How AI Is Changing Enterprise

Aaron Levie (Box CEO) argues enterprise AI adoption is ~1% into agents, token costs will converge to zero, and the software TAM could expand 5x as AI does work software previously couldn’t.

• Enterprise banks are roughly 10% into general AI chat adoption and ~1% into anything resembling agents, per Levie’s New York meetings.
• Box gross margin is 81%; a YC portfolio company scaled from 30% to 80% margins in one year as token costs dropped — exact parallel to cloud storage economics.
• Meta’s open-source commitment guarantees token costs converge to zero; any frontier model must match pricing of the second-cheapest competitor or lose customers.
• Goldman Sachs is running internal projects where AI drafts an S1 in minutes versus a six-person team — Levie cites this as proof top-tier regulated banks are leaning in early.
• Most enterprise AI spend will flow to ISVs, not internal builds; the core-vs-context framework means only proprietary workflows (drug discovery, recommendation algorithms) warrant homegrown AI.
• Non-AI-first companies face a talent retention crisis: AI-native graduates entering the workforce will refuse to work with legacy tooling, handing competitive advantage to AI-first peers.
• AI TAM expansion mirrors SaaS over on-prem: Salesforce grew addressable customers from ~10,000 to millions; AI will monetize workflows enterprises never automated before, not just convert payroll to software revenue.

2025-02-19 · Watch on YouTube