In disputes before courts and arbitral tribunals, it often comes down to which party can understand the relevant facts quicker and more comprehensively. Finding the proverbial needle in the haystack that a large dataset can represent may prove decisive. Looming deadlines, growing document volumes and multilingual datasets make this task even harder. Manual review of the entire dataset quickly exhausts the time and resources available. It has therefore become common practice to filter the data for relevant search terms first, and to limit the manual review to documents found to contain these terms. This approach, however, entails accepting the known risk that the search-term method can never be entirely accurate. Dispute resolution practice therefore requires technical tools that address this tension between precise document review and efficient use of time and other resources.
AI-enabled document analysis
This is the focus of HM Argus, Hengeler Mueller's proprietary AI-enabled platform for risk-based analyses of large datasets. Originally developed for internal investigations, HM Argus has equally proved its worth in our Disputes practice. It combines semantic document classification, which analyses full linguistic contexts rather than isolated search terms, with relevance scoring and context-based natural language search. This allows HM Argus to identify the documents that are relevant to the dispute in a focused and resource-friendly manner and to analyse them by priority for the case at hand. Unlike standard tools available on the market, HM Argus, as a proprietary tool, is designed to meet matter-specific legal requirements with flexibility, to integrate into the law firm's existing knowledge base, and to offer enhanced cost control.
HM Argus bases its analysis on a structured set of factual questions a lawyer would pose, tailored to the individual case, which are then translated into suitable prompts by our data scientists. It evaluates each document, groups hits into clusters by topic, and assigns a relevance score to them. This limits manual review to relevant documents and permits prioritising the groups of questions by relevance. Even large datasets can be analysed within a short timeframe. On the key KPIs – identifying as many relevant documents as possible from the data pool (recall) and the actual relevance of the documents identified (precision) – HM Argus achieves results that significantly exceed the corresponding KPIs of traditional keyword-based reviews.
By way of illustration: In contentious post-M&A proceedings, HM Argus helped us directly identify around 20,000 relevant documents from a dataset that contained about one million documents. On the basis of the scoring process, we were able to select about one third of the hits for prioritised review. Our review team classified approximately one in five documents reviewed as highly relevant. Overall, this meant that the preselection was automated, our lawyers could focus on the relevant parts, and the documents were swiftly organised by relevance and content assessed.
This advantage extends to every stage of arbitration proceedings, from the initial assessment of the facts and documents produced through to the preparation of the oral hearing. The same holds true for the highly complex cross-border disputes on which we advise.
Early case assessment: setting the course
When work begins on a contentious matter, risk analysis and the choice of strategy take centre stage – typically before the facts of the case have been fully established. Fact-finding and setting the course early therefore present a dilemma: Informed decision-making requires a reliable factual basis; waiting for the completion of fact-finding means losing valuable time.
HM Argus resolves this tension by analysing even large datasets within a very short timeframe and prioritising the most relevant documents to be reviewed first. After the review performed by our lawyers, the documents assessed as most relevant swiftly yield a reliable picture of the sequence of events, thus providing a reliable basis for the early case assessment.
The strategy to be pursued can therefore be chosen based on facts, not on assumptions. For our client, this means a more informed assessment of opportunities and risks, a more reliable evaluation of the settlement options available and realistic budgeting – at a time when there is still leeway in making these decisions.
Statement of facts, witnesses, preparation of the oral hearing
Once a strategy has been chosen, the focus shifts from obtaining an initial overview to working on the relevant facts. Structured prompting allows us to organise the datasets by subject matter and, based on the scoring process, to prioritise and compile the documents relevant to a certain allegation for legal review.
This analysis according to substance assists us in preparing and substantiating our own presentation of the facts, preparing hearings of witnesses and, if required, also in preparing witness statements. Moreover, documents that corroborate, qualify or challenge our presentation of the facts can be identified at a very early stage – before they are submitted by the opposing party or during the further course of the proceedings. Where the opposing party introduces previously unknown facts at short notice before or during oral hearings, such assertions can be swiftly verified and, where necessary, rebutted on the basis of the data available.
Document production in arbitration proceedings
Document production, i.e. mutual presentation of documents upon request, often plays a significant role in arbitration proceedings. Each party may submit a request for production requiring the other party to produce certain documents or categories of documents. Tight deadlines meet large datasets – this applies in two directions:
For the client to comply with production requests, its document portfolio must be reviewed against the other party's request. Documents subject to a production obligation must be identified and reviewed within tight timeframes. At the same time, privileged communication must be separated and marked, as inadvertent disclosure may jeopardise the protection afforded by legal privilege. This first step is greatly accelerated by the identification of relevant documents and scoring provided by HM Argus, freeing up time for the second step, which requires a legal assessment.
On the other hand, the opposing party's production must be analysed. The speed and quality of this analysis are crucial for the potential use of relevant information in an additional presentation of facts, the oral hearing and any cross-examinations. In this regard, especially the semantic analysis of large unsorted datasets pays off: substantive connections between claims and potentially relevant documents can be identified even if their terminology, language or layout is different from the assumptions underlying the requests submitted by us.
Large case files and mass proceedings
In proceedings before state courts, HM Argus also assists by performing a technology-based review of extensive files in complex commercial disputes. In mass proceedings, and wherever claims are standardised, case patterns, clusters, anomalies and comparability of individual cases can be analysed and systematically organised. This facilitates the handling of each individual case while permitting a consistent line of defence to be maintained across a large number of parallel proceedings in a resource-efficient manner.
Outlook
The use of AI to analyse and process the facts of a case in litigation and arbitration matters is gaining traction. Besides the economic benefits, its use directly impacts the conduct of the proceedings: From the outset, the choice of strategy is based on facts rather than assumptions, the presentation of facts can be underpinned by reliable evidence at an early stage, any document production (to be made or received) can be dealt with on time, and even extensive files remain manageable.
This is the core competency of HM Argus: clients benefit from a tried and tested tool that offers meaningful support, even under tight deadlines and challenging economic conditions, in securing the decisive advantage in fact analysis and the overall conduct of the proceedings.
