# AI adoption among Indian lawyers in 2026: the real data

**TL;DR:** The 2026 Wolters Kluwer Future Ready Lawyer Survey found that 92 percent of the 810 lawyers it interviewed use at least one AI tool in daily work, and 62 percent report weekly time savings of between 6 and 20 percent of their work week. That survey covered the United States, China and nine European countries. It did not survey India. The Thomson Reuters 2026 Future of Professionals report, drawn from more than 1,800 professionals across 62 countries, found 74 percent using AI tools several times a week but 91 percent saying their organisations fall short of what the technology could deliver. The honest reading for an Indian practice is that adoption is no longer the question, governance and verification are, and that no published 2026 survey gives you a reliable India-only adoption number.

---

## On this page

- [What the 2026 surveys actually found](#what-the-2026-surveys-actually-found)
- [The scope problem: India was not in the sample](#the-scope-problem-india-was-not-in-the-sample)
- [Adoption is nearly universal, value is not](#adoption-is-nearly-universal-value-is-not)
- [The barriers are human, not technical](#the-barriers-are-human-not-technical)
- [Shadow AI and the confidentiality problem it creates](#shadow-ai-and-the-confidentiality-problem-it-creates)
- [The work moved from searching to verifying](#the-work-moved-from-searching-to-verifying)
- [What India-specific evidence exists](#what-india-specific-evidence-exists)
- [What Indian courts have already decided about AI use](#what-indian-courts-have-already-decided-about-ai-use)
- [What this means for a two-person litigation practice](#what-this-means-for-a-two-person-litigation-practice)
- [What this means for a large firm and its junior lawyers](#what-this-means-for-a-large-firm-and-its-junior-lawyers)
- [How to read an adoption statistic before you act on it](#how-to-read-an-adoption-statistic-before-you-act-on-it)
- [A twelve-month adoption sequence you can actually run](#a-twelve-month-adoption-sequence-you-can-actually-run)
- [Frequently asked questions](#frequently-asked-questions)

---

## What the 2026 surveys actually found

Two large surveys published in 2026 carry most of the numbers you will see quoted at conferences and in vendor decks this year.

The first is the [2026 Future Ready Lawyer Survey from Wolters Kluwer Legal and Regulatory](https://www.wolterskluwer.com/en/news/wolters-kluwer-releases-2026-future-ready-lawyer-survey-report), published on 10 March 2026. It reports that 92 percent of respondents use at least one AI tool in their daily workflow, that 62 percent report time savings falling between 6 percent and 20 percent of their work week, and that 60 percent expect their organisation's investment in AI to increase over the next three years. A further 61 percent said they had growing confidence in their organisation's ability to adapt business practices, service offerings, workflows and pricing models to AI-driven efficiencies. On the market side, 54 percent expect law firms to convert improved efficiency into either a higher volume of clients or more competitive pricing.

The second is the [Thomson Reuters 2026 Future of Professionals report](https://www.thomsonreuters.com/en-us/posts/technology/future-of-professionals-2026/), drawn from a survey of more than 1,800 professionals across 62 countries in law, tax, audit, accounting, compliance, risk and global trade. It found 74 percent of respondents using AI tools several times a week and 44 percent relying on them multiple times a day. It also found that 91 percent said their organisation is falling short of what AI could deliver, that more than a third admit using AI tools their organisation has not sanctioned or cannot see, and that 18 percent said their organisation has no strategic direction on AI at all.

The two sets of numbers do not contradict each other. Wolters Kluwer asked lawyers whether they use at least one AI tool in daily work, a low bar that produces a high number. Thomson Reuters asked about frequency across a broader professional population, a stricter question that produces a lower one.

The number that matters most for planning is neither adoption figure. It is the 91 percent value gap. Nearly everyone is using something. Almost nobody thinks their organisation is getting what it should out of it.

---

## The scope problem: India was not in the sample

This is the most important qualification on every figure above, and the one that gets dropped fastest when the numbers travel.

