Reliable, Smart Workflows for Finance, Wealth and Insurance — Karthic Rao, Perpetual
Regulated finance AI workflows should minimize LLM usage, not maximize token consumption.
“in a regulated industry, token maximization is the wrong goal”
Perpetual CTO Karthic Rao presents an architecture philosophy for AI in regulated finance: build deterministic state-machine workflows and invoke LLMs only at genuine decision points, keeping intelligence minimal and purposeful. This directly counters the 'burn more tokens' guidance from large labs, arguing that reliability and compliance requirements make token maximization the wrong optimization target. The approach has practical implications for any AI deployment in regulated industries where a policy violation is not just a bug report.