A practical guide for busy professionals who want to leverage AI for legal drafting without compromising quality. Learn how to give prompts to create a first draft of the Shareholder’s Agreement by feeding the term sheet to ChatGPT.
Table of Contents
Introduction
“Rohit, can you come to my office for a minute?“
I looked up from my laptop to see Mr. Khanna, our senior partner, standing by my cubicle with that familiar expression – the one that usually meant I was about to get another research assignment that would keep me here past midnight.
“Sure, sir.” I grabbed my notepad, expecting the usual drill.
But when I walked into his corner office, he gestured toward the chair across from his desk with an unusual request. “I need you to teach me something.“
In my two years at the firm, Mr. Khanna had never asked me to teach him anything. He was the guy who could draft a shareholders’ agreement blindfolded, the partner whom clients specifically requested by name. What could I possibly teach him?
“I keep hearing associates talk about using ChatGPT for legal work,” he said, leaning back in his chair. “Yesterday, I overheard you telling Amit that you used it to draft something in twenty minutes that would normally take hours. My curiosity is piqued.“
I could not help but smile. “You want to learn how to use AI for legal drafting?“
“Do not look so surprised,” he chuckled. “I may be fifty-five, but I’m not ancient. We have that new tech startup client coming in tomorrow – TechFlow Solutions. They have already negotiated a term sheet with TechVenture Partners for their Series A round. Instead of me spending my evening doing the initial draft of their shareholders’ agreement, why don’t you show me how this AI magic works?“
This was my moment. The senior partner who had been practicing corporate law since before I was born wanted to learn from me.
“Alright, Mr. Khanna,” I said, pulling out my laptop. “Let me show you how we can draft a comprehensive shareholders’ agreement from a term sheet using nothing but smart prompting and a little AI assistance. But I need you to understand something first – this is not about replacing your legal expertise. It is about amplifying it.“
He nodded, genuinely intrigued. “Show me.“
What you need before you start
“Before we dive into ChatGPT,” I said, opening a fresh document on my laptop, “we need our roadmap. In the old days, you would gather all the commercial terms through endless client calls and negotiations. But now? The term sheet does all that heavy lifting for us.“
Mr. Khanna leaned forward. “Right, the term sheet. TechFlow already has one finalised with their investor.“
“Exactly!” I pulled up the TechFlow term sheet on my screen. “Look at this – we have TechFlow Solutions, a B2B SaaS company, taking INR 15 crore from TechVenture Partners. The term sheet gives us everything: 5-member board structure, 65% drag-along threshold, 8% cumulative preferred dividend, weighted average anti-dilution, detailed reporting requirements, even the founder non-compete terms and ESOP structure.“
If you are not familiar with these terms or need a deeper understanding of what goes into a shareholders’ agreement, check out this comprehensive guide on how to draft a shareholders’ agreement and important terms around it, post-investment, before we dive into the AI prompting process.
“This sounds like the same information I would gather anyway,” he observed.
“That is the point! We are going to use this negotiated, agreed-upon term sheet and feed it directly to the AI. Think of it like this – the term sheet is our instruction manual, and ChatGPT is our drafting assistant.“
I turned my screen toward him, showing the detailed TechFlow term sheet. “Once we upload this term sheet to ChatGPT, it can generate clauses that are specific to this exact deal rather than generic template language. And I mean specific – down to the exact dollar amounts, vesting schedules, and industry-specific metrics.”
Mr. Khanna made a note. “Interesting. So the quality of output depends entirely on the quality of the term sheet.“
“Now you are getting it. The investors and founders already did the hard work of negotiating the commercial terms. We are just translating their business deal into bulletproof legal language that complies with current Indian law.“
Setting up your workspace
“Now for the technical setup,” I said, navigating to ChatGPT on my browser. “This part is simpler than you might think.“
Mr. Khanna adjusted his reading glasses. “I have to admit, I have never actually used ChatGPT before. Is it complicated?“
“Not at all. But there are a few tricks that make all the difference.” I opened a new chat window. “First, you will want the Pro plan – it is twenty dollars a month, but trust me, it is worth it for longer outputs. The free version cuts you off mid-sentence, which is useless when you are drafting legal clauses.“
“Twenty dollars? That is approx ₹2000, less than one hour of billable time.“
“Exactly! Now, here is where most lawyers go wrong.” I opened a fresh Google Doc alongside ChatGPT. “They try to generate the entire agreement in one go. That is like asking a junior associate to draft a 40-page document without any guidance. It does not work.“
“The secret is organisation. I break everything down section by section, save each prompt I use, and paste each output into separate documents before assembling the final version.” I showed him my screen. “This way, if I need to revise just one section, I do not have to regenerate the entire agreement.“
Mr. Khanna leaned closer. “You are treating it like a systematic legal research project.“
“Bingo! And here is another pro tip – I use clear headings for everything. ‘Section 3 – Share Capital,’ ‘Section 4 – Board Composition,’ ‘Section 10 – Founder Obligations.’ It keeps me organised and makes the final assembly much cleaner.“
“This is more methodical than I expected.“
“The best part? Once you create this system, you can reuse it for every shareholders’ agreement. The folder structure, the prompt templates, the organisation method – it becomes your AI drafting playbook.“
I opened one more document. “I also keep a running list of effective prompts. Think of it as building your own legal AI library. Ready to see how the actual prompting works?“
Author’s Note: Before we dive into the specific prompts, remember that AI prompting is an art form just like legal drafting itself. Every lawyer has their own writing style, and the same applies to prompting AI tools. The eight prompts I am about to show you represent my approach, but you will develop your own style with practice. The key principle never changes: quality input determines quality output. The more specific and detailed your prompts, the more sophisticated your results will be.
