“Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance.”
That line comes from U.S. District Judge P. Kevin Castel’s decision in Mata v. Avianca, Inc., and the context is what makes it worth sitting with. Attorneys had already submitted a federal court filing containing judicial decisions that did not exist, and when those citations were questioned, ChatGPT continued to present the authorities as genuine. The firm later explained that the attorneys had not believed the technology could be “making up cases out of whole cloth,” which gets to the uncomfortable part of the story because the output did not look ridiculous enough to distrust at first glance.
What makes Judge Castel’s observation more useful than the usual warning about AI is that he did not treat the technology itself as the problem. The issue was what happened after the output appeared on screen and before it reached the court. Somewhere in that gap, verification failed.
That is the more practical question behind AI for law firms now. These tools can research, review, organize, summarize and draft, but the real decision is which work they can be trusted to handle, what still needs to be checked, and where attorney judgment needs to stay close.
Best use cases of AI for law firms
The best use cases of AI for law firms are usually straightforward. Find the information faster, get through volume, or prepare something useful for review.
Find
Some legal work starts with search.AI can shorten the time between knowing something exists and actually finding it.
Read and organize
Large matters create more material than anyone can reasonably hold in their head at once.AI can sort the first layer and make the review set smaller.
Prepare a first version
Some work already has the facts, source material and instructions.AI can bring those pieces together before the attorney starts refining them.
What can go wrong with legal AI?
Most legal AI problems are fairly predictable once you know where to look.
- Incomplete records — missing amendments, exhibits or correspondence can distort the result.
- Version control — an old playbook, earlier draft or outdated template can send the analysis in the wrong direction.
- Hallucinations — citations, quotations and facts can look credible and still be false.
- Escalation — exceptions need a clear point at which the work goes back to counsel.
- Confidentiality — client information should only enter approved tools under clear data-handling rules.
- Review fatigue — consistently good outputs can make reviewers less careful over time.
- Accountability — someone still needs to own the work that ultimately leaves the firm.
One practical test
If AI saves 30 minutes at the start and adds 40 minutes of checking and correction at the end, it has moved the work rather than reduced it.
Which AI tools for lawyers are built specifically for legal work?
General-purpose models are no longer the only option. There is now AI software for law firms built around the work lawyers already do, from research and drafting to contracts and case preparation.
The category is getting broader too. Some law firm AI products try to cover research, drafting and review in one workspace. Others stay close to one type of work, such as contracts or personal injury.
For firms comparing AI for attorneys, that difference is more useful than a long feature list. The best AI for law firms is the one built for the work the firm actually wants to move through it, with sources, permissions and review still accounted for.
Will AI replace lawyers and legal judgment?
AI can take on more of the first pass, but there is still a point where the work stops being about processing information and becomes a legal decision.
Questions a lawyer still has to answer
- Can I rely on this authority?
- Does this change the legal or commercial position?
- What does the evidence actually mean?
- What should happen next?
- Am I prepared to put my name behind this work?
That last question matters. In Fletcher v. Experian Information Solutions, AI was used in drafting a filing that contained quotations, citations and assertions the underlying cases did not support. The Fifth Circuit still looked to counsel, because the responsibility for what reached the court remained theirs.
What happens to lawyer training?
There is another issue underneath all of this. Researching a point, drafting it, getting corrected and going back through the authorities was also how junior lawyers learned to exercise judgment.
The American Bar Association has raised this concern, particularly around the loss of the “productive struggle” that comes with doing early legal work yourself. As AI for attorneys takes over more of that first attempt, firms will need to think about what replaces the experience that used to come with it.
How should law firms evaluate AI software?
Choosing AI software for law firms gets easier when the starting point is the work, not the feature list.
Questions to ask before adoption
- What work are we trying to move?
Research, contract review, intake and litigation preparation need different things from a tool. - What will it need access to?
Client files, firm knowledge, external research sources or several systems at once. - Can we trace the answer back to the source?
Especially when the output contains authorities, quotations, dates or facts someone will rely on. - What happens to client information?
Check where the data goes, who can access it, how long it is retained and whether it is used for model training. - Where does attorney review happen?
Decide that before the tool enters live work. - Does it fit the systems already in use?
A useful tool can still add work if lawyers have to build a separate process around it. - What does it save after review?
Count the checking, correction and cleanup too.
Responsible AI adoption checklist
Before putting a tool into live legal work:
- Define the task and expected output.
- Approve the sources and data the tool can access.
- Set rules for confidential and privileged information.
- Require source-level verification where accuracy matters.
- Define what must go back to an attorney.
- Assign responsibility for final review.
- Test the workflow on controlled matters first.
- Measure time saved after review and correction.
The best AI for law firms depends on the job, the material it works from and how easily a lawyer can verify what comes back. If the tool is already chosen, LegalEase can help shape the legal AI workflow, review steps and quality checks around it.
.svg)