The Wolters Kluwer methodology note is explicit. The 2026 Future Ready Lawyer Survey "included quantitative interviews with 810 lawyers in law firms and corporate legal departments across the U.S., China, and nine European countries: Germany, the Netherlands, the United Kingdom, Belgium, France, Italy, Spain, Poland, and Hungary". India is not on that list. Neither is any other South Asian jurisdiction.

So when you see "over 90 percent of lawyers now use AI" in an Indian legal newsletter, the accurate sentence is: over 90 percent of the 810 lawyers Wolters Kluwer interviewed in the United States, China and nine European countries reported using at least one AI tool in daily work. That is a real finding. It is not a measurement of Indian practice.

The Thomson Reuters survey spans 62 countries, which very likely includes India, but the published summaries do not break results down by country. You cannot extract an India figure from it either.

Indian practice also differs from the surveyed jurisdictions on four dimensions that directly drive adoption:

1. **Case-law volume and structure.** Indian judgments run long, cite heavily, and arrive from 25 High Courts plus the Supreme Court in inconsistent formats. A tool benchmarked on US or EU corpora does not inherit that.
2. **Language.** A meaningful share of district-court orders and pleadings are in Hindi, Marathi, Tamil, Bengali and other Indian languages. See [Hindi and vernacular legal research with AI](/blog/hindi-vernacular-legal-research-ai) for what that requires of a tool.
3. **Firm structure.** The Indian market is dominated by sole practitioners and small chambers, not the mid-size and large firms that fill most Western survey panels.
4. **Regulatory posture.** India has no AI statute. Governance is arriving through guidelines, court policies and the DPDP Act rather than anything resembling the EU AI Act.

Treat any adoption figure imported from a Western survey as a directional signal about where the profession is heading, not as a description of your bar association.

---

## Adoption is nearly universal, value is not

Here is the full set of figures worth carrying into a partners' meeting, with the source for each. Check the scope column before you quote any of them.

| Figure | What it measures | Scope | Source |
|---|---|---|---|
| 92% | Use at least one AI tool in daily workflow | 810 lawyers, US, China, 9 European countries | [Wolters Kluwer 2026 Future Ready Lawyer Survey](https://www.wolterskluwer.com/en/news/wolters-kluwer-releases-2026-future-ready-lawyer-survey-report) |
| 62% | Report weekly time savings of 6% to 20% of the work week | Same as above | Wolters Kluwer, 2026 |
| 60% | Expect organisational AI investment to increase over three years | Same as above | Wolters Kluwer, 2026 |
| 54% | Expect firms to convert efficiency into volume or pricing | Same as above | Wolters Kluwer, 2026 |
| 39% / 39% / 35% | Top barriers: ethics and data privacy, inadequate training, resistance to change | Same as above | [Wolters Kluwer analysis of the 2026 survey](https://www.wolterskluwer.com/en/expert-insights/legal-ai-adoption-time-savings-revenue-growth) |
| 32% | Attribute an 11% to 20% revenue increase to their use of AI | Same as above | Wolters Kluwer, 2026 |
| 74% / 44% | Use AI several times a week / multiple times a day | 1,800+ professionals, 62 countries, all fiduciary professions | [Thomson Reuters 2026 Future of Professionals](https://www.thomsonreuters.com/en-us/posts/technology/future-of-professionals-2026/) |
| 91% | Say their organisation falls short of what AI could deliver | Same as above | Thomson Reuters, 2026 |
| 78% vs 6% | Clients calling AI-enabled quality essential vs clients consistently receiving it | Same as above | Thomson Reuters, 2026 |
| 18% | Say their organisation has no strategic direction on AI | Same as above | Thomson Reuters, 2026 |
| 40% | Say their organisation now uses generative AI, up from 22% a year earlier | Professional services respondents | [Thomson Reuters 2026 AI in Professional Services Report](https://www.thomsonreuters.com/en/reports/2026-ai-in-professional-services-report) |
| 15% | Say their organisation uses agentic AI today | Same as above | Thomson Reuters, 2026 |
| 18% | Say their organisation tracks return on investment from AI | Same as above | Thomson Reuters, 2026 |

Read the last row again. Eighteen percent of organisations track whether any of this is working, and 40 percent do not know whether return on investment is measured at all. Adoption has run ahead of measurement, which is why the value gap persists.

