AI creates value. Human judgement builds trust.

Trust has always been fundamental to the relationship between lawyers and their clients. My legal training taught me that earning that trust requires more than knowing the law. It requires questioning assumptions, examining the relevant facts and sources and being able to explain and defend a legal conclusion.

As artificial intelligence (AI) becomes an increasingly important part of legal practice, I find these principles more relevant than ever. AI offers considerable opportunities to improve the way we conduct research, analyse information and approach complex legal questions. For me, the challenge lies in making the most of these opportunities while ensuring that efficiency and technological progress continue to serve the quality and reliability of the legal advice we provide.

AI adoption is already widespread. According to the Forvis Mazars C-suite Barometer, 95% of leaders confirmed at the start of the year that AI is having an impact on their organisations. But the next challenge is not simply adopting more technology. It is turning AI into measurable and sustainable value, supported by strong governance, quality data and clear accountability.

For me, the same principle applies to legal practice. The value of AI does not come simply from using it more or working faster, but from applying it where it genuinely improves the way we work while maintaining the quality and reliability of the advice we provide.

In my Corporate/M&A practice, I use AI to navigate specialised legal sources more efficiently, structure information and explore different approaches to legal questions. These tools are developing rapidly and already contribute meaningfully to our daily work. The question is therefore not whether to use AI, but how to use it responsibly: where does it add value, where are its limitations and what is needed to turn information into reliable legal advice?

Why I still go back to the source

Generative AI can produce persuasive legal analysis within seconds. It may also overlook factual distinctions, rely on incomplete information or refer to an authority that does not support the conclusion drawn from it. The quality of the language does not necessarily reflect the reliability of the underlying analysis.

Legal research is particularly demanding because the appropriate legal assessment often depends on circumstances that are not immediately apparent. The law applicable today may not be the law that governed a particular action when it occurred. Transitional provisions, contractual arrangements or specific factual circumstances can fundamentally change the legal assessment.

This is why the ability to critically examine a legal conclusion remains so important. Whenever an AI tool refers to a judgment, statutory provision or particular interpretation, I verify the reference against the underlying source. The relevant question is not only what the authority says, but whether it actually supports the position being considered.

My international work and studies in International Business Law have reinforced the importance of this approach. Legal concepts that appear similar may have different consequences across jurisdictions. An analysis that is correct under one legal framework cannot automatically be transferred to another.

AI can make research more efficient and unfamiliar legal materials more accessible. It can also challenge initial assumptions by suggesting arguments or perspectives worth examining. Determining which of those arguments are legally sound and relevant to a particular matter, however, remains an essential part of a lawyer’s work.

In M&A, context changes everything

This becomes particularly evident in legal due diligence. AI can assist in identifying change-of-control clauses, comparing contractual provisions, detecting inconsistencies and organising findings across substantial volumes of documentation. These capabilities can be particularly valuable when working under demanding transaction timelines.

But no two transactions are identical.

A consent requirement that appears significant when reviewing a contract in isolation may ultimately be manageable in the context of a particular acquisition. Conversely, a seemingly routine provision may become critical once the target’s business, the transaction structure or the client’s commercial priorities are considered.

This is where the distinction between identifying information and providing legal advice becomes particularly important.

When an AI tool flags a potential issue during a due diligence review, I verify it against the underlying documentation and assess its legal relevance, materiality and implications for the transaction and the client.

In that sense, effective AI use in legal work requires clarity about where technology can support the process and where human judgement remains essential. The Forvis Mazars report, AI at scale: from experimentation to real business value, highlights the importance of human-in-the-loop validation mechanisms and recommends evaluating technology choices based on the specific business need, use case, data environment and risk level.

For legal due diligence, this distinction is particularly relevant. AI may help identify, organise and compare information, but determining its legal significance and implications for a particular transaction remains a matter of professional judgement.

Equally important are risks that may not be apparent from the available documentation. A data room may be incomplete. A contractual risk may only become apparent when considered alongside a corporate or regulatory issue. Sometimes, the relevant question arises from understanding how the target actually operates rather than from reviewing a particular document.

An efficient review of the information available does not necessarily establish that all relevant information has been considered.

Good due diligence is therefore not about producing the longest list of findings. It is about identifying and understanding which risks matter, explaining them clearly and assessing their significance against the client’s specific objectives.

Not every identified risk needs to be eliminated. A client may decide to accept certain risks, seek contractual protection or require an issue to be addressed before closing. What matters is that the decision is made with a clear understanding of the legal and commercial implications.

The value of due diligence is not the number of issues identified. It lies in the quality of the judgement applied to those that matter, including the ability to ask what may be missing.

How will junior lawyers develop their judgement?

AI also raises an important question about the next generation of lawyers.

Much of the legal knowledge and judgement we develop comes from the work itself. My own training involved extensive research, working through legal literature and judgments, learning where to look and recognising when an apparently straightforward answer required closer examination.

AI is already changing that process. Repetitive work should not be preserved simply because previous generations of lawyers spent hours performing it. Greater efficiency is a genuine opportunity. At the same time, if AI increasingly takes over some of the tasks through which junior lawyers have traditionally developed their research and analytical skills, we need to think more deliberately about how those skills will be built in the future.

There is no simple answer. But perhaps the question is no longer whether junior lawyers should learn the traditional way or learn to work with AI. They will need to learn both.

This reflects a wider challenge for organisations as AI becomes more embedded in everyday work. The AI at scale report emphasises that people remain at the centre of AI transformation, with successful adoption depending on leadership, workforce enablement and clear communication about how AI supports growth and performance.

For the legal profession, this raises an important question: as technology changes how some traditional tasks are performed, how do we ensure that the analytical skills and professional judgement developed through those tasks continue to be built?

Learning to use AI should not come at the expense of learning how to question it. Understanding where to find reliable sources, how to evaluate competing interpretations, when further investigation is necessary and, importantly, when an answer does not quite make sense remains fundamental to becoming a good lawyer.

Trust is ultimately about responsibility

Attitudes towards AI in legal practice range from considerable resistance to genuine enthusiasm. Technological developments that genuinely improve our work should be embraced. At the same time, responsible use requires more than confidence in the capabilities of a particular system.

This comes down to practical questions: Is the tool appropriate for the information involved? Have the relevant sources been verified? Are the limitations of the analysis understood? And has consideration been given to what might be missing?

Our professional obligations remain equally relevant. Confidentiality, the protection of client information, professional diligence and appropriate transparency must be considered regardless of the technology used.

If anything, AI reinforces the need to understand how information is handled. Forvis Mazars' latest cyber security research highlights the convergence of AI governance, data privacy and cyber risk, recommending that organisations manage AI governance and data privacy through shared policies, controls and vendor oversight strategies rather than treating them as entirely separate disciplines.

The way we use AI also matters to our clients. They should be able to rely on advice that reflects their objectives, the relevant legal and commercial risks and an independent professional assessment. Depending on the circumstances and the significance of AI’s contribution, this may also require careful consideration of whether and how its use should be communicated to clients.

AI has considerable potential and is likely to become increasingly integrated into legal practice. It can make information more accessible, improve efficiency and allow us to devote more attention to the questions that matter most to our clients.

But professional judgement involves more than reaching a technically correct conclusion. It requires understanding the client’s circumstances, assessing the significance of legal risks, recognising uncertainty and taking responsibility for the advice provided.

AI moves fast. Trust takes longer. And in legal practice, trust is ultimately built not by the technology we use, but by our ability to question its output, understand its limitations and stand behind the advice we give.