Every commercial lawyer has watched the same scene. The commercial terms are settled, the schedules are agreed, everyone is tired, and someone opens the boilerplate to deal with the last few clauses. Governing law goes in without discussion. Notices are copied from the last deal. And then the arbitration clause, the provision that will govern how this relationship is unwound if it ever goes wrong, is agreed in about four minutes by two people who each privately believe it will never be needed. It is the clause negotiated last and litigated first, and the effort it receives is almost exactly inverted from the effort it deserves.
The cost of that inversion is not theoretical, and it is not measured in legal fees alone. A defective arbitration clause buys you satellite litigation: an application to appoint an arbitrator because the mechanism you wrote has failed, a challenge to the arbitrator you appointed because the mechanism you wrote was impermissible, a fight about which court supervises because you used the words seat and venue as though they were synonyms, a jurisdictional objection because you carved out some disputes and never defined the boundary. Every one of those is a proceeding that happens before anybody looks at the merits of your actual dispute, and it is paid for out of the same budget, argued by the same counsel, and reported to the same board.
This article is a drafting guide for Indian in-house counsel and GCs, and it is written around a single argument: the arbitration clause deserves the drafting effort that its litigation probability actually justifies, not the effort its position in the document suggests. It covers what a workable clause must contain, the seat and venue distinction that determines who supervises you, the unilateral appointment defect that has quietly invalidated a very large number of clauses sitting in Indian contract portfolios today, the multi-tier and carve-out drafting that produces jurisdictional fights, and how to audit a legacy portfolio for all of it before a dispute finds it for you.
The proposition in one line
You are not drafting a clause about arbitration. You are drafting the procedural rules of a proceeding you cannot foresee, against a counterparty who will by then be hostile, to be read by a judge who was not in the room. Ambiguity you can live with in a commercial clause becomes a separate lawsuit in an arbitration clause, because the other side's incentive is no longer to make the contract work.
The Midnight Clause Problem
The arbitration clause is sometimes called the midnight clause, for the obvious reason. Three habits produce most of the damage, and all three are habits of a tired drafter rather than a careless one.
The first is copying. A clause is lifted from a previous contract, which lifted it from a template, which was drafted for a different transaction type, a different counterparty profile and, quite often, a different decade of Indian arbitration law. Nothing about it is examined, because it looks like the clause everyone uses. The second is compression. Both sides want to close, so precision is traded for speed, and precision in this clause is exactly what disputes about this clause consume. The third is optimism. Neither side wants to spend the last hour of a deal discussing how they will fight, so the clause is agreed in the tone of a formality, and formalities are drafted loosely.
Every hour saved on the arbitration clause is borrowed at a rate set by the person who later wants to avoid it.
There is a structural asymmetry worth naming. At the time of drafting, both parties want the contract to work, so both read the clause charitably. At the time of dispute, one party very often wants the contract not to work, and reads the clause looking for a defect. A clause is therefore always drafted under one set of incentives and construed under the opposite set. Good drafting simply means writing for the second reader.
What a Workable Clause Must Actually Contain
Under the Arbitration and Conciliation Act, 1996, Section 7 defines the arbitration agreement. That definition is the floor, not the target. A clause that satisfies Section 7 can still be so incomplete that everything about how the arbitration runs has to be resolved by a court before it starts. The elements below are the ones whose absence reliably produces a fight.
An unambiguous intention to arbitrate
The clause must show that the parties agreed to refer disputes to arbitration and to be bound by the award. Permissive language, discussed below, is where this most often fails. Write shall, not may, and do not mix arbitration language with language contemplating a court trial of the same dispute.
Scope, drawn deliberately
Which disputes go to arbitration. Broad formulations covering all disputes arising out of or in connection with the contract exist because narrow ones generate arguments about whether a particular claim falls inside. If you narrow the scope, do it because you mean to, and define the boundary in terms someone else can apply.
Number of arbitrators and how they are appointed
One or three, and a mechanism that works when the other side refuses to cooperate. The appointment mechanism is the single most defect-prone element in Indian contracts, and the reason has a name: unilateral appointment.