The gap between the 78 percent of corporate clients who call AI-enabled quality improvements essential and the 6 percent who say they consistently receive them is the commercial version of the same problem. Clients priced AI into their expectations before most providers delivered against it. [A 2026 buyer's guide to AI tools for lawyers in India](/blog/best-ai-tools-for-lawyers-india) sets out the criteria a client-facing tool has to clear.

---

## The barriers are human, not technical

Wolters Kluwer's own analysis of the 2026 results names the three top obstacles to further AI implementation: ethical concerns related to AI and data privacy at 39 percent, inadequate training at 39 percent, and resistance to change at 35 percent.

None of those is a software problem. All three are solved by decisions a managing partner can make in a week and by work that takes a quarter.

**Ethics and data privacy at 39 percent.** For an Indian practice this is not abstract. Pasting a client's file note into a consumer chatbot is a disclosure to a third party, and the Digital Personal Data Protection Act, 2023 governs how you handle personal data in that file. The practical control is not a policy document, it is a tool choice: use something where you know where the data sits and what happens to it. [Legal AI data residency in India](/blog/legal-ai-data-residency-india) sets out what to ask a vendor, and [the DPDP Rules 2025](/blog/dpdp-rules-2025) covers the obligations that now attach.

**Inadequate training at 39 percent.** This is the barrier most firms misdiagnose as a tooling problem. The failure mode is buying a licence, sending a link, and assuming a competent lawyer will work out the rest. They will work out enough to be dangerous. What they will not work out alone is prompt discipline, the difference between a retrieval-grounded answer and a generated one, or the verification steps that keep a fabricated citation out of a filing. [How to vet legal AI for citation accuracy](/blog/how-to-vet-legal-ai-citation-accuracy) is the training material most firms skip.

**Resistance to change at 35 percent.** Some of this resistance is well founded. A senior counsel burned once by a junior's AI-assisted memo citing a case that does not exist is applying exactly the standard the profession requires. The answer is not enthusiasm, it is a workflow where verification is built in rather than bolted on.

The Thomson Reuters data supports the same reading. As [LawSites reported from that study](https://www.lawnext.com/2026/06/thomson-reuters-future-of-professionals-report-warns-of-widening-gap-between-ai-adoption-and-ai-value.html), when professionals were asked why AI strategies stall, the most common answers were that tools were not yet in place at 47 percent, that people were not trained to work in the intended way at 43 percent, that strategy was not translated into clear operational priorities at 32 percent, and that there was no shared understanding of the plan at 30 percent. In firms with a named AI strategy, 66 percent said AI was meeting or exceeding expectations. Where there was no active strategy, that fell to 22 percent.

A named strategy triples the reported value. That is the finding with the largest payoff in either survey, and it costs nothing but a decision.

---

## Shadow AI and the confidentiality problem it creates

The Thomson Reuters report found that more than a third of professionals use AI tools their organisation has not sanctioned, or use sanctioned tools in ways the organisation cannot see. LawSites puts the figure at 34 percent overall, rising to 41 percent among professionals who say their organisation is moving too slowly on AI.

Read those two numbers together. Shadow AI is a supply problem more than a discipline problem. Lawyers who cannot get a decent tool through the front door bring one in through the browser, and they do it more when they think the firm is behind.

For an Indian practice this creates a specific exposure. A junior drafting a reply in a matrimonial matter who pastes the client's affidavit into a free consumer chatbot has moved identifiable personal data, including data about health, finances and family, onto infrastructure the firm has no contract with and cannot audit. If the client later asks what happened to their file, the firm has no answer.

The fix is uncomfortable but simple. Sanction something usable. A firm that provides no approved tool is not preventing AI use, it is only preventing supervised AI use. [Free versus paid legal AI in India](/blog/free-vs-paid-legal-ai-india) works through what the free option actually costs once you account for that exposure.

