For most of the last decade, "legal tech" meant a handful of narrow tools: e-discovery platforms, contract repositories, and practice-management software. Generative AI changed the conversation almost overnight. Drafting, summarising, research and first-pass review — tasks that once consumed the early years of a lawyer's training — can now be attempted by a model in seconds. The question for law firms is no longer whether to engage with this technology, but how, and under what safeguards.

Where AI is genuinely useful today

The strongest use cases share a pattern: high volume, repeatable structure, and a human who can verify the output quickly.

• Document review and due diligence. Flagging change-of-control clauses, termination rights or unusual indemnities across hundreds of contracts is exactly the kind of pattern-matching these systems handle well.
• First drafts and clause libraries. Producing a workable first draft of a routine agreement, NDA or notice — which a lawyer then edits — saves time without removing judgment.
• Summarisation. Condensing long judgments, deposition transcripts or regulatory circulars into structured briefs for a partner or client.
• Knowledge management. Making a firm's own precedents, opinions and past work product searchable in plain language rather than by keyword.

The risk that made headlines

The cautionary tale most lawyers now know is Mata v. Avianca (S.D.N.Y. 2023), where counsel filed a brief citing cases that a chatbot had simply invented, and were sanctioned. The lesson is not that AI is unusable, but that a fluent answer is not a verified answer. Large language models predict plausible text; they do not check a citator. Any workflow that places model output in front of a court or client without independent verification is a professional-conduct incident waiting to happen.

The unit of accountability in a law firm is still the lawyer who signs. Technology can change who drafts; it cannot change who answers.

Confidentiality and data protection

Client confidentiality is the second fault line. Pasting a privileged document into a public AI tool may amount to disclosing it to a third party and, depending on the provider's terms, may feed it into future training. Firms need to ask hard procurement questions: Where is data stored and processed? Is it used for model training? Who at the vendor can access it? Can it be deleted on request?

In India, the Digital Personal Data Protection Act, 2023 adds a statutory layer for any matter involving personal data, while professional-secrecy duties under the advocate–client relationship apply regardless. An enterprise deployment with contractual data-handling guarantees is a very different proposition from a free consumer chatbot — and firm policy should say so explicitly.

What a sensible firm policy looks like

1. Approved tools only. Maintain a short list of vetted systems; prohibit client data in anything else.
2. Mandatory verification. Every citation, quotation and statement of law generated with AI assistance must be checked against a primary source before it leaves the firm.
3. Clear supervision. Define who may use AI for what, and who reviews the result. Treat AI output like work from a very fast, very confident junior with no accountability.
4. Client transparency. Disclose material AI use where engagement terms, regulation or client expectations require it — and be ready to explain billing when a task that took six hours now takes one.
5. Training. The biggest risk is not the tool; it is the lawyer who does not understand its failure modes.

The business-model question

The deeper disruption may be economic rather than technical. If the billable hour rewards time spent, then a tool that compresses time threatens revenue. Firms that respond by hiding efficiency gains will lose trust; those that move toward fixed fees, value-based pricing and higher-order advisory work are more likely to turn productivity into advantage. Likewise, the traditional apprenticeship model — juniors learning by doing document review — needs rethinking. If the routine work disappears, firms must deliberately design other routes for young lawyers to build judgment.

Regulation and the courts are catching up

Courts and bar bodies around the world are issuing guidance on AI use, from disclosure requirements for filings to restrictions on its use in judicial decision-making. India's courts have also experimented with technology for translation and case-management support, while stressing that adjudication itself remains a human function. Expect the rules to keep evolving — firms should assign someone to track them.

Conclusion

AI will not replace lawyers, but lawyers who use it carefully will outpace those who ignore it, and those who use it carelessly will find out why professional-responsibility rules exist. The winning posture is neither hype nor prohibition: adopt where verification is cheap and value is high, protect client data relentlessly, and keep a human accountable for every word that goes out under the firm's name.

This post is written for general discussion and does not constitute legal advice.