Ricardo Barreiro-Deymonnaz
Barreiro Abogados, Buenos Aires

Introduction

Construction arbitration is experiencing a crisis of evidentiary inflation. On complex infrastructure and engineering projects, the documentary record has grown from filing cabinets to terabytes. Claims and counterclaims routinely number in the hundreds. And yet, the fundamental architecture of dispute resolution – a panel of arbitrators, a hearing measured in days or weeks, advocates presenting and testing evidence in oral submissions – has not changed in any fundamental respect.

The conventional response to this mismatch is technological: use AI-assisted review to identify relevant documents, deploy sophisticated document management platforms, leverage search algorithms to navigate the record. These tools are useful. But they do not address the underlying problem. The underlying problem is that size has become a proxy for seriousness in construction arbitration, and no institution with sufficient authority is challenging that assumption.

This article argues that the record management crisis is, at its root, an institutional failure. And the institution best positioned to address it is the tribunal itself.

The anatomy of record inflation

Construction arbitration has a record problem. Not a document management problem. Not a technology problem. A record problem – and the difference matters. International construction arbitration is inherently characterised by the massive volume of documents involved, such as daily progress reports, drawings, specifications and the vast correspondence generated over the lifespan of a project.

This problem is not new. According to the Queen Mary University of London’s 2019 International Arbitration Survey: Driving Efficiency in International Construction Disputes,^[1] 66 per cent of surveyed professionals identified large amounts of evidence as one of the most defining features of these disputes. The document overabundance was highlighted by 42 per cent of respondents as a primary cause of procedural inefficiency, causing delays and heavily burdening an already complex evidentiary process.

To mitigate this issue, the survey’s conclusions pointed towards two main solutions: a strategic shift in case management and the decisive adoption of new technologies. Procedurally, the report recommended that legal teams focus on presenting core documents rather than producing massive volumes of information, with 51 per cent of respondents favouring limiting document production to highly pertinent documents. Notably, 33 per cent of users were willing to completely forego the document production or disclosure phase to save time and resources. On the technological front, the report concluded that it was highly beneficial to use common platforms to centralise electronic disclosure. Furthermore, 40 per cent of respondents believed that adopting technological automation for document production and review could significantly increase the efficiency of these document-heavy arbitrations without compromising the quality of the dispute resolution process.

It is clear, therefore, that as long ago as 2019, Queen Mary University of London’s survey had already identified record volume and its management as a significant driver of procedural inefficiency in construction arbitration. The prescribed remedy, as the survey’s respondents envisaged it, lay primarily in technological solutions.

We are now in the artificial intelligence era, with powerful tools at our disposal which not only help with document production and review automation as envisaged by Queen Mary University of London’s survey but also go much further into advanced and complex analysis of thousands of documents in short periods of time. And yet, the record problem seems to have worsened rather than improved.

The growth of the project record is driven by three reinforcing dynamics. First, modern project management software – Aconex, Procore, Primavera, BIM platforms – generates records automatically and continuously. Every RFI, every schedule update, every document transmittal is logged, timestamped and stored. The system does not distinguish between a document that matters and one that does not.

Second, legal culture rewards volume. In a system where omission is a professional liability and inclusion is relatively costless, rational advocates include everything. The submission of a comprehensive bundle signals preparation. The omission of a potentially relevant document creates adverse inference risk, or so the argument runs.

Third, and most importantly, tribunals seem to have historically been passive recipients of whatever parties choose to submit. This passivity is not inevitable. It is a choice – or, more precisely, a habit.

Record abundance does not equal probative clarity

Probative value – the capacity of a document to prove or disprove a fact in issue – does not scale with volume. A single contemporaneous site diary entry written on the day of a critical event may be more probative than 10,000 automatically generated system logs. A concise programme narrative prepared at the time of a delay event may do more to illuminate causation than a forensic delay analysis reconstructed years later from scheduling metadata.

The traditional legal principle of best evidence has been quietly displaced,
in construction arbitration practice, by a principle of all available evidence. These are very different standards. The former requires parties to identify and present their most probative evidence. The latter simply requires them to produce everything they have and invite the tribunal to sort it out.

The consequences of this displacement are significant. Tribunals are asked to absorb volumes of material that cannot meaningfully be reviewed in the time available. Advocates are incentivised to over-produce rather than curate. And the probative signal is buried in documentary noise.