The same Thomson Reuters study found professionals are clear about what they want from a sanctioned tool: 96 percent said it must safeguard confidential data, 94 percent said outputs must be grounded in authoritative content, and 90 percent said it must produce reasoning that can be explained and defended. Yet 41 percent said they lack access to AI tools designed for professional work and built on verified professional content. Those figures come from the LawSites summary of the report.

---

## The work moved from searching to verifying

The most useful description of what has actually changed in daily practice does not come from a percentage. It comes from a partner at one of India's largest firms.

Speaking at the India Today Conclave in New Delhi on 8 March 2025, Rishabh Shroff, Partner at Cyril Amarchand Mangaldas, said: "What would take a young lawyer 20, 30, 40, 50 hours to do? AI can do it in five seconds." In the same session, [reported by Business Today](https://www.businesstoday.in/bt-tv/video/rishabh-shroff-ai-can-do-in-five-seconds-what-takes-young-lawyers-20-50-hours-467270-2025-03-08), he described AI as disrupting labour arbitrage and raised its effects on hiring, promotion and mentorship in law firms. He also noted his firm's demonstration of a United States language model and the absence of Indian language models built for law firms.

That last point has aged into the central question for the Indian market. The compression Shroff describes is real. Whether its output can be trusted depends on whether the model retrieved Indian judgments or predicted what an Indian judgment would plausibly say.

The practical consequence is that the bottleneck in legal research has moved. Twenty years ago the hard part was finding the four cases that mattered. Today the hard part is confirming that the six cases a tool handed you exist, say what the summary claims they say, and are still good law. The hours saved on the first task are partly spent on the second, and a firm that budgets for the saving without budgeting for the verification will produce faster work of lower quality. [AI legal research without the hallucination risk](/blog/ai-legal-research-india) sets out the five-step check that closes the gap, and [how to cite Indian judgments](/blog/how-to-cite-indian-judgments) covers the citation form a court will accept.

This is why the shape of the tool matters more than its speed. A system that retrieves from an indexed corpus of Indian judgments and shows the paragraph it relied on lets you verify in seconds. A system that generates plausible text and offers a citation as decoration makes verification a separate research project. [Niyam](https://niyam.ai) is built on the first pattern, with answers tied back to the judgment text they came from.

For the mechanics of that step, [the duty to verify AI output](/blog/lawyer-duty-verify-ai-output) sets out what the profession expects of you, and [good-law checking](/blog/good-law-checking) covers the part that automated summaries most often get wrong.

---

## What India-specific evidence exists

There is no Indian equivalent of the Wolters Kluwer panel. What exists is fragmentary, and it is worth being clear about which pieces are hard and which are soft.

**Hard: government and court activity.** The [India AI Impact Summit 2026](https://www.newsonair.gov.in/india-ai-impact-summit-2026-ai-is-rapidly-transforming-indias-legal-ecosystem-with-ai-powered-research-tools), held at Bharat Mandapam in New Delhi in February 2026, included legal-sector AI among its exhibits. Akashvani News reported from the summit that AI tools in the Indian legal sector are being applied to client queries, document review and backend case management. Saakar S Yadav, Managing Director of Lexlegis.AI, told Akashvani that the company's product would make drafting legal documents efficient and fast. That is vendor commentary at a government event, not adoption data.

**Hard: the policy layer.** The Ministry of Electronics and Information Technology released the India AI Governance Guidelines on 5 November 2025, built on seven principles including trust as the foundation, people first, fairness and equity, innovation over restraint, accountability, understandable by design, and safety, resilience and sustainability. India still has no AI statute. The Chambers and Partners Artificial Intelligence 2026 guide for India records that AI is regulated through existing law, including the Information Technology Act, 2000, the DPDP Act, the Bharatiya Nyaya Sanhita, 2023 and the Copyright Act, 1957, and that the proposed Digital India Act has sat in draft since 2023.

