Kannada, Tamil, Telugu or Marathi-speaking district comes in the state language. If your appeal or revision goes to a High Court that works in English, you need an accurate English rendering of that order fast, and a mistranslated operative portion is not a cosmetic error. It changes what you argue on appeal. [Vernacular legal research](/blog/hindi-vernacular-legal-research-ai) covers how AI translation actually performs against certified translation on legal text, and where it still needs a human check. **Keeping track of dates across courts.** A solo who appears in a district court, a consumer forum and a High Court in the same week is running three separate diary systems in their head, because those three fora do not share a cause list format or a numbering convention. Missing a date is not a productivity problem. It can be the difference between an application being heard and a matter being dismissed for non-prosecution. [Checking case status on eCourts and the NJDG](/blog/check-case-status-ecourts-njdg) is the free starting point for solving this, before any paid diary tool enters the picture. These four bottlenecks are the only ones worth building a stack around. Everything else, from marketing to accounts to office management, is a different problem with a different set of solutions, and folding it into the same tool decision is how a solo ends up paying for features they never open. --- ## The first week Do not try to fix all four bottlenecks in the same week. Pick the one that cost you the most time in the last thirty days and start there. **Day one and two: map what already exists for free.** Register on the [eCourts Services](https://services.ecourts.gov.in) portal for every court where you have a matter pending. Bookmark the [National Judicial Data Grid](https://njdg.ecourts.gov.in) for your district and your High Court bench. Create a free account on [Indian Kanoon](https://indiankanoon.org) if you do not already have one. None of this costs anything, and all three are used daily by advocates across India, including in firms that also pay for commercial databases. **Day three: pick one recurring draft and build a template.** Take the document type you draft most often, whether that is a legal notice, a bail application or a written statement, and turn your best past draft into a clean template with placeholders. This is not an AI step. It is the groundwork that makes any AI drafting tool useful later, because a tool that fills placeholders in a structure you trust is far more reliable than one asked to invent structure from nothing. **Day four and five: test one AI tool on one real but low-stakes task.** Do not point it at a matter with a hearing in three days. Pick something with slack in the timeline, such as a first draft of a notice for a matter that is still two weeks from a deadline, and see what the tool actually produces before you decide whether it earns a place in your week. **Weekend: decide what you learned.** Did the tool save you time net of the time spent checking its output? If yes, use it again next week on a slightly higher-stakes task. If no, drop it and try a different tool for the same task, or accept that the task stays manual. A solo practice does not have the slack to run a six-month pilot the way a firm might. One week of honest testing on real but low-risk work tells you more than a vendor demo ever will. --- ## The free-plus-cheap stack before you pay for anything Before any paid tool, four resources cost nothing and cover most of what a solo needs day to day. **eCourts and the National Judicial Data Grid** give you case status, cause lists, and order copies for district courts, High Courts and the Supreme Court in one place, run by the government and free to use. This is the baseline for the date-tracking bottleneck described above, and it works even in districts with no other digital infrastructure. **Indian Kanoon** is a free search engine for Indian case law, covering the Supreme Court, High Courts and tribunals, with full text search across judgments going back decades. For a quick check of whether a point has been decided before, it is usually the fastest free option available, though it does not flag whether a judgment has since been overruled or distinguished, which is a separate problem covered further down. **A general AI chatbot's free tier**, used only for non-confidential drafting scaffolding such as an outline for an argument or a first pass at plain-language client correspondence, never for anything containing a client's name, facts, or documents. The confidentiality section below explains why that line matters and where it comes from. **Your own template library**, built once and reused, which is free in the sense that it costs only the time you already spent drafting the original documents. This stack has a real gap: none of these four tools verifies whether a citation is still good law, and none of them is built specifically for Indian legal research, which means none of them indexes Indian judgments the way a purpose-built legal research tool does or shows you the underlying paragraph a claim is based on. [Choosing an Indian case law search engine](/blog/choosing-indian-case-law-search-engine) and [free versus paid legal AI in India](/blog/free-vs-paid-legal-ai-india) go through that gap in more detail. A research tool built to show its source paragraph, the way a citator with [good law checking](/blog/good-law-checking) does, closes exactly that gap, and is the first thing worth paying for once the free stack has been in daily use for a few weeks and you can point to the specific task it is not covering. --- ## What a solo should automate and what stays manual | Task | Automate | Stay manual | |---|---|---| | First-pass search for authorities on a settled point | ✓ | | | Deciding which authority to actually cite in a filing | | ✗ | | First draft of a routine document from a template | ✓ | | | Final read of any document before it is signed and filed | | ✗ | | Translating a vernacular order for your own understanding | ✓ | | | Relying on that translation for the operative portion without checking the original | | ✗ | | Tracking