Snowflake’s Plan to Cut AI Costs
THE SO WHAT
Snowflake’s focus on model routing and multi-model risk management is a clear signal that enterprise AI is entering its cost-optimization phase, not its “just ship something” phase. If your stack still assumes one default frontier model, you’re overpaying and under-hedged—start instrumenting usage and routing across smaller, cheaper models where quality allows.
READ THE SOURCE
MORE FROM THE WIRE
Applied AICommerce AI is fragmenting. Here is why that matters.
Commerce AI is splitting into point solutions, orchestration layers, and bespoke in-house stacks—outcomes are diverging because integration, not model choice, is the constraint. If you own a P&L, you should be mapping where AI touches pricing, merchandising, and service this quarter and consolidating around a small number of orchestrators instead of adding yet another point tool.
Applied AIAlina Kukarina on why AI transformation begins before choosing the technology
The fact that nearly 40% of companies tracking AI cost savings see less than expected underscores that the blocker is operating model, not tooling. Before you pilot another copilot, clarify which decisions you want to change, who owns them, and how you’ll measure impact—or you’ll just add another dashboard to ignore.
Applied AIMy iPhone keyboard was becoming infuriating for typing AI prompts— I changed these 3 settings, and ChatGPT is now much easier to use
Prompt UX is now constrained by mobile input friction as much as by model quality. If your product assumes heavy natural language input on phones, you need to design around autocorrect, text expansion, and voice—not just ship a chat box.
Applied AIWe're asking the wrong question about the cost of enterprise AI
Focusing on token pricing while ignoring the infra behind each request is how teams end up with surprise GPU and networking bills. If you own an AI budget, push vendors for end-to-end cost per workflow—not per 1,000 tokens—and model how that scales under real usage.