Seat, stated as the seat
The legal seat determines the supervisory jurisdiction over the arbitration. Use the word seat. Do not leave a court to infer it from a sentence about where hearings will be held.
Venue, if you want flexibility
If you want hearings to be capable of being held somewhere convenient, say so separately and say expressly that this does not change the seat. Venue flexibility is a legitimate commercial want; expressing it carelessly is how seats get argued about.
Language, rules and governing law
The language of the proceedings. Whether the arbitration is institutional, under named rules, or ad hoc under the Act. The governing law of the contract, and, where there is any international element, the law governing the arbitration agreement itself, which is a separate question and frequently unaddressed.
Two of those deserve a note. Institutional arbitration under a named set of rules imports an appointment mechanism, a fee schedule and an administrative apparatus, which removes a great deal of what otherwise has to be drafted and, more importantly, removes the need to go to court when the other side simply refuses to appoint. Ad hoc arbitration is perfectly workable and often cheaper, but it puts the entire burden of procedural design on your clause. Choose deliberately rather than by default, and if you choose ad hoc, draft as though you will have to run the process against a party who wants it to fail.
Before it leaves your desk, a clause should answer the following questions on its own face, without anyone having to construe anything:
- Is arbitration compulsory? If the word may appears anywhere near the obligation to refer, fix it.
- Which disputes are covered, and is the boundary applicable by a stranger? If classifying the dispute requires judgement, you have built a preliminary issue.
- How is the tribunal constituted if the other side does nothing at all? If the answer depends on their cooperation, the mechanism has not been designed, it has been assumed.
- Where is the seat, and does the clause say seat? If a court would have to infer it, it is not fixed.
- Which institution and which rules, by their exact current names? If neither, the clause has to carry the whole procedure itself.
- What language, and how many arbitrators? Two sentences that cost nothing now and cost an application later.
- Does it match the clause in every related contract in the same transaction? If not, you have licensed parallel proceedings.
Seat Versus Venue, and Why It Decides Who Supervises You
This is the distinction that costs Indian parties the most money for the least reason. The seat of an arbitration is a legal concept: it fixes the law governing the arbitration and identifies the courts with supervisory jurisdiction over it. The venue is a physical concept: where the hearings happen. They are frequently the same place, which is why so many drafters treat them as one thing, and they are not one thing.
Why it matters is entirely practical. Supervisory jurisdiction is what determines which court hears an application to appoint an arbitrator under Section 11, which court entertains an application for interim measures under Section 9, which court hears a challenge to the award under Section 34, and which court hears an appeal under Section 37. If the seat is contested, then the first thing that happens in your dispute is a fight about which court is even permitted to decide anything, conducted in at least two courts at once, before a single word has been said about the underlying claim.
The Indian jurisprudence on this is developed. Bharat Aluminium Co v Kaiser Aluminium Technical Services, decided in 2012 and universally called BALCO, is the foundational decision on seat and the reach of Indian courts. BGS SGS SOMA JV v NHPC Ltd (2019) addressed how a clause naming a place is to be read where the parties have used the language of venue, and how a designated place may be understood as the seat. The point for a drafter is not to master that case law. It is to notice that both decisions exist because clauses were drafted in a way that left the question open, and that a clause which says the seat of arbitration shall be Mumbai, and hearings may be held at any location the tribunal considers convenient, does not generate the argument at all.
A test you can apply in ten seconds
Read your clause and ask: if the counterparty filed a Section 9 application in a court on the other side of the country tomorrow, does this clause tell me, in words, why that court is the wrong one? If the answer requires you to construe a sentence about where meetings will be held, the clause has not fixed a seat, it has invited an argument about one.
The Unilateral Appointment Defect in Legacy Portfolios
If you audit one thing in your contract portfolio, audit this. It is the single most common defect in Indian legacy contracts, and it exists because of a drafting convention that was entirely standard for decades and is now a serious problem.
The convention was to give the stronger party control of the tribunal. Clauses said that disputes would be referred to the sole arbitration of the Managing Director of the employer, or of a person nominated by the Chairman, or to an arbitrator appointed by one named party. In infrastructure, construction, public procurement, distribution and lending documentation, this was ordinary. It reflected bargaining power and nobody found it remarkable.