Prompting ChatGPT — Section by Section
“Alright, here is where the magic happens,” I said, cracking my knuckles theatrically. “But first, let me show you the wrong way to do this.“
I typed into ChatGPT: “Draft a shareholders’ agreement.“
Mr. Khanna watched as ChatGPT generated a generic, templated response. “That is… not very useful.“
“Exactly! It is like asking your junior to ‘draft something legal.‘ You get garbage.” I deleted the response. “Now watch this.“
I uploaded the TechFlow Solutions term sheet to ChatGPT first. “See what I’m doing? I’m giving ChatGPT the complete commercial context before I ask it to draft anything. The term sheet becomes our instruction manual.“
“Smart approach.“
“Now, instead of asking for everything at once, I break it down systematically. Each prompt builds on the term sheet details and targets one specific section. Here is my formula – I use eight strategic prompts to create a comprehensive agreement.“
“First, let us start with introduction & definitions.”
I typed exactly as I had it saved in my prompt library:
Prompt: Draft the introductory and definitions section of a shareholders’ agreement for TechFlow Solutions Private Limited, an Indian private company incorporated under the Companies Act, 2013. The agreement involves Series A investment by TechVenture Partners. Include definitions for Company, Shareholders, Equity Shares, Series A Preference Shares, Board of Directors, ROFR, Tag-Along, Drag-Along, Liquidation Preference, and Anti-dilution.
The output was immediate and comprehensive.
“See how specific that was? The company name, jurisdiction, exact legal framework, specific investment context, and precise definitions required based on the term sheet. Notice I asked for ‘comprehensive’ definitions – that is key for a sophisticated agreement.“
You can click here to access the ChatGPT chat and see how it generated clauses based on the subsequent prompts in the following part of the article.
“Next is share capital and issuance rights.”
Prompt: Draft a clause describing authorised share capital of INR 10 crore divided into 1 crore equity shares of INR 10 each. Include issuance of 3,00,000 Series A Preference Shares at INR 500 per share to TechVenture Partners for INR 15 crore investment. Include language on future issuances, pre-emptive rights for existing shareholders, and anti-dilution protection using the weighted average (broad-based) method.
Another sophisticated clause appeared, incorporating mathematical formulas.
“Notice how I am getting the actual anti-dilution formula, not just a description. The AI can handle complex legal mathematics when you are specific.“
“Then governance and board composition.”
Prompt: Create a clause on board composition for a 5-member board: 2 founder directors (Rahul Sharma as CEO and Priya Patel as CTO), 2 TechVenture Partners nominee directors, and one independent director with supply chain industry background. Include appointment procedures, quorum of 3 directors (with at least one founder and one investor nominee), decision-making process, and board observer rights for TechVenture Partners.
Mr. Khanna leaned forward as an extensive governance section was generated. “This is remarkably thorough. It includes board observer rights, specific quorum requirements, and even meeting notice periods.“
Click here to see the clause ChatGPT produced.
“Now the complex part – transfer restrictions.”
Prompt: Write a clause covering share transfer restrictions for TechFlow Solutions, including Right of First Refusal (ROFR) with 15-day exercise period, Tag-Along rights for transfers of 10% or more shareholding, Drag-Along rights requiring 65% or more of total share capital, and exceptions for transfers to affiliates, family members, and ESOP issuances.
“Perfect. See how I am building this systematically? Each section builds on the previous one, and everything flows from the term sheet. I specifically mentioned ‘comprehensive’ and ‘detailed’ because I want professional-grade output.“
“Financial reporting and dividend policy – this gets SaaS-specific.”
