LexisNexis: Is your legal work “too complex” for AI?

When you take the time to consider it, the number of steps legal teams have to work through before a matter properly gets moving is mind-boggling. There is the research, the scoping, the client conversations, the drafting, the fact-checking and the proof-reading, often before the substantive legal work has even reached its most difficult point. It is little wonder so many lawyers are turning to AI.

Our survey of private practice lawyers in large law firms found that four-fifths use AI tools for legal research. Roughly two-thirds use the technology for knowledge management, large-scale document review, document analysis and client document drafting.

This is not confined to junior lawyers working through high-volume tasks. Usage cuts across practice areas and seniority levels, and in many cases senior lawyers are more engaged than their junior colleagues.

That does not mean every legal workflow is ready for AI, or that complexity is a weak excuse for resisting change. Sometimes the work really is too complex for a particular AI use case, especially where the task depends on incomplete facts, shifting client objectives, sensitive negotiations, commercial judgement or an understanding of risk that cannot be captured properly in a prompt.

Oliver Bethell, Chief Technology Officer at Travers Smith, puts the point plainly.

“A lot of what we do as a firm is bespoke and complex, and doesn’t lend itself to automating end-to-end.”

That is an important distinction. The sensible question is not whether AI can be pushed into every piece of legal work, but where it genuinely reduces effort without creating more work in review, correction or risk management. If a tool produces a draft that takes longer to check than it would have taken to write, the use case is weak. If it helps a lawyer find the right source, compare documents faster, test a line of reasoning or spot an inconsistency, it starts to earn its place.

Human judgement should still sit with the lawyer

Much of the debate around AI in legal work becomes unhelpful because it treats the work as one large category. In practice, legal matters are made up of many different activities, some of which are more suitable for AI than others.

AI may not be the right tool for forming a final view on a novel point of law, managing a delicate negotiation or balancing competing commercial risks. Those are areas where context, judgement and responsibility sit firmly with the lawyer.

But around those moments are many tasks that absorb time and attention: reviewing bundles, summarising background material, comparing clauses, checking documents against known positions, searching internal knowledge and producing an initial draft that a lawyer can then reshape. These tasks are not low-value simply because they can be supported by technology. They are often the work that allows good legal judgement to be applied properly.

The issue for firms is deciding which parts of a workflow can be improved safely and which should remain more tightly controlled. That requires a more precise discussion than whether legal work is, in broad terms, too complex for AI.

Adoption is ahead of integration

The survey data suggests lawyers are already finding uses for AI, whether or not firms have fully embedded it into their operating models. The more difficult task is turning individual usage into a consistent, governed way of working.

Only 30% of legal professionals at large firms said AI is embedded in their team’s strategy and operations. Adoption on its own does not create firm-wide value. It can just as easily create variation, with different teams using different tools, applying different standards and making different assumptions about what is appropriate.

Alex Bazin, COO & CTO at Lewis Silkin, says the benefits of AI will remain limited unless firms change the way work is organised.

“Very few firms have rewired their underlying processes, feedback loops, and expectations to drive consistency of usage. Until that happens, the gains from AI will stay trapped with individuals rather than compounding across the firm.”

Mark Rendall, CIO at Taylor Wessing, makes a similar point.

“Many firms are seeing experimentation and pockets of use, but without clearly embedded use cases, these don’t translate into sustained, organisation-wide impact.”

This is where complexity becomes a management issue as much as a technical one. If AI use is left to individual preference, firms have no reliable way of distinguishing between appropriate and inappropriate use. One lawyer may use AI to summarise background material, another may use it to generate a first draft, while another may use it for something much closer to legal analysis. Without agreed use cases, standards and review expectations, the boundary between useful support and unacceptable risk becomes too loose.

AI has to fit the workflow

For complex legal work, AI cannot operate as a generic tool that sits outside the workflow and is used whenever an individual lawyer happens to remember it exists. It needs to be connected to trusted content, secure systems, approved use cases and clear review standards. Otherwise, the technology may be available, but the working method remains largely unchanged.

Hélder Santos, Head of LegalTech and Innovation at Bird & Bird, says the biggest barrier is trust.

“AI needs to be carefully connected to a firm’s systems, processes, and ways of working, in a way people actually trust and use.”

That trust is shaped by more than the model itself. It depends on the sources the tool draws from, the data it can access, the way outputs are presented, and the confidence lawyers have in checking and using them.

Rachael Birch, IT Services Manager at Redkite Solicitors, points to a related problem around ownership.

“It is often not clear whether AI sits with IT, risk, or the legal teams themselves. As a result, many AI tools are simply bolted onto existing workflows rather than embedded as a seamless part of legal or administrative processes.”

That “bolted on” model may be tolerable for limited experimentation, but it is not enough for more complex work. If the tool is not part of the workflow, lawyers are left to decide for themselves how to use it, when to trust it and how much review is enough.

Build your workflows on trusted sources. Lexis+® with Protégé™

Data quality limits what AI can do

Poor data quality is another practical reason why some AI use cases struggle, particularly in larger firms where knowledge and matter information may sit across different systems.

Jason King, Head of Platform at Taylor Rose, sees this as one of the biggest barriers to scaling AI.

“Legal data is often fragmented across systems, inconsistently labelled, and embedded in siloed workflows, which limits what AI can reliably do at scale.”

That is especially relevant to complex work because context matters. If the information available to the AI is fragmented, incomplete or inconsistently labelled, the output will reflect those weaknesses. The problem may not be that the legal question is too complex in principle, but that the firm’s data environment is not yet strong enough to support the use case.

Taylor Rose is addressing this by moving to operate on the same platform across the group, supported by centralised data mastering and storage, zero-copy access and AI embedded directly into workflows.

King says the biggest gap between adoption and integration comes when firms deploy AI as standalone tools rather than building it into everyday legal workflows. That point is worth taking seriously, because the more complex the work, the more important the surrounding system becomes.

Trust remains the threshold

Lawyers will not use AI seriously unless they trust it, and their caution is justified. Our survey found that 85% of legal professionals at large law firms are concerned about relying on inaccurate or fabricated information.

Bethell says trustworthy AI has two dimensions.

“Trusting that the data is secure, and trusting that the model will produce reliable answers.”

Both are essential. Lawyers need to know client data is protected, but they also need to know where an answer came from, whether it is grounded in appropriate sources and how it can be checked. In legal work, the usefulness of an answer depends heavily on whether a lawyer can verify and stand behind it.

This is why legal-specific AI tools matter. Across all respondents, 72% said they feel more confident using AI that is grounded in legal sources, rising to 79% among lawyers at large firms.

Stuart Greenhill, Senior Director of Segments at LexisNexis UK, makes the point clearly.

“In legal work, confidence is not enough. Authority matters. Validation matters. Security matters. If you cannot stand behind the output, it is not legal AI. It is just AI.”

The opportunity for firms is to make better decisions about where AI can safely support legal work and where it should not be used. That means looking beyond the tool itself and focusing on the quality of the data, the authority of the sources, the clarity of the workflow, the review process and the professional judgement required.

AI will not remove the complexity of legal work, and nor should it. But where the use case is clear, the information is reliable and the governance is strong, it can help lawyers manage parts of that complexity with greater confidence