**Hard: courtroom deployment.** Adalat AI, a speech-to-text tool for court proceedings, reached roughly 4,000 courtrooms by the end of 2025, and Kerala made it compulsory in every trial court from 1 November 2025. That is a real adoption number for one specific tool in one specific function. [Adalat AI and real-time transcription in Indian courts](/blog/adalat-ai-real-time-transcription-courts) covers the accuracy and confidentiality questions it raises. On the research side, the Supreme Court's own [electronic reports and neutral citations](/blog/e-scr-neutral-citations) have made judgments machine-readable in a way that makes grounded retrieval possible at all.

**Soft: practitioner surveys.** Manupatra has run a survey titled "AI Adoption and Its Impact in the Legal Industry: An Indian Perspective", summarised on Bar and Bench. The summary describes growing familiarity with AI among Indian legal professionals and names implementation cost, ethical implications and a shortage of skilled professionals as the main barriers. It publishes no headline percentages, so it cannot be used as a source for a number.

The honest position is this. Indian courts and Indian government are moving visibly. Indian practitioner adoption is almost certainly rising. Nobody has published a defensible percentage for it, and anyone quoting one to you should be asked where it came from.

---

## What Indian courts have already decided about AI use

While the surveys measure behaviour, Indian courts have been setting rules, and those rules bind you regardless of what any adoption figure says.

The Kerala High Court published its Policy Regarding Use of Artificial Intelligence Tools in District Judiciary on 19 July 2025, the first binding AI policy issued by an Indian High Court for its subordinate judiciary. The policy is addressed to District Judges and Chief Judicial Magistrates, covers all AI tools including generative language models, and applies to court computers, personal laptops and phones used for judicial work. Its core prohibition is that AI software must not be used to arrive at any findings, reliefs, order or judgment. AI is defined as assistive only, and responsibility for the content and integrity of a judicial order stays with the judge.

At the national level, the Supreme Court's AI Committee released the Draft Regulations for Use of Artificial Intelligence in Courts, 2026 on 3 June 2026. These are draft provisions open for comment, not binding law. Their centre of gravity is disclosure and verification: a party or counsel who used AI in preparing a document must say so at the time of filing, and responsibility for fabricated or false AI-generated content sits with the person who filed it. [Regulation 19 and the AI disclosure duty in pleadings](/blog/regulation-19-ai-disclosure-pleadings) works through the filing obligation in detail, and [what the Supreme Court's draft AI rules actually say](/blog/supreme-court-ai-rules-india) covers the broader framework.

The Chambers guide for India also records that several High Courts have developed policies on AI use by lawyers and judges, prescribing disclosures, verifications and consequences for violations. If you practise across multiple High Courts the rules are not uniform, so check the position in each. [What the draft rules require of the tools themselves](/blog/supreme-court-ai-rules-what-tools-need) is a separate question from what they require of you.

A ninety-two percent adoption statistic from Europe tells you nothing about whether your filing complies with the disclosure duty in the court you are appearing before. Only one of those two questions can get you into trouble.

---

## What this means for a two-person litigation practice

Most Indian lawyers do not work at a firm with an AI strategy, a knowledge management function, or a budget line for pilots. The surveys above are written for organisations that have those things. Here is the translation for a practice that does not.

**Assume the time saving is real but smaller than advertised.** The Wolters Kluwer finding is 6 to 20 percent of the work week for 62 percent of respondents. On a fifty-hour week that is three to ten hours, and it comes mostly from research and first-draft work rather than from hearings, client meetings or travel.

**Spend the saving on verification, not on more matters.** The fastest way to convert an AI time saving into a professional negligence problem is to take on more work with the freed hours and skip the checking step. Build the check in first. Take the volume once the check is habitual.

**Pick one function and get it right.** Research is usually the right first choice for a litigation practice, because the output is checkable against a primary source in minutes. Drafting is the harder second step because errors are less visible. [Where AI genuinely helps litigators](/blog/ai-legal-research-litigators-vs-corporate) sets out the split.

**Do not use a consumer chatbot on client facts.** This is the one non-negotiable. [ChatGPT for lawyers in India](/blog/chatgpt-for-lawyers-india) covers what a general model can and cannot do for Indian law, and [AI hallucinated citations in India](/blog/ai-hallucinated-citations-india) covers what has already gone wrong when lawyers assumed otherwise.

