GSTAT’s first decade: five institutional choices that will decide how it works
Benjamin Cardozo observed that judges do not merely find law; they make it by filling the gaps a statute leaves open.
The Goods and Services Tax Appellate Tribunal (GSTAT) has now entered its operational phase with its benches sitting nationwide. It applies a young law full of known and unknown ambiguities and will spend its first decade filling the gaps.
What it decides will shape GST jurisprudence for a generation, and whether it earns legitimacy depends less on its rulings than on its institutional choices. Five of such choices we pitch here.
Equip the bench before the hearing begins
Reasoning about economic consequences cannot rest on the adversarial record alone. The GSTAT needs a dedicated research cell—modelled on the Supreme Court’s research assistant scheme—that maps precedent and assembles legislative history before a hearing. Where a question is novel or affects the whole GST ecosystem, especially in cases with many micro, small, and medium enterprises (MSMEs) and the like, an amicus curiae can supply the argument the parties lack the incentive to make. This is not administrative luxury; it is a jurisprudential investment.
Put the right case before the right bench
Which bench hears a case should be a deliberate design, with consistency and efficiency as the design choices. When the same question is being fought before many benches at once, the President can route them all to one “issue bench” that settles it in a single ruling, not a dozen conflicting ones; a weekly “sector day” does the same for a whole sector’s disputes.
The law already lets a small appeal that turns only on facts, not law, be heard by a single member, not a full two-member bench. Every appeal today, however modest, first goes to a two-member bench merely to decide whether it is simple enough for one—spending the time of two judges on a sorting question. Sorting cases is an administrative task, not a second round of litigation.
Let the order and its consequences travel together
A taxpayer who pays the 10% pre-deposit, wins before the GSTAT, then watches that money stay frozen while the department appeals has not received the justice promised. Interest on a pre-deposit refund is a statutory entitlement, not administrative grace; what is missing is the machinery, which the GSTAT can supply by a simple direction—as a 2014 circular once did for central excise.
Accountability is the larger point. The Tribunals Reforms Act, 2026, has created a National Tribunals Data Grid and mandated annual performance review—yet the GSTAT sits outside its First Schedule. Rather than wait to be brought in, the tribunal should adopt that discipline itself: a live dashboard of pendency by age, orders against reservation dates, and disposals set beside reversal rates, so speed is never bought at the cost of reasoning.
Close the loop with the GST Council
The GSTAT now sees exactly where the law fails—the tariff entries that breed endless classification fights, the input-credit denials litigated in bulk though the tax was paid. A short annual “red flag note” to the Council, surfacing the provisions generating the most avoidable litigation, would turn that experience into correction at source. The Council already responds to litigation; what is missing is a formal channel. It needs no amendment—only the will to build a self-correcting tax culture in place of an adversarial one.
Employ AI, do not shun it
While the Supreme Court has flagged an “alarming” trend of petitions citing AI-hallucinated judgments that do not exist, the answer is not prohibition; it is deployment matched to capability. AI that retrieves and monitors can be a diligent clerk that reads everything and forgets nothing—and there the GSTAT should put it to work. A model watching every order across 31 benches can mark the moment an issue is decided one way on one bench and differently on another, so that the bench, in its wisdom, may refer the conflict to a larger bench and settle it for good.
It can also flag where the department, having accepted a GSTAT ruling in one jurisdiction, contests the identical point in another—so that an accepted position is communicated as settled across all benches, not re-litigated on a taxpayer-by-taxpayer basis. So deployed, AI keeps the bench consistent.
These five share a theme: each keeps the GSTAT from mistaking formal vindication for real relief, and each measures the tribunal against the economy it serves rather than the cases it clears. The statute is imperfect—bringing the GSTAT within the appointment and oversight framework Parliament has just legislated for its peers may help—but most of what the moment demands is already within reach.
Cardozo’s counsel was a disposition, not a formula: approach the gaps sincerely, and build law the economy can rely upon. The GSTAT’s first chapter is being written now. India cannot afford to get it wrong.
Manjunath A.N. and Arjun Raghavendra M. are former officers of the Indian Revenue Service.