AI in Practice: Research, Drafting, Review, Discovery, Access to Justice
There is no single "legal AI" — there is a right tool for each task, and a wrong tool that gets you sanctioned. Learn to match them.
The hook
“A generic chatbot and a grounded research engine can answer the same question with the same confidence — but only one of them can show you where the answer came from. In practice, the gap between those two is the gap between competent lawyering and a misconduct finding.”
What you'll be able to do
- Survey the real legal-AI landscape across five workflows — research, contract drafting/review, discovery, due diligence, and access to justice — including the Indian tool stack and the judiciary's own systems.
- Operationalize the core skill of this module: match the tool to the task, choosing on the basis of grounding, provenance, and verifiability rather than fluency.
- Distinguish generative chatbots from grounded, citator-backed research engines, and explain why provenance is the deciding factor for legal reliability.
- Recognize the practical realities of legal-AI adoption — where AI beats the lawyer baseline (document Q&A, summarization) versus where humans still win (contract redlining), and the commercial 'credit trap' of subscription tools.
In short
Module 6 turns from how AI fails and how to prompt it to where AI actually gets used in legal practice. Students survey real workflows and tools — legal-research engines, contract drafting and review, e-discovery with Technology-Assisted Review, due diligence, and access-to-justice applications — including the Indian commercial stack and the Supreme Court's own judiciary systems. The organizing skill throughout is "match the tool to the task": choosing on grounding, provenance, and verifiability, because the whole module is about using the right AI well.
The AI bridge
The whole module is "use the right AI well." Matching the tool to the task is itself a responsible-use skill: choosing a grounded, citator-backed engine over a generic chatbot for case-law research is how you build provenance and verifiability into the workflow from the start — and avoid being the lawyer in Mata who trusted a tool that had no connection to the databases it cited.
In this module
- 01
The legal-research engine landscape: grounded, citator-backed tools (Thomson Reuters CoCounsel; Lexis+ with Protege; vLex Vincent) and the Indian stack (Manupatra AI, SCC Online's conversational assistant, CaseMine/AMICUS, LegitQuest, Indian Kanoon, VIDUR AI, BharatLaw.AI) — and why grounding and citators are the feature that matters for safe AI use, not fluency.
- 02
Contract drafting, review, and due diligence tools (Spellbook, Luminance, Kira, DraftWise) — and the practical reality that AI can beat the lawyer baseline on document Q&A and summarization, while humans still win on contract redlining, so you must match the tool to the task and verify accordingly.
- 03
E-discovery and Technology-Assisted Review (TAR): predictive coding as an established, judicially approved use of AI in litigation (Da Silva Moore v. Publicis Groupe, S.D.N.Y. 2012) — a model of AI deployed transparently and defensibly, which is exactly how responsible legal-AI use should look.
- 04
Access to justice: legal-aid chatbots and document automation that widen access — paired with the responsibility caution that a fabricating or sycophantic model harms exactly the unrepresented users who can least afford to verify it.
- 05
The Indian judiciary's own AI stack (SUPACE, SUVAS, TERES, LegRAA), framed as 'AI assists, never decides; the judge owns the integrity of the decision' — the institutional version of the human-in-the-loop discipline the course teaches.
- 06
Practical realities of adoption: the generative-vs-grounded distinction decides reliability; the subscription 'credit trap' shapes which tools you can actually run; and provenance plus a verification step is non-negotiable whichever tool you pick — the right AI, used well, with its sources checkable to source.
The interactive demos
Every idea is a Mirror Move
Run it on the room, show it inside the machine, prove it live on a real AI, then name the skill.
Generic chatbot vs. grounded tool: the provenance gap
On us
Poll the room: ask students to commit to which they would trust more for an Indian case-law question — a fluent general chatbot answer or a grounded research engine's answer — and capture the split before revealing anything.
In the machine
A generative chatbot predicts a plausible-looking answer with no connection to the databases it appears to cite, while a grounded, citator-backed engine retrieves from real sources; same confidence on the surface, opposite reliability underneath.
Live AI
Run the same research question live through a generic chatbot versus a grounded tool, and compare the provenance and citations each returns — show whether each citation can be traced to a real, verifiable source.
The skill
Match the tool to the task: for legal research, choose a grounded, citator-backed engine and verify provenance to source — fluency is not authority.
The lab
Tool Fit
Students take one realistic legal matter and break it into sub-tasks (e.g., research the point of law, summarize the documents, draft a clause, review/redline a contract, run discovery). For each sub-task they decide which tool or tools they would use — generic chatbot, grounded research engine, contract tool, TAR/e-discovery, or the judiciary's approved systems — and justify the choice on grounds of provenance, grounding, and how they would verify the output.
Deliverable
A tool-fit map for the matter: each sub-task paired with the chosen tool(s), an explicit justification of provenance and grounding, and a verification step for each output (how the student would check it to source before relying on it).
Key sources & cases
Da Silva Moore v. Publicis Groupe (S.D.N.Y. 2012)
US case in which a court gave judicial approval to predictive coding (Technology-Assisted Review) in e-discovery — the module's example of AI deployed transparently and defensibly in litigation.
Thomson Reuters CoCounsel; Lexis+ with Protege; vLex Vincent
Grounded, citator-backed legal-research engines surveyed as the provenance-first alternative to generic chatbots.
Indian research stack: Manupatra AI, SCC Online conversational assistant, CaseMine/AMICUS, LegitQuest, Indian Kanoon, VIDUR AI, BharatLaw.AI
The India-first set of legal-research and case-law tools; grounding and citators are the deciding feature.
Contract tools: Spellbook, Luminance, Kira, DraftWise
Contract drafting, review, and due-diligence tools surveyed for the drafting/review workflow.
Indian judiciary stack: SUPACE, SUVAS, TERES, LegRAA
The Supreme Court of India's own AI systems — framed as 'AI assists, never decides; the judge owns the integrity of the decision.'
2026 India tool guides and comparative benchmark
Sources named in the brief: CaseMine, iPleaders, LiveLaw 'AI in Courts'; and the comparative benchmark (CoCounsel/Harvey/Lexis) showing task-by-task strengths.
Readings
- Brief Section 6, Module 6 entry — legal-research engines, contract/review/due-diligence tools, e-discovery and TAR (Da Silva Moore v. Publicis), access to justice, and the Indian judiciary stack (SUPACE, SUVAS, TERES, LegRAA)
- Brief Section 4 — the CT-to-legal-reasoning mapping (decomposition/IRAC underpins breaking a matter into tool-matched sub-tasks)
- 2026 India tool guides: CaseMine, iPleaders, LiveLaw 'AI in Courts'
- The comparative benchmark (CoCounsel/Harvey/Lexis) showing task-by-task strengths
- Richard Susskind, Tomorrow's Lawyers and Online Courts and the Future of Justice (context for access to justice and changing workflows)
Next module
Module 07 / 08
Law OF AI: Bias, Evidence, Liability, IP, Data Protection
law-of-AIBring a credit-bearing AI course to your students
A 16-hour course that treats using AI well as a professional duty — one a council can approve, and a graduate can defend in court.