As a digital investigation method, the eDiscovery process is an essential part of the processing and investigation of data content in the context of, for example, litigation, arbitration proceedings, special investigations, internal investigations and the handling of data leaks. In this context, companies and law firms are confronted with an overwhelming amount of data. In order to process and analyse this data in detail, a precise approach is essential.
Artificial intelligence (AI) can significantly reduce project costs
Such projects were previously characterised by the development of search filters, search term lists and often a large-scale manual document review. The technology has developed significantly and AI approaches such as predictive coding or technology-assisted review have been able to reduce project costs considerably in some cases under certain conditions. Nevertheless, eDiscovery projects can become expensive as previous AI approaches have reached their limits. Generative AI is revolutionising the way companies approach their eDiscovery processes and requires a rethink of how data is identified and viewed. At the same time, important requirements must be placed on traceability and quality control. And finally, those involved in the project must also be trained to "interrogate" the software and thus the AI in a meaningful way in the project context - sound prompt engineering can be a decisive success factor.
Michael Sauermann
Partner, Audit, Regulatory Advisory, Forensic
KPMG AG Wirtschaftsprüfungsgesellschaft
Documents are not considered individually, but as a whole
Generative AI makes it possible to analyse large amounts of data quickly and in a targeted manner with regard to the subject of the investigation. It can extract relevant information, classify documents, identify links and create summaries. This technology offers considerable advantages as it increases the efficiency and accuracy of document searches and analyses. By using AI-supported algorithms, large volumes of data can be searched and analysed in less time. In addition, documents are not only categorised intelligently, but also prioritised, allowing project teams to focus on the crucial aspects of a case or project. Generative AI can also uncover patterns, connections and nuances, which can lead to robust legal decisions. The nature of the review thus changes from an individual consideration, e.g. of a single document, to an embedding in the overall context.
The improved efficiency can ensure that often lengthy and cost-intensive eDiscovery processes are avoided. With the support of generative AI, companies can therefore optimise the data analysis process, which leads to a significant reduction in processing and review times. This is particularly important in legal contexts, where deadlines must be met and any delay can have serious consequences. The points mentioned above significantly reduce the time required compared to a purely manual review. Companies and law firms can concentrate on the essentials: strategically analysing the data found to process the case and develop well-founded decisions.
Innovations and possible applications of generative AI
Generative AI is able to recognise patterns and correlations within the data that may not be immediately obvious to humans, which can lead to valuable insights. Search queries are formulated in natural language to better understand and consider the searcher's goals. Current software products can also provide an assessment of why the document found is considered relevant in the context of the project. Potential risks are also addressed. Instead of limiting itself to specific keywords, AI therefore enables a context-based search that also takes into account terms used synonymously and related concepts. This makes it easier to find relevant documents that might otherwise be overlooked.
By "teaching" the AI a specific project context and objectives, the AI can automatically classify documents according to relevance and sensitivity. Accordingly, documents can be automatically categorised as "relevant", "not relevant" or "privileged". This significantly reduces the manual effort involved in reviewing documents, speeds up the entire eDiscovery process and can also improve the quality of the results. Companies and law firms no longer have to sift through extraordinary amounts of documents, but can concentrate on the most critical information.
The performance of the generative AI improves with every interaction. It learns from the user's input and continuously adapts its models, which leads to a steady improvement in search results.
Many modern solutions offer user-friendly dashboards and interfaces that enable users to work efficiently with the technology even without in-depth technical knowledge. This promotes the acceptance and effective use of the technology in various departments of a company.
Generative KI kann umfassende Dokumentensätze analysieren und prägnante Zusammenfassungen der wichtigsten Informationen erstellen. Dabei werden die Ergebnisse bereits in Texten zusammengefasst - ob als Fließtext oder in Stichpunkten, je nach Auswahl der Nutzerinnen und Nutzer. So ist es beispielsweise möglich, die Kommunikation zwischen zwei Personen (sog. Custodians) zusammenzufassen und in einer Zeitleiste darstellen zu lassen. Die Einsichtnahme in die Dateninhalte wird damit sehr niederschwellig. Das hilft Anwaltskanzleien schnell einen Überblick über einen Fall zu gewinnen und die wichtigsten Argumente oder Beweise zu identifizieren, was den Entscheidungsprozess effizienter gestaltet.
Advantages for companies and law firms
- Accuracy and faster insights: Through context-based searches, analysts have insights and simultaneous access to the context of documents and their content. AI models can also reduce human error in document classification and evaluation.
- Cost savings: A large part of the cost of eDiscovery projects is usually incurred during the document review process. By reducing the number of man-hours required, companies save time and money. Cost efficiency is currently a critical success factor - in light of this, the quality-assured implementation and application of AI solutions offers a clear competitive advantage.
- Faster results: The automation of routine tasks speeds up the entire process. This is particularly important in time-critical situations where quick reactions and answers to crucial questions are required. For example, companies and law firms can classify and analyse newly acquired evidence (e.g. data from another custodian) more quickly, which can influence the success of their cases.
- Improved compliance: With more in-depth, targeted, comprehensive and quality-assured data analyses, companies can ensure that they comply with legal requirements and internal guidelines. This minimises the risk of legal consequences and protects the company from potential consequences, including penalties.
The capabilities of the relevant software products in eDiscovery projects are remarkable. New insights can be gained in a very short time. At the same time, however, there are also risks that should be taken into account.
Possible risks:
- Incorrect classification: Generative AI can classify documents incorrectly or overlook important information. These errors can lead to legal disadvantages if not recognised and addressed by the project team.
- Bias and discrimination: AI models may have unintended biases ("algorithmic bias") that originate from the training data and may lead to unfair results. This can lead to certain documents or information being systematically favoured or disadvantaged, which can jeopardise the integrity of the presentation of results and the eDiscovery process.
- Regulatory and legal risks: Handling sensitive data requires special care. The use of AI is subject to regulatory or legal requirements. Insufficient compliance and disregard for regulatory provisions and laws can lead to legal consequences. For this reason, the current legal situation should always be kept in mind. eDiscovery projects are also often characterised by an international focus. Accordingly, the legal situation in the respective countries should also be taken into account.
- Complexity of integration: The implementation of generative AI in existing eDiscovery workflows can be complex and often requires adjustments to the infrastructure and, in particular, the data processing and quality assurance processes. In addition, project participants need to be trained.
These tasks and challenges require careful planning and implementation to maximise the benefits of generative AI in the eDiscovery process while minimising potential risks.
Feel free to contact me at any time.