Arora predicts a 90% drop in token costs, which he sees as bullish for AI adoption, as it will drive increased consumption. He criticizes enterprises for only marginally adapting current workflows instead of redesigning their companies around AI, predicting that roles in marketing, HR, and finance could be cut in half within three years. He argues that AI applications will have opinions, unlike traditional SaaS, and that value will accrue not just at the model layer but also in memory and context, creating a significant moat. Arora warns that CEOs must lead AI transformation themselves rather than delegating it, and that companies should buy off-the-shelf AI applications rather than building from scratch. He also discusses the need for better guardrails and the inevitability of Chinese open-source models in the ecosystem.