Dr Jana Julia Hübler sets out the core principles for the responsible and effective use of AI and considers what the growing role of AI means for the next generation of professionals.
Artificial intelligence is no longer just a future consideration for legal practice. It is already in use across the profession, including in restructuring, turnaround and insolvency work. Yet for many practitioners, the conversation remains focused on tools and technology rather than the more fundamental question of how professionals should engage with AI in a way that is genuinely useful, legally sound and professionally responsible.
This question carries particular weight in our profession which is defined by time pressure, incomplete information and high-stakes decisions and where errors can have serious consequences. In that environment, AI can either provide meaningful support or introduce additional risk, depending on how it is used.
Starting with the right question
The starting point for responsible AI use is a practical question: What exactly am I trying to achieve?
Restructuring and insolvency work typically begins under pressure. Information is incomplete, documents are disorganised and deadlines are short. Legal, commercial and tactical considerations must also be brought together quickly. In that environment, AI can provide genuine support by assisting with initial document reviews, extracting key facts, identifying gaps, preparing issue lists and supporting the drafting process. However, the starting point must always be a clear understanding of the task at hand.
This is important because different AI systems are built for different purposes. Some are better suited to summarising and drafting, while others perform better at document-based analysis, structured workflows or comparative tasks. Sometimes, generative AI tools like ChatGPT, Copilot Harvey and Legora are not the right solution. Traditional automation or a well-designed spreadsheet may be more appropriate. Only once the task has been clearly defined can a professional decide which tool is suitable. Responsible AI use therefore begins before a prompt is even written.
Quality instructions, quality output
The second principle is familiar but often underestimated: garbage in, garbage out.
AI systems respond strongly to the material and instructions they are given. Even if the task is unclear, the output may appear fluent and well-structured. This is where the risk lies. A polished answer is not necessarily a reliable one.
AI systems are highly responsive to the input and instructions they receive. The risk lies in the fact that if a task is described imprecisely, the system will fill in the gaps, but not necessarily in the way the user intended.
A good prompt is not a magic formula. It is more like a clear work instruction given to a capable but uninformed colleague. It should define the task, its context, the relevant documents, the expected output and the limits of the answer.
This can make a substantial difference. Asking an AI system to ‘summarise these documents’, for example, leaves almost everything open to interpretation, such as what information is important, what to look for, what to flag and what to ignore. The system will produce an answer, but it will be based on its own interpretation of the task. A more useful set of instructions, which mirrors the way a professional might approach a task, would be to prompt the system to identify key facts, payment arrears, liquidity issues, deadlines, termination rights, financing risks, governance concerns and missing information. This is likely to result in a more detailed output.
The same logic applies to how tasks are sequenced. Professionals do not move directly from a pile of documents to providing final advice. They first read the documents, analyse them, and build on that analysis before drafting a submission, preparing advice, or formulating a position. AI work should follow the same structure. Rather than one all-encompassing prompt, a better approach is a workflow in which each prompt builds on the output of the previous one. This workflow would start with an overview, followed by a focused analysis of specific issues and then a drafting step informed by both. Each stage creates the foundation for the next. This is how AI can support the kind of structured, step-by-step thinking that results in professional work of a high standard.
Professional judgement remains critical
The third principle is the ‘human-in-the-loop’ principle.
Human control begins before the process starts: Professionals using public versions of AI, rather than those where the information is contained within the company, should not enter confidential, personal, or professionally privileged information unless they are satisfied that doing so is permitted under applicable confidentiality obligations, data protection requirements, and professional secrecy rules.
Once the output has been produced, the facts, sources, reasoning, completeness and legal relevance must all be verified.
Before communicating information to a client, court or another party, ask yourself: Is this legally correct, commercially sensible and appropriate for the situation? This decision requires professional expertise and cannot be delegated to AI.
AI can help professionals work faster and more systematically, which is crucial in our profession given the tight deadlines involved. However, it does not reduce professional responsibility. Professional judgement is crucial throughout the process, not just for signing off at the end.
Impact on new professionals
AI will also change what is expected of new professionals entering the field. Future lawyers and restructuring, turnaround and insolvency professionals will be expected to have AI competencies, such as the ability to formulate precise instructions, structure complex tasks and work efficiently with AI tools in their day-to-day work.
However, this is only part of the story. It will be equally important for new professionals to know how to manually review documents, identify legally relevant facts, conduct analyses and recognise when something does not make sense. Only someone who can perform an analysis independently can critically assess the reliability of AI output. If this ability is lost, AI becomes a risk rather than a support tool. The greater danger is not that AI produces an incorrect answer. It is that the user fails to recognise the error.
AI competence and traditional professional competence are not mutually exclusive. New professionals will need both.
Conclusion
The most valuable applications of AI in restructuring, turnaround and insolvency tend to be the most practical, such as creating an initial overview of documents, extracting dates and amounts, identifying gaps, comparing versions and organising large volumes of information quickly. These tasks provide real value in the early stages of a case. However, AI should not be treated as the final authority.
The key is to use AI purposefully by starting with the task, not the tool and giving clear instructions. Professional judgement is central to critically evaluating the output. This is how AI can become genuinely useful to the profession.
This article is based on a presentation given by Jana Julia Hübler, a lawyer at CMS Germany, at the R3 & INSOL Europe International Restructuring Conference in London.
Coming soon: Discover how AI is transforming restructuring, turnaround and insolvency work in R3 and Alph4’s forthcoming report, featuring practical case studies, expert insights and guidance for professionals.