Prompt: Generate a clause on dividend policy and financial reporting for TechFlow Solutions. Include 8% cumulative preferred dividend on Series A shares (if declared), prohibition on common share dividends until preferred obligations are met, audited annual accounts within 90 days, quarterly MIS within 30 days including revenue metrics and customer acquisition data, and monthly operational updates on customer count, ARR growth, burn rate, and cash position.
Mr. Khanna watched the screen as a sophisticated financial governance clause appeared. “It includes specific SaaS metrics – ARR growth, churn rates, and burn rate. That is incredibly industry-specific.“
“That is because the term sheet told us exactly what TechVenture Partners negotiated for their B2B SaaS investment. The AI understands the business context.“
Click here to see the clause ChatGPT produced.
“Exit Rights – this is where it gets really sophisticated.“
Prompt: Draft an exit clause for TechFlow Solutions with three options: IPO with registration rights for Series A shareholders, third-party sale (with drag-along and tag-along mechanisms), and company buyback at fair market value subject to Companies Act compliance. Include 1x non-participating liquidation preference for Series A shares, ensuring TechVenture Partners receives INR 15 crore or pro-rata share of proceeds, whichever is higher.
“See how it is structuring multiple exit scenarios with specific valuation mechanisms?” I pointed at the screen. “In a traditional template, you would get one generic exit clause. Here, we are getting sophisticated exit mechanics with actual liquidation waterfalls because I specified exactly what was negotiated.“
“Dispute resolution with recent legal developments.”
Prompt: Create a dispute resolution clause for TechFlow Solutions with mediation first for 30 days, then binding arbitration where each party appoints one arbitrator and they jointly appoint a presiding arbitrator within 15 days. Arbitration seated in Mumbai under the Arbitration and Conciliation Act, 1996, conducted in the English language.
“Perfect escalation structure,” he noted as the clause generated. “And it even addresses the recent Supreme Court concerns about arbitrator appointments – that is current legal awareness.“
“Founder obligations – this is crucial for startups.”
Prompt: Draft a comprehensive founder obligations section for TechFlow Solutions including: (1) non-compete obligations for Rahul Sharma and Priya Patel during tenure and 2 years post-exit specifically in supply chain/logistics SaaS sector, (2) non-solicitation of employees, customers and business partners, (3) key man insurance requirements of INR 5 crore each for CEO and CTO with Company as beneficiary, and (4) full-time commitment obligations with specific tenure requirements.
“This is a separate section entirely,” Mr. Khanna observed. “Most basic agreements do not have this level of founder-specific obligations.“
“Finally, ESOP and miscellaneous provisions.”
Prompt: Write standard miscellaneous clauses for TechFlow Solutions shareholders’ agreement, including Governing Law (India under Companies Act 2013), Entire Agreement, Notices, Severability, Counterparts, founder non-compete obligations (2 years post-exit in supply chain/logistics SaaS space), key man insurance requirements (INR 5 crore each for CEO and CTO), and 15% ESOP pool creation with 4-year vesting and 1-year cliff.
I leaned back as the final section appeared. “There you have it. Eight strategic prompts, eight comprehensive sections, all flowing directly from the term sheet. Total time? About fifteen minutes of actual prompting.“
Click here to see the clause ChatGPT produced.
Mr. Khanna was quiet for a moment, scrolling through all the generated sections in my document. But as he read deeper, his expression changed from impressed to amazed.
“Rohit, this is remarkable. Your eight prompts generated exactly what I expected, but with much more sophistication than typical AI output.”
He was scrolling through multiple sections now.
“The AI essentially took my eight strategic prompts and executed them at a senior associate level,” I said, genuinely impressed.
“Instead of giving me basic template clauses, it generated sophisticated legal language that properly implements everything I requested from the term sheet.“
“And the quality?” Mr. Khanna asked.
“It is… it is actually senior associate level work, maybe even junior partner quality.” He paused.
“This would have taken me four to five hours to draft from scratch, and I’m not sure I would have included this level of mathematical precision in the anti-dilution formulas or this degree of procedural detail in the transfer restrictions.“
“Ah,” I grinned, “that is where the lawyer’s brain really kicks in. The AI did the heavy lifting, but now we need to do what only we can do – make it coherent, consistent, and legally bulletproof.“
Assembling the final document
“This is where your legal training becomes crucial,” I said, opening a new master document. “The AI gave us eight comprehensive sections that perfectly address what I prompted for, and they are sophisticated. But they are not talking to each other yet.“
“We need to be the conductor of this orchestra.” Mr. Khanna pulled his chair closer.
“What do you mean?“
“Watch.“
I began copying the responses generated by ChatGPT to a Word file.