**Cost is not the barrier it was.** The entry price for purpose-built Indian legal AI is now low enough that the free-versus-paid decision turns on risk rather than budget. A tool grounded in Indian judgments removes a class of failure a general chatbot cannot. [Niyam](https://niyam.ai) is priced for exactly this kind of practice.

---

## What this means for a large firm and its junior lawyers

Shroff's point about labour arbitrage is the one large Indian firms have not yet answered.

If a task that took a first-year associate forty hours now takes minutes, the billing model, the staffing model and the training model break at the same time. Billing is the easiest of the three, since 54 percent of Wolters Kluwer respondents already expect firms to move toward volume or competitive pricing. Staffing is harder. Training is hardest.

Here the Thomson Reuters data is directly on point. As reported by LawSites from that study, 48 percent of professionals said they were worried about AI's impact on the development of independent judgment, and 71 percent said those early in their careers need structured support from experienced peers to build the skills AI risks displacing.

The forty hours a junior spent on a research memo were not purely wasteful. Some of that time was how a junior learned which authorities carry weight and how a line of cases develops. Compressing the task to five seconds removes the drudgery and the apprenticeship together. A firm that does not rebuild the second will find in five years that it has a partner track with nobody on it.

Practical responses that do not require a committee:

- Make the junior verify the AI output against the primary source and record what they checked. That preserves the reading, removes the copying.
- Have juniors write the argument, not the summary. [Summarisation](/blog/ai-judgment-summarisation) is the part the tool does.
- Keep at least one research task per matter deliberately manual for first and second-year lawyers.

There is a retention dimension too. LawSites reports from the same study that 62 percent of professionals said access to professional-grade AI tools would factor into accepting a new role, and that among those already using such tools, roughly one in three would turn down a role that did not offer them, against 12 percent of those without access.

---

## How to read an adoption statistic before you act on it

Most AI adoption numbers you will encounter this year are second-hand, and the errors are predictable. Use this table before you repeat one.

| Check | Do this | Not this |
|---|---|---|
| Scope | ✓ Name the jurisdictions surveyed in the same sentence as the number | ✗ Present a worldwide figure as an Indian figure |
| Sample | ✓ State the sample size and who was interviewed | ✗ Say "lawyers say" with no denominator |
| Question wording | ✓ Distinguish "uses at least one tool" from "uses AI daily for substantive work" | ✗ Treat every adoption figure as measuring the same thing |
| Source level | ✓ Confirm the figure on the surveying organisation's own page or report | ✗ Quote a vendor blog citing a newsletter citing a press release |
| Date | ✓ Give the publication date and the fieldwork date | ✗ Say "a recent survey" |
| Population | ✓ Check whether the survey covered lawyers only or all professionals | ✗ Apply a cross-profession figure to litigation practice |
| Metric type | ✓ Separate self-reported time savings from measured outcomes | ✗ Treat a perception figure as a productivity measurement |
| Derived claims | ✓ Use the range the survey published | ✗ Convert "6 to 20 percent" into a single average the survey never reported |

The Wolters Kluwer finding is that 62 percent of respondents fall in a band between 6 percent and 20 percent of the work week. Any single average derived from that band is somebody's arithmetic, not the survey's finding. If you need a number for a business case, use the band and say it is a band.

---

## A twelve-month adoption sequence you can actually run

The survey finding that a named AI strategy raises reported value from 22 percent to 66 percent is the argument for doing this in a defined order rather than opportunistically. The sequence below is deliberately slow at the start.

```mermaid
flowchart TD
    A[Month 1: name an owner and write one page] --> B[Month 2: pick one function, usually research]
    B --> C[Month 3: choose a tool grounded in Indian judgments]
    C --> D[Months 4 to 6: run it on live matters with mandatory verification]
    D --> E{Are citations checking out<br/>and hours actually saved?}
    E -->|No| F[Change tool or change workflow, repeat months 4 to 6]
    E -->|Yes| G[Months 7 to 9: write the firm AI policy and disclosure practice]
    G --> H[Months 10 to 12: extend to drafting, measure ROI, train juniors]
    F --> D
```

**Month 1. Name an owner and write one page.** The page states which tools are approved, which are prohibited, what may never be pasted into a tool, and who signs off on verification. One page beats a policy nobody reads.