Section 12(5) of the Arbitration and Conciliation Act, 1996, read with the Seventh Schedule, renders persons falling within specified categories of relationship with a party ineligible to act as arbitrator. The Supreme Court then addressed the consequence for appointment clauses of exactly this kind. TRF Ltd v Energo Engineering Projects Ltd (2017) concerned the position where a person who is himself ineligible to act as an arbitrator nominates one. Perkins Eastman Architects DPC v HSCC (India) Ltd (2019) developed the point in the context of unilateral appointment, where one party alone controls the constitution of the tribunal.
This is a portfolio problem, not a drafting problem
The reason this matters to a GC rather than to a drafter is arithmetic. Every long-form contract your organisation signed before this line of authority developed, and every contract signed since by someone working from an old template, may carry an appointment mechanism that cannot be operated. You will not find out at signature. You will find out at the moment you most need the tribunal constituted, when the counterparty declines to participate in an appointment they know is vulnerable, and you spend the first year of your dispute in an application under Section 11 instead of in the arbitration. The defect is silent, it is systemic, and it is discoverable today for the cost of reading your own contracts.
The remediation is unglamorous and mostly consists of not repeating it. In new drafting, avoid any mechanism in which one party alone controls the constitution of the tribunal, and prefer either institutional appointment under named rules or a genuinely balanced mechanism for a three member tribunal. In legacy contracts, identify the affected clauses, assess with counsel how the current position bears on each, and where the relationship is live and the counterparty is cooperative, fix it by amendment or in the next renewal rather than leaving it to be discovered in a dispute. The precise operative position on ineligibility, waiver and appointment continues to develop, so treat this section as a reason to review with counsel rather than as a settled statement of what your particular clause is worth.
Pathological Clauses, and What to Write Instead
A pathological clause is one that expresses an intention to arbitrate while making arbitration difficult or impossible to conduct. Most of them are recognisable on sight once you know the shapes. The table below sets out the ones that recur in Indian contracts, what a better formulation looks like, and, crucially, what the defect actually costs when it is exploited.
| Pathological drafting | Better drafting | Why it matters |
|---|---|---|
| Disputes may be referred to arbitration. Permissive language, sometimes combined with a reference to the jurisdiction of a named court over the same disputes. | Any dispute arising out of or in connection with this Agreement shall be finally resolved by arbitration, and the parties agree to be bound by the award. | Permissive wording invites the argument that the parties conferred an option rather than an obligation, so the threshold question becomes whether there is a binding agreement to arbitrate at all. Under Section 7 that is the foundation of everything downstream. |
| Arbitration shall be held at New Delhi. A single place, described only as where arbitration is held, with nothing said about the seat. | The seat of arbitration shall be New Delhi. Hearings may be held at any venue the tribunal considers convenient, and this shall not alter the seat. | Supervisory jurisdiction for Sections 9, 11, 34 and 37 follows the seat. Leaving it to be inferred is how parties end up litigating in two courts about which court may hear them, which is the purest form of wasted cost. |
| Sole arbitrator appointed by the Managing Director of the Company. Or by a nominee of one named party. | A sole arbitrator appointed by a named arbitral institution under its rules, or a three member tribunal with one arbitrator nominated by each party and the presiding arbitrator appointed by the two nominees or by the institution. | This is the unilateral appointment defect. Section 12(5) with the Seventh Schedule, and the line of authority including TRF Ltd v Energo Engineering Projects Ltd (2017) and Perkins Eastman Architects DPC v HSCC (India) Ltd (2019), make the mechanism vulnerable, and the vulnerability surfaces exactly when you need the tribunal constituted. |
| Disputes shall be settled amicably, failing which by arbitration. No process, no trigger, no time limit on the amicable stage. | Either party may give written notice of dispute. If the dispute is not resolved within a stated number of days of that notice, either party may refer it to arbitration. Escalation is a condition of reference and expires automatically at the end of that period. | An unbounded precondition gives a reluctant respondent an argument that the reference is premature. Multi-tier clauses are fine; multi-tier clauses without a defined start, a defined end and a defined trigger are an invitation to delay. |