The technology paradox

AI-assisted document review is frequently presented as the solution to the record management problem. AI can review hundreds of thousands of documents in hours, identify thematic clusters, flag potential inconsistencies and generate privilege logs with minimal human input. These are genuine capabilities and they represent real value.

But AI review, at its current state of development, answers the question of which documents touch on a given topic. It does not answer the prior question: which of those documents is genuinely probative of a specific fact in issue? The advocate who uses AI to identify all documents potentially relevant to a claim and then submits them all has not curated the record. That advocate has industrialised the production of an uncurated record.

More troubling still is the role of AI as a record generator. Modern project management platforms produce continuous streams of automatically-generated records – clash detection reports, sensor data logs, automated workflow notifications, version histories. These records exist. They are technically part of the project record. But they were not generated by a human being making a decision or observing an event. Their probative value, in most cases, is minimal. They are the digital exhaust of software systems doing what software systems do.

We risk building a dispute resolution system in which the project record is increasingly composed of documents that no person wrote, describing events that no person witnessed, submitted in volumes that no tribunal can meaningfully review.

The institutional responsibility of tribunals

Tribunals have the tools to address this problem. They have always had the tools. The IBA Rules on the Taking of Evidence in International Arbitration permit tribunals to limit document production to what is genuinely relevant and material. The Prague Rules provide a model for active tribunal management of the evidentiary process. Institutional rules under the International Chamber of Commerce (ICC), the London Court of International Arbitration (LCIA), the Singapore International Arbitration Centre (SIAC) and comparable institutions vest broad procedural authority in tribunals. None of this is new.

What is new is the scale and urgency of the problem. In a large dispute with a terabyte record, the passive receipt of whatever parties choose to submit is not a neutral procedural choice. It is an abdication of the tribunal’s responsibility to ensure that the dispute resolution process functions effectively.

The counterargument deserves to be taken seriously. Parties to arbitration are autonomous actors who bear responsibility for the manner in which they present their case. If a party chooses to submit thousands of documents without properly guiding the tribunal as to where it should look and why, that is a strategic choice – and its consequences, including the risk that the tribunal fails to find the probative needle in the documentary haystack, fall on the party that made it. Tribunal intervention, on this account, risks distorting the adversarial process and creating due process exposure that prudent arbitrators rightly seek to avoid.

There is something to this view. Party autonomy is not merely a procedural convenience in international arbitration; it is a constitutive value of the system. And the due process concern is real: an arbitrator who imposes documentary limits risks a losing party arguing, on the way to annulment proceedings, that its right to present its case was impaired.

But the argument proves too much. A tribunal’s responsibility is not simply to provide a neutral arena in which parties deploy whatever evidentiary strategy they choose. It is to administer justice – and that function carries with it a duty to ensure that the process is capable of reaching the facts.
A party that buries a probative document in a ten-thousand-page bundle has not assisted the tribunal in finding the truth; it has transferred to the tribunal a curation burden that was always the party’s own to discharge. Permitting this transfer is not neutrality. It is a form of institutional abdication dressed as procedural deference.

The contrast with domestic court litigation is instructive, though it should not be overstated. Construction disputes that reach state courts – including technically complex, document-intensive ones – do not routinely generate evidentiary records of the scale common in international arbitration. The difference is not the nature of the disputes. It is the culture of the forum. Domestic judges exercise their case management authority with a directness and confidence that is sometimes absent in arbitral tribunals and they do so without routinely generating the due process challenges that arbitrators fear. The lesson is not that arbitration should replicate domestic litigation. It is that the exercise of legitimate procedural authority is compatible with due process; indeed, it is one of its requirements. A process that cannot manage its own evidentiary record is not protecting the parties’ right to be heard. It is simply failing to function.

The practical implications are not radical. Tribunals should establish documentary management protocols at the first procedural conference. Parties should be required, when producing documents, to articulate not merely their relevance but their probative value – the specific fact in issue they are said to prove or disprove. Hearing bundles should be subject to document or page limits, with exceptions available on reasoned application. Parties should be required, in advance of the hearing, to identify the specific documents they intend to rely upon in oral submissions.

