LPM: Data will decide if AI will work in your firm
Nigel Williams, product director and interim head at LexisNexis Enterprise Solutions, explores how structured data and operational discipline optimise AI usage.
Data will decide if AI will work in your firmNigel Williams, product director and interim head at LexisNexis Enterprise Solutions, explores how structured data and operational discipline optimise AI usage
AI is unfolding in a similar pattern across law firms. First comes the excitement. AI solution demonstrations feel like magic. Instant summaries. Drafted clauses in seconds. Insights surfaced from thousands of documents in moments. Boards lean forward. Innovation committees accelerate. Budgets shift.
Then comes the quieter, profound realisation. AI is powerful, but entirely dependent on the quality of the data it is fed. AI does not solve data problems or correct structural disorder. It amplifies it.
If client entities are duplicated, AI will draw on duplicated records. If matter naming conventions are inconsistent, AI will misinterpret them. If metadata is incomplete, AI will generate incomplete insight.
The danger of this all? AI output will still look confident.
The myth of the quick fix
AI is often positioned as a shortcut — a way to leapfrog process discipline, and as a tool that can sit above existing systems and make sense of whatever lies beneath. However, the governance and risk teams will instinctively understand that no technology can reliably interpret chaos.
Furthermore, feeding the AI with poor-quality data risks horrific results — there have already been numerous public examples of lawyers relying on AI-generated case law that did not exist.
The firms that treat AI as an overlay rather than a foundation risk building something impressive looking but structurally unstable.
Structured data means clarity
In the legal context, structured data means that clients are defined consistently across the firm, matter records contain agreed and standardised fields, risk indicators are captured in the same way every time, documents are linked to matters in a coherent structure, and so on.
In essence, structured data provides clarity. This is important when using AI. For example, if a firm wants AI to identify trends across matters, but the matters are recorded inconsistently, the analysis will be flawed. If it wants AI to support risk review, but risk data is captured sporadically, the conclusions will be unreliable.
This is unlike in human-led manual processes, where lawyers correct or compensate instinctively for inaccuracies and inconsistencies in data. They’ll remember which spreadsheet holds which information. They’ll know where to look for that one critical document. They’ll know which colleague is likely to have the information, and such.