| Carve-outs by subject matter. Claims relating to intellectual property, or to payment, or to termination, excluded from arbitration and reserved to a court. | Either no carve-out, or a carve-out drawn by remedy rather than subject matter, such as preserving the right to seek urgent injunctive relief from a court without affecting the tribunal's jurisdiction over the merits. | Subject matter carve-outs mean somebody must first classify the dispute, and in a real dispute both parties classify it in whichever direction suits them. You have created a jurisdictional fight as a preliminary to the actual fight. |
| Rules of a body that does not administer arbitrations, a misnamed institution, or an appointing authority that no longer exists. | The full and current name of the institution and of its rules, checked at the time of drafting, with a fallback if the institution ceases to administer arbitrations. | An inoperable appointment mechanism sends you to court under Section 11 to have an arbitrator appointed. That is months you did not budget, and it happens before the case begins. |
| Silence on language, number of arbitrators and governing law of the arbitration agreement, in a contract with a foreign counterparty. | State the language. State one or three. State the governing law of the contract, and where there is any cross border element, address the law governing the arbitration agreement expressly. | Each silence is a preliminary application waiting to happen. In cross border matters the law governing the arbitration agreement can determine questions of validity and scope, and it is the element most often left unstated. |
| Identical but separately drafted clauses across a suite of related contracts, each naming a different seat, institution or number of arbitrators. | One clause, drafted once, replicated verbatim across the suite, with consolidation and joinder addressed where the contracts are genuinely interdependent. | Related contracts with divergent clauses produce parallel proceedings over one commercial dispute, with inconsistent outcomes and no efficient route to bring them together. This is the most expensive defect on this list. |
Tiers, Multiple Parties, and the Clock
Multi-tier preconditions
Escalation clauses requiring negotiation between senior executives, or mediation, before arbitration are common and often commercially sensible. Whether the precondition is mandatory, and what follows if a party proceeds without satisfying it, depends on how the clause is drafted and how a tribunal or court construes it on the facts. The drafting response is straightforward. Make the trigger objective, which usually means a written notice. Make the period finite. Say expressly what happens at the end of it, which should be that either party may refer the dispute to arbitration. And say whether the period counts for limitation purposes, because a precondition that quietly consumes limitation is a trap for the party that complies with it.
Multiple parties and multiple contracts
Modern commercial arrangements rarely sit in one document. There is a framework agreement, several work orders, a guarantee, a side letter and an escrow arrangement, and they were drafted at different times by different people. If each carries its own arbitration clause with its own seat and its own institution, a single commercial failure produces several arbitrations that cannot be consolidated and may reach inconsistent conclusions. Where a transaction is genuinely interdependent, use one clause across the suite, and address joinder and consolidation expressly by reference to institutional rules that provide for them. Where a guarantor or an affiliate is expected to be bound, make that explicit rather than assuming a tribunal will find it.
Timelines and cost
Section 29A of the Act imposes a time limit for making the award, with a mechanism for extension. The operative period, how it is computed, the categories of arbitration it applies to and the route to extension should be confirmed against the provision as it currently stands rather than assumed from memory, because this is an area that has been amended. What a drafter should take from it is that the statute contemplates a disciplined timetable, and that clauses which build in long preconditions, cumbersome appointment mechanisms and multiple tiers are consuming a clock that is not unlimited. Cost allocation is worth a sentence too: silence leaves it entirely to the tribunal, which is a legitimate choice but should be a choice.
Interim relief
Section 9 provides for interim measures by a court, and it is what a party turns to when the tribunal does not yet exist and assets, evidence or the subject matter of the dispute are at risk. Two drafting points follow. First, the seat determines which court that is, which is another reason to state the seat in terms. Second, if you have carved out a right to seek urgent relief from a court, say expressly that seeking it does not waive or affect the agreement to arbitrate, because a counterparty will otherwise argue exactly that.
Auditing the Portfolio Before a Dispute Audits It for You
Everything above is easy to apply to the next contract you draft. The harder and more valuable exercise is the one nobody is asked to do: going back through what you have already signed. A GC who discovers the unilateral appointment defect during a live dispute has discovered it at the worst possible price. A GC who discovers it in a planned review has discovered it at the best.
