Edition Eight — learn to draft before you let a machine draft for you

It happens at every session. I can almost time it.

Twenty minutes in, a hand goes up somewhere near the back, and the question arrives. Sir, how do you see the use of AI in law practice?

Then the rest come faster, because someone has broken the ice. Is it all right if we draft agreements using AI? What about pleadings? Can a legal due diligence be done this way?

I have been asked this in more rooms than I can count, and I have started noticing the tone. It is not curiosity. It is anxiety wearing the clothes of curiosity.

• • •

And I do not blame a single student for asking.

Look at what the market is telling them. Crores of cases are pending before our courts. Every legal technology company alive has looked at that number and seen a gap wide enough to build an industry inside. The pitch is everywhere — faster, cheaper, at scale. A twenty-two-year-old reading that will reasonably conclude that the profession she is entering may be done by machines before she turns thirty.

So the question is fair. My answer is simply not the one that is expected.

Before any argument about whether these tools are good or bad — and I have views — here is my plain instruction.

Learn to draft first.

Study contracts. Real ones. Study pleadings and filings until you can see their architecture. Understand why a clause sits where it sits, what it is guarding against, and what happens when it is missing.
Then use the tools and use them well. But in that order.
Let me show you what happens when the order is reversed.

• • •

What happens when the order is reversed

Last month a term sheet reached me for a private equity investment. Eight pages.

The fund manager knows me. He rang and said he had asked his lawyer to simplify it and make it shorter. Then he laughed and said — my lawyer has probably used some AI for this.

Quite likely, I said.

The revised version came back at six pages. Six, from eight, after an express instruction to make it short.

Why? It carried a list of twenty-six items requiring affirmative votes. Twenty-six. And a good number of them made no sense for this transaction at all — they had been imported from every other term sheet in existence and dropped in. Then came an elaborate right of first refusal, and a full deadlock resolution mechanism, both worked out in detail.

Neither belongs in a term sheet. Those are matters for the definitive agreements. A term sheet exists to record the commercial understanding — what is being bought, at what price, on what conditions, with what protections in principle. It is short precisely because it is not the contract.

• • •

Now notice what actually went wrong

Nothing in that document was incorrect.

The clauses were competently written. The right of first refusal was a real one. The affirmative vote list was a genuine list of the kind that appears in investment documents. Hand it to someone who has never sat through a negotiation and it looks impressive — thorough, careful, professional.

That is the trap.

The failure was not accuracy. It was judgment. The drafter could not tell what did not belong.

And that is the whole of drafting. Not what you put in — what you leave out, and knowing which document a thing belongs in.

A model trained on every term sheet ever written will hand you the average of all term sheets. But no transaction is average. Yours has a particular investor, a particular promoter, and a particular fear sitting on each side of the table. The protection that matters most in your deal may appear in none of the others.

Only someone who knows the basics can see that.

And here is the part that should concern you most. These tools turn every lawyer into a reviewer. A junior who cannot draft cannot review. Output will land on your desk from your first week, and you will be asked whether it is right — and you will have no way of knowing.

• • •

A practical exercise

So here is a practical exercise. Do it properly and you will be ahead of most of your batch.

Take something simple. A Leave and License Agreement.

One.

Ask an AI model for a detailed memo on the structure of a Leave and License Agreement, with an explanation of the critical clauses. Read it as a student, not as a consumer. For every clause, ask why it exists.

Two.

Now instruct it to prepare three drafts. One entirely in favour of the Owner. One entirely in favour of the Licensee. One neutral. Put them side by side and watch what moves — the deposit, the lock-in, the notice period, the repairs, the indemnity. That comparison will teach you more about negotiation than a semester of lectures.

Three.

Now do the work yourself. Take one of those drafts and mark it up by hand as though you were acting for the other side. For every change you make, ask what you are protecting against.

Then repeat the whole exercise across ten good commercial agreements. Ten. Not two.
At the end of it you will have built the one thing no model can build for you — the judgment that lets you use the model.

• • •

I am not against these tools. I use them. The profession you are joining will use them far more than mine ever did, and the lawyer who refuses them will be left behind by the lawyer who does not.

But be clear about which lawyer you are training to become. Not the one who produces the draft. The one who can look at a draft and say: this is six pages and it should be two, these twenty-six items should be four, and this clause belongs in an entirely different document.

That lawyer is not replaceable. The other one always was.

So here is my question for you. When the output lands on your desk — and it will, sooner than you think — will you be able to tell whether it is right?

— Nitin

• • •

Connecting Dots — every fortnight, I take a deal, a policy, or a person and tell you the thinking behind it. Insights and instincts.

I’m a corporate and M&A lawyer in Mumbai. For three decades I’ve advised global companies and their boards on coming into India, on joint ventures, and on mergers and acquisitions, and I still do. Busy as that keeps me, I always make time for younger professionals — through this letter, and through my conversations with law students — because staying close to young minds keeps me learning. If something here struck a chord, write back. I read every reply.