I save articles, project notes, and scraps of thinking because they might help with a question later. The trouble is remembering which one applies when the question finally arrives. Giving a collection to AI can make retrieval easier. I get more from it when I start with a situation I can describe.
Suppose someone is looking for a portfolio topic after working at an exhibition. I would ask what they were responsible for, what visitors repeatedly asked, and whether they ever changed the way something was explained. With that account in hand, the knowledge base can bring up relevant cases and research methods. Then we can see which line of inquiry the person can actually pursue. A good-looking idea can wait if they cannot reach the people or observe the situation it depends on.
I use a similar test when I write project notes. Would someone joining today understand the problem and the decisions already made? A developer needs implementation detail. A collaborator may need to understand why the work is worth doing. A new team member may need the history first. AI can help arrange the same material for each reader. I still need to give it the context and check what it inferred. Keeping track of the original question, the suggestions I received, and the reasons I changed direction also gives me something concrete to return to when I write up the project.