Vinoo Ganesh explains that AI can generate financial work quickly, but the speaker argues that citations and model evaluationsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. alone do not establish whether a figure is correct or whether a firm's rules were followed. The central challenge is making the process behind a result checkableAgent completion verification checks evidence that an agent achieved the requested outcome rather than merely producing intermediate work..
The proposed architecture has three parts. Atomic provenanceData provenance records where information came from, how it changed, and which people, systems, or processes handled it. lets a model identify a source figure while deterministic software retrieves and validates itDeterministic software verification checks an output with a repeatable procedure that produces the same result for the same inputs and state.. Scope determinism assigns planning to the model and extractionInformation extraction identifies selected facts and relationships in source material and converts them into a structured form. and arithmetic to code. Derivation chains preserve the inputs and transformations behind calculated figures such as margins and valuations.
In the Q&A, the speaker identifies filings, deterministic calculations and internal documents as inputs to the provenance chain. The speaker says customers want an AI analyst for repetitive reading and first-pass modeling tasks, while portfolio managers retain responsibility for investment decisions.
Watch on YouTube