Assemble the population
You cannot audit clauses you cannot find. Pull together every live contract of material value, including framework agreements, work orders, guarantees and side letters, and note which of them relate to the same commercial arrangement. Expect the assembly step to be the longest part, and expect to find agreements nobody had listed.
Extract the arbitration clause as structured data
For each contract, record the same fields: seat as stated, venue as stated, number of arbitrators, appointment mechanism, institution and rules if any, language, governing law, escalation preconditions and any carve-out. The point of structuring it is that defects become visible as patterns rather than as individual readings, and patterns are what you fix at scale.
Flag the unilateral appointment mechanisms first
Sort for any clause in which one party, or a person connected to one party, appoints the sole arbitrator or nominates the tribunal. This is the highest yield search in the whole exercise and it is mechanical. Treat every hit as a matter to review with counsel rather than as a conclusion, because the consequence for any given clause depends on its terms and on the position as it currently stands.
Then flag seat ambiguity and clause divergence
Second pass: clauses that name a place without saying seat, and suites of related contracts whose clauses do not match. The first creates a supervisory jurisdiction fight, the second creates parallel proceedings. Both are cheap to fix in an amendment and expensive to fix in a dispute.
Rank by exposure, not by count
You will not remediate everything, and you do not need to. Rank the flagged contracts by value, by remaining term and by the realistic probability of dispute, and fix the top of that list. A defective clause in a low value contract that expires in four months is not where your attention belongs.
Fix at the natural moment
Amendment for value requires the counterparty to agree, and the cheapest moment to obtain agreement is when they want something: a renewal, an extension, a variation, a new work order. Hold the flagged list against the renewal calendar so that the fix rides on a conversation that was going to happen anyway.
Close the loop on the template
None of this is worth doing if the template that produced the defect is still in use. Fix the standard clause, retire the old one from every precedent bank and drafting note where it survives, and make the standard clause the default in the intake process rather than something a drafter has to remember to reach for.
Where CourtMesh Fits
A clause audit fails for a boring reason: nobody can find the contracts, and once found, nobody can hold the extracted data anywhere it will stay current. That is a contract lifecycle problem before it is an arbitration problem.
CourtMesh offers a contract lifecycle management app, My Agreements, which gives the portfolio a single place to live, with intake, approvals, and obligation and renewal tracking. For this exercise, three things follow from that. The contracts are in one repository rather than in a drive, a drawer and three inboxes, which is what makes the population assemblable at all. The clause fields you extract can be held against the contract as structured data rather than in a spreadsheet that will be stale within a quarter. And because renewals and notice windows are tracked, the flagged list can be held against the renewal calendar, so the amendment conversation happens at the moment the counterparty is already at the table.
On the research side, CourtMesh provides unified search across the Supreme Court, 25 High Courts, District Courts and Tribunals over a corpus of roughly 310 million cases drawn from official government portals, with AI case analysis. That is how you check what the courts are currently doing with clauses shaped like yours, and how you follow a developing line of authority rather than relying on a note written when a template was last revised.
The limits should be stated as plainly as the capability. CourtMesh does not draft your arbitration clause, does not tell you whether a particular appointment mechanism survives, and does not give legal advice. Whether a specific clause in a specific contract is enforceable is a question for counsel on the facts and on the position as it stands when you act. What the platform does is make the portfolio visible, keep the extracted terms current, and put the case law in one search instead of several.
Fix the clause before the dispute finds it
The arbitration clause is agreed in the last four minutes of a negotiation and then carries the entire procedural architecture of a dispute that may run for years. Name the seat in terms and separate it from the venue. Use an appointment mechanism that does not depend on one party's goodwill, because unilateral appointment is the most common defect in Indian legacy portfolios and it surfaces exactly when you can least afford it. Bound your escalation tiers, align the clauses across a contract suite, and preserve the route to interim relief. Then do the unglamorous part: pull the portfolio together, extract the clause terms as data, rank by exposure, and fix at renewal. CourtMesh My Agreements gives that exercise a home, and unified search across the Supreme Court, 25 High Courts, District Courts and Tribunals gives you the case law behind it. Neither replaces counsel, and neither pretends to.
Explore CourtMesh
