◇ Could this help me?
Yes for the first half, with real reservations about the second. Making institutional knowledge answerable rather than merely written down is a straightforwardly good idea, because the failure of internal documentation is almost never that it does not exist, it is that nobody can find the one paragraph they need, and that is exactly the gap an assistant with access to the material closes. The part not to generalize is dropping design docs, design reviews, and roadmaps: that works at a company with unusually senior staff, direct customer access, and a product whose direction is set by research, and in most organizations those artifacts are what keep people aligned rather than bureaucracy to be removed. General practice for the knowledge-access idea, and the speaker's own admission that he has never been more wrong is the right lens for the rest. Judge it as one company's experiment rather than as a model to copy.
I want to make the knowledge in my own work answerable instead of just written down somewhere. Help me identify the questions that get asked repeatedly and currently require finding a document or asking a person, then tell me what material would need to be reachable for those questions to be answered reliably, and what would go wrong if the answer were confidently wrong. Be specific about which questions are safe to answer this way and which ones carry enough consequence that they should still route to a person. Do not propose removing any existing documentation until we have tested whether the answers hold up.