“First, I paste everything in order – Introduction, Definitions, Share Capital, Board Composition, Transfer Restrictions, Financial Reporting, Exit Rights, Dispute Resolution, and Miscellaneous.”
“Note that the 10. FOUNDER OBLIGATIONS and 11. ESOP clause as provided by ChatGPT must come before the Miscellaneous clause.”
“Anyways, the good news is that ChatGPT maintained excellent consistency because we fed it the term sheet context upfront.“
Click here to see the first draft of the Shareholder Agreement.
I scrolled through the assembled document, showing him the comprehensive 11-page shareholders’ agreement.
The critical review phase- where legal expertise becomes essential
“Now comes the really important part, sir,” I said, assembling the sections into a master document. “The AI gave us sophisticated clauses, but this is where your experience becomes absolutely crucial.“
Mr. Khanna leaned forward, studying the screen. “You mean the legal review?“
“Exactly. We have completed phase one – the structural drafting. But phase two requires the kind of legal judgment that only comes from years of practice.“
Where AI limitations become apparent
I scrolled through our assembled document. “The AI generated professional-grade language for everything I requested. But look at this board composition clause – you would immediately spot whether appointing two nominee directors and removing them ‘at any time by written notice’ actually complies with section 161 of the Companies Act.“
Mr. Khanna examined the clause carefully. “The board structure looks reasonable, but…”
“But you are thinking about the removal procedures, are you not? Section 169 requires a special resolution for the removal of directors, and there are specific notice requirements outlined in section 169(1). The AI included clean language about removal ‘by written notice’ but missed the statutory compliance requirements.”
He nodded thoughtfully. “And these share transfer restrictions – the 15-day ROFR period looks standard, but does it align with section 58 requirements for private company share transfers?”
I pointed to the transfer section. “Exactly! The AI-generated sophisticated ROFR language, however, does not supersede section 58 of the Companies Act, which specifically governs private company share transfers. The clause mentions ‘Transfer Notice’ procedures, but we need to verify it complies with the statutory framework for restricted transfers.”
The review that requires senior judgment
“This is where your experience matters most, sir,” I said. “The AI cannot cross-check against Companies Act compliance the way you naturally would.“
Mr. Khanna was already taking notes. “Legal compliance first – do the board appointment procedures follow section 161? Do the transfer restrictions properly invoke section 58? Then commercial reality – will these mechanisms actually work when founders want to exit?”
“Exactly! You are already spotting the statutory gaps the AI missed.”
Where decades of practice make the difference
“The AI has never had to explain to the ROC why a director removal did not follow proper procedures,” I observed. “But you know which provisions will create compliance headaches during actual implementation.“
Mr. Khanna studied the governance section. “This observer rights clause for TechVenture looks comprehensive, but observers are not recognised under the Companies Act framework. We need to structure this carefully to avoid unintended director liability.“
“And the drag-along provision requiring 65% approval…” he continued.
“You would want to cross-check that against section 230 merger provisions to ensure it does not accidentally trigger different statutory requirements,” I finished. “That is exactly the kind of statutory knowledge the AI cannot apply.“
The senior lawyer’s advantage
“In twenty minutes, we generated a sophisticated first draft,” I said. “But it is not ROC-ready until someone with your experience reviews it for Companies Act compliance, procedural accuracy, and regulatory alignment.“
Mr. Khanna nodded approvingly. “The AI accelerated the structural drafting dramatically. But ensuring statutory compliance – that still requires someone who actually files these documents and deals with registrar queries.“
“Which is why AI makes the perfect assistant, sir. It handles the complex clause architecture so you can focus on the legal compliance that actually matters for filing and enforcement.“
Final note: the evolution of legal practice
Six months later, Mr. Khanna had become the firm’s most vocal advocate for AI-assisted drafting. During a partner’s meeting, he shared our TechFlow experience.
“The AI handled the mechanical complexity beautifully,” he explained to the room. “But what surprised me was how much better my legal advice became when I was not bogged down in drafting mechanics.”
He paused, looking around at his colleagues. “I could focus entirely on statutory compliance, commercial risks, and strategic considerations. The clients got both speed and superior legal protection.“
The senior partner nodded thoughtfully. “So, are you saying AI made you a better lawyer?“
“Not exactly,” Mr. Khanna replied. “AI gave me more time to be the lawyer I have always been – just without the drudgery.“
As I listened from the back of the room, I realised we had stumbled onto something profound. The future was not about AI replacing lawyers or lawyers resisting AI. It was about intelligent collaboration.
The best legal documents are not generated by humans or AI alone – they are crafted by both, working in perfect partnership.
Allow notifications
