Every vendor demos well. The question is which tool survives contact with your matters, your documents and the way your people actually work.
Start a conversation with the AI Adoption Concierge, already scoped to tool selection & evaluation. Choose the question closest to yours, or describe your situation directly.
Firms rarely fail at AI adoption because they picked the wrong product. They fail because they picked without deciding what problem they were solving, ran a pilot that could not produce a negative result, and chose something that does not sit inside the systems people already have open. The market is also moving faster than any procurement cycle, which argues for evaluating capability categories and switching costs rather than committing to a single vendor as though it were a decade-long platform decision. This part of the practice is about asking better questions before signing anything, and designing an evaluation that could actually tell you no.
Three questions in sequence: what kind of tool, does it work on our material, and will it fit where people work.
The questions that separate tools, and the ones vendors would rather you did not ask.
investigateDesigning a trial that can fail — and measuring something other than enthusiasm.
investigateThe factor that most reliably decides whether a tool gets used — and the one weighed least in selection.
investigateHow the Institute approaches selection — criteria and process, never a recommendation of a named product.
They solve different problems and most firms end up with both. Legal-specific platforms that retrieve from an actual corpus of authority are materially better positioned for research, because there is a real document behind each citation rather than a generated one. General-purpose assistants are often better for drafting, summarising and reasoning over material you supply, and they improve faster. The decision that matters more than the category is which is approved for what task, and whether client-confidential material may go into either — that turns on the terms, not the label.
Enough that inaction is a decision, and less than the marketing implies. Survey figures vary widely with how the question is asked: some report AI use among legal professionals jumping from roughly 19% in 2023 to 79% in 2024, while others put lawyers actively using AI tools at around 30% in 2026, up from about 11% in 2023. The consistent findings are the gaps rather than the headline: larger firms adopt substantially more than solo practitioners, and only a small minority of firms — on the order of 8% — have adopted anything universally. Most adoption is partial and uneven.
Long enough to cover a real cycle of the work and short enough that the market has not moved underneath you — commonly six to twelve weeks. What matters more than duration is that it runs on genuine matters with a defined success measure agreed in advance. Pilots that run on curated sample documents, with enthusiastic volunteers and no pass mark, produce a positive result every time regardless of the tool, which is why so many firms have bought something that then went unused.
No, and it will not. Rankings in this market age within months, they cannot be validated against any firm's specific matters, and an institution connected to anyone who builds in this space cannot credibly publish them. What the Institute does is set out the categories, the evaluation criteria, the questions worth asking a vendor, and the failure patterns — so a firm can reach its own answer and defend it internally.
Describe the task you are trying to improve. The Institute will help you design an evaluation that could actually tell you no.