cause lists and hearing dates across courts | ✓ | | | Confirming today's cause list before leaving for court | | ✗ | | Summarising a long judgment to find the relevant paragraphs | ✓ | | | Deciding whether a judgment applies to your facts | | ✗ | | Checking whether a citation has been overruled or distinguished | ✓ | | | Signing off on that check before it goes into a filing | | ✗ | The pattern across every row is the same. AI is fast at narrowing a large set down to a small one, whether that set is judgments, draft language, or translated text. A human, specifically the advocate whose name is on the vakalatnama, is the only one who should make the final call on what goes into a filing. [The advocate's duty to verify AI output](/blog/lawyer-duty-verify-ai-output) covers where that duty comes from and what it looks like in practice. --- ## When paying is justified, and when it is not Paying for a tool is justified when three conditions are all true. The task is one you do often enough that a few minutes saved per instance adds up over a month. The free option you tried on that task produced output you had to rework so heavily that the time saved was marginal. And you can name the specific gap the paid tool closes, not a vague sense that it might help. Paying is not justified when a free government resource already does the job, when the task happens rarely enough that the subscription cost outweighs the time saved over a year, or when you are buying a tool because a colleague uses it rather than because you have tested it against your own work. [The real cost of legal research software in India](/blog/legal-research-software-cost-india) sets out how to think about subscription cost against billable hours saved, which matters more for a solo than for a firm because there is no shared cost to spread it across. If you do decide a citation-verification tool is worth paying for, check the vendor's own pricing page rather than a reseller's quote, since published tiers for solo practitioners vary. Niyam, for instance, publishes an Advocate tier on its own pricing page aimed specifically at solo practitioners, priced separately from its higher-volume Professional and Firm tiers, with a paid three-day trial rather than a permanent free tier. Read the published terms directly at [niyam.ai](https://niyam.ai) before committing, and compare that structure against whatever else you are considering, because pricing structures across Indian legal AI vendors differ enough that a decision made on a colleague's recommendation alone is not a decision made on your own numbers. The honest failure mode to watch for is the subscription nobody opens after the first month. That happens when a solo pays for a tool during a busy week under time pressure, then never returns to it once the immediate crisis passes, because the tool was never tested against a real recurring task in the first place. The first-week testing approach above exists specifically to prevent that outcome. --- ## The confidentiality problem when there is no firm around you A firm has an IT policy, sometimes a signed data processing agreement with its vendors, and at minimum a partner whose job includes asking whether a tool is safe to put client material into. A solo advocate has none of that, and the temptation to solve a confidentiality question with a free general chatbot instead of a tool built for legal work is strongest exactly where the risk is highest, on a matter where you are working alone at midnight before a hearing. The professional obligation here is not a matter of general caution. Section 132 of the [Bharatiya Sakshya Adhiniyam, 2023](https://www.indiacode.nic.in/handle/123456789/20063), the successor to Section 126 of the Indian Evidence Act, 1872, protects communications between an advocate and a client made in the course of and for the purpose of that engagement from disclosure, and that protection does not lapse when the engagement ends. The [Bar Council of India's rules on professional standards](https://www.barcouncilofindia.org/info/rules-on-professional-standards) carry the confidentiality duty across into professional discipline, so a breach is not only an evidentiary problem. Pasting a client's facts, documents, or a draft built from them into a general-purpose AI tool whose terms allow that input to be retained or used for model training sits uncomfortably close to that line, and a solo advocate has no compliance officer to catch it before it happens. Two separate things follow from this. First, read the terms of service of any AI tool before you put client material into it, specifically the clause on whether your inputs are used to train the underlying model or retained beyond the session, and prefer tools that state plainly that they do not. Second, the Digital Personal Data Protection Act, 2023 adds a separate and additional layer where the material includes personal data, since a solo practice processing personal data as a data fiduciary carries its own consent and purpose-limitation obligations regardless of what your professional confidentiality duty already requires. [DPDP Rules 2025](/blog/dpdp-rules-2025) covers what those obligations actually require in practice, including for practices too small to have a dedicated compliance function. The practical rule that follows from both duties is simple to state and easy to forget under deadline pressure. Client-identifying facts, documents and drafts go only into tools whose data-handling terms you have actually read and are comfortable with. Everything else, including general research on a point of law with no client facts attached, can go into a free general tool without the same risk. --- ## The verification habit that keeps you out of trouble On 2 July 2026 the Supreme Court of India decided [Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr.](/blog/sc-ai-generated-case-law-2026), 2026 INSC 668, Civil Appeal No. 11950 of 2025, and set aside orders of the National Company Law Tribunal and the National Company Law Appellate Tribunal because the NCLT had relied on six precedents that were fake, wrongly cited, or carried paragraphs that do not exist in