The largest gains in the profession sit next to the only place where a mistake becomes a published opinion with your name on it.
Start a conversation with the AI Adoption Concierge, already scoped to litigation workflow. Pick a starting point, or describe your situation directly.
Litigation is where AI has both the most to offer and the least room for error, and the two facts are not in tension so much as they demand different treatment at different stages. Discovery, transcript work and record review are volume tasks that compress enormously. Preparing to examine a witness, framing a theory of the case and deciding what to argue do not compress at all. Between them sits everything filed with a court, where a growing number of judges have entered standing orders about AI use and where the reported sanctions decisions come from. The practical approach is to stage the work: use the tools hard on the volume, keep them well away from the judgement, and apply the strictest verification to anything that will be filed.
The gain falls as the judgement content rises. The court-facing risk does the opposite.
The strongest case. Volume, defined questions, checkable output.
Depositions, medical records, long correspondence. Finding what is in there, fast.
Assembling a timeline from a large record. Strong, with sources verified.
Useful for orientation. Every authority verified to source before it is filed. No exceptions.
Can suggest lines of questioning. Cannot judge which one this witness will break on.
Judgement built on experience of this court and this opponent. Does not transfer.
How litigators keep this safe.
This is the area where the consequence of a lapse is public.
A growing number require disclosure of AI use, certification that citations were verified, or both. They are not uniform across a courthouse, let alone a jurisdiction. Check the individual judge before every filing.
It depends on the judge, which is why checking is a per-filing task rather than a policy you set once. A number of judges have entered standing orders requiring disclosure of AI use in filings, certification that any AI-assisted research was verified against original sources, or both — and these vary between judges in the same courthouse. The practical answer is that someone checks the individual judge's orders before every filing, treated as part of the filing checklist rather than something remembered.
It is concrete and the pattern is consistent: the reported cases overwhelmingly involve citations to authority that does not exist, filed without verification. Consequences have included monetary sanctions, referrals to disciplinary bodies, striking of filings and findings recited in published opinions that follow the lawyer permanently. Notably, the failure is not usually the AI use — courts have been considerably more forgiving of firms that used a tool and verified than of ones that used it, did not check, and then were slow to admit what happened.
For the preparation, usefully; for the examination itself, no. Analysing a long record to find inconsistencies, building a chronology and surfacing where a witness has said different things at different times is exactly the kind of volume work that compresses well, and it can be genuinely better than a human pass over thousands of pages. What it cannot do is judge which inconsistency is worth pursuing with this witness in front of this factfinder, or read the room and change course. That is the skill, and it is unchanged.
It is already appearing in proportionality disputes and will appear more. If a review that previously justified a burden argument can now be done at a fraction of the cost, the burden argument is weaker, and opposing counsel will make that point. The corollary is that a party resisting discovery on cost grounds should expect to be asked what technology it considered. This is a live area, courts are working through it unevenly, and it is worth watching in your jurisdiction rather than assuming settled practice.
Describe the matter. The Institute will help you separate the volume from the judgement.