Yijia Liang explains why long-running financial analysis involves shared contextContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task., human reviewHuman-in-the-loop AI keeps a person involved in reviewing, correcting, approving, or guiding an AI system's work. and sensitive data that differ from typical personal assistant tasks. He describes an approach in which deal teams collaborate with the same agent through a shared inbox.
Yijia Liang describes building custom spreadsheet, presentation and document engines after evaluating desktop plugins, virtual machines and existing libraries. He presents the company’s reported performance and fidelity results as product claims.
Yijia Liang discusses parallel agent searches across large collections, source citations and iterative write-render-check loopsAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior.. The talk closes with his view that finance teams can delegate more workAI agent delegation assigns a bounded task, context, authority, and expected result from one participant to an agent or another agent. while retaining oversight and judgment.
Watch on YouTube
