Every legal search has a stopping rule, though it is rarely written down. The consultation ends. A hearing approaches. The advice is due while other files are still open. At some point the lawyer stops looking, not because the law has yielded every relevant answer, but because the work must move on.
This is not a defect of character. It is the condition under which most professional judgment is exercised. The question is what that limit does to the argument. An authority that was never found cannot be weighed. A contrary line that would take another afternoon to trace may never enter the room. A familiar case becomes the centre of the analysis partly because it is familiar.
The most important promise of AI research is therefore not that it can produce more words per hour. It is that it may change what a lawyer has time to consider. A question asked in ordinary language can surface decisions whose vocabulary differs from the words counsel first tried. A broader first set of candidates can reveal a split, a qualification, or an analogy worth pursuing. The saved time matters when it is spent examining those discoveries, not when it is converted into a longer answer no one has checked.
There is a catch. Faster retrieval can create a larger burden of verification. Twenty plausible judgments are not automatically better than five carefully read ones. A tool that merely expands the pile has shifted work rather than improved it. The gain lies in helping a practitioner locate the promising cases sooner, understand why they may matter, and move into the underlying decisions while there is still time to change course.
That is the problem we built Zalari to address. Today we are opening an early preview, shaped around the Zimbabwean, South African, and Labour Court collections in our Law Reports archive. It is a chance to test what a question asked of that material might bring into view, not a claim that the work of finding or judging is finished. Drafting, in particular, is experimental: useful to explore, but not yet dependable for production legal work. Fuller capabilities will follow, and we will keep sharing what improves as the preview develops. The intellectual work of deciding what matters remains the lawyer’s.
The difference may be especially consequential where a practitioner cannot put a large team on a question. It will not erase disparities of experience, resources, or access to a complete record. It may, however, make a wider first encounter with the reported law possible within the hours actually available. That is a modest claim with a serious implication: better legal work should not depend entirely on who can afford the longest search.
Think of a question you once answered under a hard deadline. Ask it again in Zalari, with the jurisdiction and forum made explicit. Do not count the cases it returns. Ask whether, in the time you would really have had, it brings a useful authority or a troubling counterexample into view early enough to matter. Then read that decision. The point of the exercise is not to admire the speed of the machine; it is to see what your saved attention can do.
Speed is valuable in law when it buys the time to think more carefully. If it merely makes us more confident sooner, it has cost us something instead.
