Arvind Jain argued that enterprise AI value depends less on using the largest models and more on connecting models to company context, tacit knowledge, and measurable work. He said “token maxing” has become a cost problem, with enterprises needing budgets, guardrails, and smarter routing between frontier, open-source, and smaller models.