This Week in AI for Ridiculously Busy People

What It Covers

AI's shift from subsidized token consumption to usage-based pricing is reshaping enterprise strategy, with companies like Uber and Walmart already capping employee AI usage while the market develops cost-cutting architectural solutions.

Key Questions Answered

• •Token Efficiency Architecture: Enterprises must now treat token management as a core business function. Model routing systems like Factory's native routing maintain state-of-the-art performance while cutting costs by 25%, making intelligent model selection a measurable competitive advantage worth implementing immediately.
• •Hybrid Model Stacking: Harvey's collaboration with Fireworks AI demonstrates that pairing an open-weight worker agent with a frontier model advisor outperforms the frontier model alone on legal tasks at a fraction of the cost—a replicable architecture pattern for any domain-specific enterprise deployment.
• •Post-Training for Cost Reduction: Microsoft and McKinsey post-trained a model on McKinsey-specific tasks, achieving GPT-4.5-level performance at one-tenth the cost. Domain-specific fine-tuning is now a viable cost strategy, not just a performance strategy, for organizations with well-defined task categories.
• •Codex Sites Feature: Codex's new "Sites" feature converts any in-platform document or project into a deployable website or web app in a single click, currently available to business and enterprise users—making shareable, functional web outputs a standard unit of knowledge work.