I have noticed that AI work often changes the brief after the project has already started. A team begins with one problem, then the tool reveals a different constraint or a different opportunity.

That can be useful, but it can also make the design process slippery. If the brief keeps changing silently, the team loses track of what it is trying to prove.

I try to pause and rewrite the brief when AI changes the shape of the work. What is the user problem now? What should stay human-controlled? What new risk did we introduce?

The brief does not need to be rigid. It just needs to stay visible as the team learns.