Test the whole workflow.
Consider a recurring contract review. AI may produce a first assessment quickly, while the legal team still spends hours finding amendments, correcting the business context and checking unsupported conclusions. The relevant measure is the effort and quality of the complete review, including the decision it supports.
Begin with a real task and observe how it is done. Identify the costly preparation, repeated explanation and uncertain handoffs. An improvement may involve better information, a simpler process or a different allocation of responsibility as well as AI. Test it on representative work, including missing information and difficult exceptions.
Make handover a test.
A convincing demonstration is only one moment. Give the responsible colleague a new task and the materials they will actually retain. Can they find the relevant sources, operate the workflow, assess the answer and recognise when they need help?
If the colleague must reconstruct hidden instructions or repeatedly ask the implementer what an output means, there is more work to do. A useful handover includes source context, operating methods, checks, unresolved issues and clear ownership. It should help the next person continue without having to repeat the original project.
Choose continuing support deliberately.
Some teams will operate the result themselves. Others will want continuing technical support, legal judgment or a shared arrangement. These are legitimate choices. Make the responsibilities, cost of upkeep and route to change the arrangement intelligible.
This is one practical expression of structural change in legal work. You retain more knowledge and a greater ability to assess your options, including how you use outside counsel. Judge the result through useful completed work, the quality of decisions and the independence you can exercise.