None of this prevents parties from maintaining comprehensive document repositories for their own reference. None of it prevents the tribunal from requesting additional documents if it considers them necessary. What it does is shift the burden: instead of tribunals sifting through everything parties choose to submit, parties must do the intellectual work of identifying what is genuinely probative – what is the best evidence to support their case as opposed to all the evidence available for that purpose.

A note on civil law and common law

The civil law/common law divide is frequently invoked in discussions of evidence in international arbitration. Civil law tribunals are said to be more comfortable with an inquisitorial model; common law advocates are said to be more accustomed to comprehensive document disclosure. There is some descriptive truth to this.

But the framing obscures more than it reveals. The habits of over-production characteristic of common law practice and the habits of under-production sometimes seen in civil law practice both fail in the context of large disputes. The real divide is cultural rather than jurisdictional, and international arbitration, at its best, draws on both traditions. A tribunal with the confidence to exercise its procedural authority – to demand less when parties offer too much and to request more when parties have withheld genuinely probative material – serves the process better than a tribunal that passively receives whatever arrives.

Turning principle into practice: mechanisms and enforcement

Identifying the tribunal’s institutional responsibility is a necessary first step. But it immediately raises two harder questions: precisely how should a tribunal exercise its authority, and what happens when parties do not comply?

On the first question, the answer lies in graduated procedural architecture. At the first procedural conference, the tribunal should issue a documentary management order that does three things: establishes the categories of documents parties may produce without leave; sets a default limit on hearing bundle size; and – most importantly – requires each document submitted to be accompanied by a brief statement identifying the specific fact in issue it is said to prove or disprove. This last requirement is the most transformative. It forces advocates to perform the intellectual work of curation before submission. A party that cannot articulate what a document proves has no business submitting it.

Beyond this foundational order, tribunals have a range of further tools: hearing bundle protocols issued well in advance of the hearing, with genuine page or document limits and a procedure requiring leave to exceed them; bifurcation or phased hearings that confine the documentary record for each module to what is genuinely relevant to the issues in that phase; and strict annexure limits on witness statements and expert reports, which are the primary vehicles through which documentary inflation enters the record.

On the second question – non-compliance – the position is more demanding. Tribunals are understandably reluctant to exclude documents, fearing that a losing party will raise due process objections on the way to annulment proceedings. But this concern, while real, is frequently overstated. Due process requires that parties have a fair opportunity to present their case. It does not guarantee the right to present it in whatever manner the party chooses, regardless of the tribunal’s procedural orders. A party that submits 3,000 documents in defiance of a 500-document limit has had its opportunity. It chose to exercise that opportunity in breach of the tribunal’s order. The tribunal may properly decline to admit the excess.

Costs sanctions and, in appropriate cases, adverse inferences are also available. Where a party’s conduct appears designed to obscure rather than illuminate – to bury an inconvenient document in a 1,000-page bundle rather than present probative evidence transparently – that conduct is itself a form of evidentiary manipulation that the tribunal may address.

The deepest practical challenge is the epistemic one. Presented with a 1,000-page bundle, how can a tribunal determine whether document 347 is genuinely probative without reading all 1,000 pages? The answer, in most cases, is that it cannot – and this is precisely the structural function of the oversized bundle. It transfers the curation burden from the submitting party to the tribunal. The solution is not for tribunals to read faster. It is for tribunals to require, in advance of the hearing, structured reliance schedules identifying each document, the witness or submission in which it will be used and the fact it is said to prove. Parties that cannot complete this exercise have not curated their record. Tribunals that receive it have a navigational framework that makes the bundle manageable. The presumption shifts: a document not identified in a reliance schedule carries no weight by default and the party wishing to rely on it must seek leave.

Conclusion

The record management crisis in construction arbitration will not be solved by better technology or by more sophisticated document management platforms. It will be solved when tribunals are willing to exercise the authority they already possess: to impose meaningful documentary discipline from the outset of proceedings; to require parties to identify their best evidence rather than submit all their evidence; and to insist that the project record, as presented in arbitration, reflects genuine probative value rather than mere volume.

More is not necessarily better. That principle is not controversial in the abstract. The challenge is to translate it into procedural practice. Tribunals are the institution best placed to do so. The question is whether they are willing.

Ricardo Barreiro-Deymonnaz is a partner at Barreiro Abogados in Buenos Aires. He can be contacted at rbarreiro@bodlegal.com.