the real judgments. The bench of Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe held that courts must adopt a zero-tolerance mode for producing, citing or using AI-generated precedents without verification, and that citing such material without verification is misconduct on the part of an advocate. The full judgment is on the [Supreme Court's own judgment portal](https://api.sci.gov.in/supremecourt/2025/52338/52338_2025_5_1501_71939_Judgement_02-Jul-2026.pdf) and on [Indian Kanoon](https://indiankanoon.org/doc/113338666/). This did not come from a solo practitioner cutting corners. Paragraph 16 of the judgment records that the fake authorities in that case were not cited by counsel at the bar at all. They were introduced by the adjudicating authority's own research. The lesson for a solo advocate is not that AI use itself is dangerous. It is that fabricated citations can enter a case from any point in the process, including one you did not control, and the only defence available to the advocate whose name is on the filing is having verified what actually went in. Verification against Pooja Ramesh Singh has to clear two separate checks, because the six failed citations in that case failed in two different ways. Three did not exist at all, which any search of a reporter or database catches immediately. Two were real, correctly cited Supreme Court judgments with paragraphs attributed to them that are not actually in those judgments, which a check that stops at confirming the case exists will miss entirely. [How to vet legal AI for citation accuracy](/blog/how-to-vet-legal-ai-citation-accuracy) sets out the specific test for the second failure mode, and it comes down to one question: does the paragraph you are about to cite actually appear in that judgment, in those words, at that paragraph number. [How to cite Indian judgments](/blog/how-to-cite-indian-judgments) and [the difference between primary and secondary legal sources](/blog/primary-vs-secondary-legal-sources) cover the mechanics of checking a citation against the reporter it claims to come from. Build this into a fixed habit rather than a judgment call made fresh each time. Before any AI-suggested citation goes into a draft, open the primary source, in a reporter or on the court's own portal, and confirm both that the case exists and that the paragraph you are quoting is actually in it. A citator that flags whether a judgment is still good law before you rely on it, rather than leaving you to discover an overruling authority at the hearing, removes one whole category of this risk, but it does not remove the second check. Confirming that the paragraph text itself is real still has to happen every time, on every citation, regardless of which tool produced the first draft. For a solo advocate this costs a few minutes per citation and no institutional infrastructure at all. It is the single habit that separates AI use that saves time from AI use that ends in a Rule 17 or Rule 36 inquiry. --- ## A week-one adoption sequence ```mermaid flowchart TD A[Day 1: register on eCourts for pending matters] --> B[Day 1: bookmark NJDG for your district/bench] B --> C[Day 2: create free Indian Kanoon account] C --> D[Day 3: build one template from your best past draft] D --> E[Day 4: pick one low-stakes recurring task] E --> F[Day 4: test one AI tool on that task only] F --> G[Day 5: check output against primary sources] G --> H{Saved net time?} H -->|Yes| I[Day 6: reuse on slightly higher-stakes task] H -->|No| J[Day 6: try a different tool or keep task manual] I --> K[Day 7: read data-handling terms before any client-identifying input] J --> K K --> L[Week 2: repeat for the next bottleneck] ``` The sequence deliberately keeps every AI decision behind a manual step. Registration and bookmarking cost nothing and carry no confidentiality risk. Template-building uses only your own past work. The AI test itself is confined to one task with slack in the timeline, and the terms-of-service check happens before, not after, any client-identifying material is involved. By the second week you have real evidence, from your own practice, of what is worth paying for. --- ## Frequently asked questions ### Is it ethical for a solo practitioner to use AI for drafting court documents? Yes, using AI to produce a first draft is not itself a professional conduct issue. The obligation that applies is the same one that applies to any draft, AI-assisted or not: the advocate whose name is on the vakalatnama must verify the content, including every citation, before it is filed. [Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd.](/blog/sc-ai-generated-case-law-2026) makes clear that citing unverified AI-generated material is what creates the misconduct exposure, not the drafting tool itself. ### Can a solo advocate use free AI tools on client matters? Free general AI chatbots can be used for research that contains no client-identifying facts, such as a general question about how a point of law has been decided. They should not be used for anything containing a client's name, documents, or facts unless you have read the tool's terms on data retention and training use and are satisfied with them, because the confidentiality duty under Rule 17 of the BCI Rules continues regardless of which tool you used. ### What happens if an AI tool cites a fake case in my petition? If it reaches a filing, you carry the exposure described in Pooja Ramesh Singh, where the Supreme Court held that citing AI-generated precedent without verification is misconduct on the part of an advocate. The fix is procedural, not technological: check every AI-suggested citation against the primary source, confirming both that the case exists and that the paragraph you are quoting actually appears in it, before it goes into any draft you sign. ### Does the Bar Council of India regulate AI use by advocates specifically? Not yet through a dedicated AI rule. In Pooja Ramesh Singh, the Supreme Court directed the Bar Council of India to constitute a committee and prescribe a guiding principle and disciplinary action for AI-generated precedent misuse. Until that framework is published, the existing