**Month 2. Pick one function.** For litigation, research. For a transactional team, [first-pass contract review and drafting](/blog/ai-contract-drafting). Do not start with two.

**Month 3. Choose a tool grounded in Indian judgments.** The test is not the demo, it is whether the tool shows the paragraph its answer came from and whether that paragraph exists. [Choosing an Indian case-law search engine](/blog/choosing-indian-case-law-search-engine) and [native legal AI versus generic GPT](/blog/native-legal-ai-india-vs-generic-gpt) cover the criteria.

**Months 4 to 6. Run it on live matters with verification mandatory.** Every citation checked against the judgment, every check recorded. This is the phase most firms skip and the one that determines whether anything else works.

**Months 7 to 9. Write the policy and the disclosure practice.** Given the draft Supreme Court regulations and the High Court policies already in force, your filings need a settled position on when AI use is disclosed and who signs. Build it before a court asks.

**Months 10 to 12. Extend and measure.** Only 18 percent of organisations track return on investment, per Thomson Reuters. Measure hours on research per matter, citation error rate caught at verification, and turnaround time from instruction to first draft.

Four questions separate a usable system from a demo: does it retrieve from Indian judgments rather than predict them, does it show the source paragraph, does it flag overruled authority, and where does your data sit. [Niyam](https://niyam.ai) answers those four in its product documentation, and any vendor you evaluate should answer them in writing.

---

## Frequently asked questions

### Does the 92 percent AI adoption figure apply to Indian lawyers?

No. The 92 percent figure comes from the 2026 Wolters Kluwer Future Ready Lawyer Survey, which interviewed 810 lawyers in the United States, China and nine European countries: Germany, the Netherlands, the United Kingdom, Belgium, France, Italy, Spain, Poland and Hungary. India was not in the sample. The figure is a useful directional signal for where the profession is heading globally, but it is not a measurement of Indian practice and should never be quoted as one.

### Is there a reliable survey of AI adoption among Indian lawyers?

No published survey gives a defensible India-only adoption percentage as of August 2026. Manupatra has run a survey titled "AI Adoption and Its Impact in the Legal Industry: An Indian Perspective", but the open summary on Bar and Bench describes findings qualitatively without publishing headline percentages. If someone quotes you an Indian adoption number, ask which survey, what sample size, and what question was asked.

### How much time does AI actually save a lawyer?

The 2026 Wolters Kluwer survey found 62 percent of respondents reporting time savings between 6 percent and 20 percent of their work week. That is a self-reported range, not a measured outcome, and it covers lawyers in the surveyed Western and Chinese markets. On a fifty-hour week the band works out to roughly three to ten hours. Most of that saving lands in research and first-draft work.

### What are the biggest barriers to AI adoption in law firms?

Wolters Kluwer's analysis of its 2026 survey names three: ethical concerns related to AI and data privacy at 39 percent, inadequate training at 39 percent, and resistance to change at 35 percent. None is a technology problem. The Thomson Reuters 2026 data points the same way, finding that stalled AI strategies are most often blamed on tools not being in place and on people not being trained rather than on the capability of the software.

### What is the AI value gap?

It is the finding in the Thomson Reuters 2026 Future of Professionals report that 91 percent of professionals say their organisation is falling short of what AI could deliver. The report attributes the gap to the organisations deploying AI rather than to the technology. In organisations with a named AI strategy, 66 percent said AI was meeting or exceeding expectations for creating value. Where there was no active strategy, that fell to 22 percent.

