AI->Human Communication: How to?

TL;DR: AI and human consume information differently. AI works best when given dense and comprehensive information with detailed instructions, but human likes it concise. How can we make AI->Human communication more effective?

Detailed description: In the current surge of AI agent developments, it is clear that the human->AI communication is most effective when the human user can clearly articulate their request, but how about AI->Human, especially when there are too much information to show?

For example with research agents, they read large collections of papers, code, and data, and possibly do some experiments on their own. In existing research agent works, the tendency is to generate thousands of lines of code and possibly a 100-page report. No human being will possibly read all that in details, and to fully trust the AI-produced research work, we need to verify that nothing fishy happens in the middle.

Imagine a scenario that you are joining some ongoing research project. It is likely not publishable, but there are some progress. How do you quickly onboard with what has been done? Or, you went for a break and want to catch up with the literature you care about (which probably has 100 new papers), how do quickly catch up with all that information?

These type of problems happens a lot with the AI->Human communication. It may sound like a classic summarization problem, but it actually requires a lot of careful thinking about what are the most effective ways to help human understand a dense source of information