duty to verify what you cite, drawn from ordinary professional conduct standards, applies to AI-assisted work exactly as it applies to any other draft. ### Can I use ChatGPT for legal research in India? You can use a general chatbot for background research on a point of law with no client facts attached, as a starting point rather than an authoritative source, since general models are not built on an indexed corpus of Indian judgments and cannot show you the paragraph a claim is based on. Any citation it produces still needs the same primary-source check described above before it goes anywhere near a filing. ### Is Indian Kanoon free? Yes, Indian Kanoon is a free search engine covering Supreme Court, High Court and tribunal judgments, usable without a paid account for basic search and full-text reading. It does not flag whether a judgment has since been overruled or distinguished, so a separate good-law check is still needed before relying on anything it surfaces. ### How do I verify a citation before filing? Open the primary source, either the court's own judgment portal or a reporter, and confirm two things separately: that the case exists under the cited name and citation, and that the specific paragraph or quoted text actually appears in that judgment. [How to vet legal AI for citation accuracy](/blog/how-to-vet-legal-ai-citation-accuracy) sets out both checks in detail, since checking only that the case exists misses fabricated paragraphs attached to real judgments. ### What is the difference between eCourts and the NJDG? eCourts Services gives you case-specific information: status, cause lists, and order copies for an individual matter across district courts, High Courts and the Supreme Court. The National Judicial Data Grid gives you aggregate pendency and disposal data for a court or a case type. A solo advocate uses eCourts daily for individual matters and the NJDG occasionally for context on how long a category of case typically takes. ### Can I put client documents into a general AI chatbot? Only after reading that tool's terms on whether inputs are retained or used for model training, and generally not for documents containing a client's name or case-specific facts. Rule 17 of the BCI Rules, tied to the confidentiality protection in Section 132 of the Bharatiya Sakshya Adhiniyam, 2023, continues to bind you regardless of which software you use, and a solo has no compliance function to catch a mistake before it happens. ### Does the DPDP Act apply to a solo law practice? Yes, if your practice processes personal data, which most litigation and transactional practices do, you are a data fiduciary under the Digital Personal Data Protection Act, 2023, with consent and purpose-limitation obligations that apply regardless of firm size. [DPDP Rules 2025](/blog/dpdp-rules-2025) covers the specific obligations, including for practices with no dedicated compliance staff. ### What is Section 132 of the Bharatiya Sakshya Adhiniyam? It is the current provision protecting communications between an advocate and client made in the course of and for the purpose of the engagement from disclosure, corresponding to the former Section 126 of the Indian Evidence Act, 1872. Rule 17 of the BCI Rules ties the professional confidentiality duty directly to this section, and that duty continues even after the engagement has ended. ### How much does a citation-verification tool cost for a solo practitioner? Published pricing varies by vendor and should be checked on the vendor's own pricing page rather than taken from a colleague's recommendation. Niyam, for example, publishes a separate Advocate tier for solo practitioners on its own pricing page at [niyam.ai](https://niyam.ai), distinct from its higher-volume tiers, with a paid short trial rather than an ongoing free tier. Compare the published terms of whichever tools you are considering before deciding. ### Should I translate vernacular judgments with a general translation tool? For a first read to understand what an order says, yes, a general translation tool is a reasonable starting point and faster than waiting for a certified translation. For anything that will be quoted in an appeal or revision, particularly the operative portion of an order, get the translation checked against the original text, because a translation error in the operative part changes what you can argue. [Vernacular legal research with AI](/blog/hindi-vernacular-legal-research-ai) covers where AI translation is reliable and where it still needs a human check. ### What is the fastest free way to check if a judgment is still good law? There is no complete free substitute for a citator that tracks subsequent judicial treatment, since that requires an indexed database of which later judgments cite, follow, distinguish or overrule an earlier one. Indian Kanoon's citation list on a judgment page is a useful starting point, but it does not classify how a later case treated the earlier one, so a manual read of the citing judgments is still required if you rely on it as your only check. ### Can AI draft a bail application for me? AI can produce a structured first draft from a template and the facts you provide, covering the grounds, the statutory framework and the standard prayer clause. It cannot assess whether your specific facts meet the threshold a particular bench has been applying, which is a judgment call that has to stay with the advocate. [Drafting a bail application](/blog/drafting-bail-application-guide) sets out what the structure should contain regardless of who produces the first draft. ### What happens if my practice has no IT support and a client raises a data-handling concern? Address it the same way a firm would, just without a dedicated person to do it: read and be able to explain, in plain terms, what happens to a client's data in every tool you use, including whether it is retained, whether it trains an underlying model, and where it is stored. If you cannot answer that for a tool, do not put client material into it. This is a smaller version of the same duty every advocate carries under Rule 17 of the BCI Rules, not a separate obligation created by firm size.