### What is shadow AI and why does it matter for a law firm?

Shadow AI is the use of AI tools an organisation has not sanctioned, or use of sanctioned tools in ways it cannot see. Thomson Reuters found more than a third of professionals doing this. For a law firm it creates an unmanaged confidentiality exposure, because client data moves onto infrastructure the firm has no contract with and cannot audit. Providing an approved, usable tool is the practical control.

### Does using AI mean I have to tell the court?

If the Supreme Court's Draft Regulations for Use of Artificial Intelligence in Courts, 2026 are adopted in their current form, yes. The draft, released on 3 June 2026, requires a party or counsel who used AI in preparing a filed document to disclose that at the time of filing, and places responsibility for fabricated content on the filer. These are draft provisions open for comment, not yet binding. Individual High Courts have their own positions, so check the forum you are appearing before.

### Can a judge in India use AI to decide a case?

No. The Kerala High Court's policy of 19 July 2025 prohibits AI software from being used to arrive at any finding, relief, order or judgment in the district judiciary, defines AI as assistive only, and leaves responsibility for the content and integrity of a judicial order with the judge. The Supreme Court's draft regulations draw the same line between assistive and decisional use.

### What did Rishabh Shroff say about AI and junior lawyers?

Speaking at the India Today Conclave on 8 March 2025, Rishabh Shroff, Partner at Cyril Amarchand Mangaldas, said: "What would take a young lawyer 20, 30, 40, 50 hours to do? AI can do it in five seconds." In the same session he described AI as disrupting labour arbitrage and raised its consequences for hiring, promotion and mentorship in law firms, and noted the absence of Indian language models built for law firms.

### Has the work of legal research changed, or just the speed?

The work has changed. Finding candidate authorities is now fast. Confirming that the authorities exist, say what a summary claims, and remain good law is now the bottleneck. A tool that retrieves from an indexed corpus and shows the source paragraph makes that confirmation quick. A tool that generates text and attaches citations as decoration turns it into a separate research task.

### Does India have a law regulating AI?

Not as of August 2026. AI in India is governed through existing statutes, including the Information Technology Act, 2000, the Digital Personal Data Protection Act, 2023, the Bharatiya Nyaya Sanhita, 2023, the Copyright Act, 1957 and the Consumer Protection Act, 2019. The Ministry of Electronics and Information Technology released the India AI Governance Guidelines on 5 November 2025 as a principles-based framework. The proposed Digital India Act has remained in draft since 2023.

### How many organisations actually measure whether AI is working?

Very few. The Thomson Reuters 2026 AI in Professional Services Report found that only 18 percent of professionals say their organisation tracks return on investment from AI, and another 40 percent do not know whether it is measured at all. Firms that measure hours saved per matter and citation error rates caught at verification put themselves in a small minority and gain the evidence to expand or stop.

### Are clients asking law firms to use AI?

Yes, and there is a gap. The Thomson Reuters 2026 Future of Professionals report found 78 percent of corporate clients saying AI-enabled quality improvements from their providers are very important or essential, while only 6 percent said most or all of their providers actually deliver them. Expectation has moved faster than delivery, which is a commercial risk for firms that have not started.

### Should a small Indian practice wait until the tools mature?

Waiting has a cost that is now visible. Client expectations have already moved, courts have already started setting rules on disclosure and verification, and the entry price for purpose-built Indian legal AI is low. The cautious position is not to abstain, it is to adopt one function, verify everything against primary sources, and expand only once the verification habit holds.

### What should I ask a legal AI vendor before buying?

Four questions. Does the system retrieve from Indian judgments or generate text that resembles them. Does it show the specific paragraph an answer relies on. Does it flag whether an authority has been overruled or distinguished. Where is client data stored and processed, and what are the retention terms under the DPDP framework. Get the answers in writing.

### Will AI reduce the number of junior lawyers Indian firms hire?

The pressure is real but unresolved. Shroff described AI as disrupting labour arbitrage, which is the economic basis of the pyramid model. Thomson Reuters found 48 percent of professionals worried about AI's effect on the development of independent judgment and 71 percent saying early-career professionals need structured support from experienced peers. Firms that compress the work without rebuilding the training will run out of promotable lawyers before they run out of work.