AI & Enterprise

AI use rises as costs feel burdensome, token engineering is coming

As companies use more AI, the costs can feel burdensome, lifting interest in ways to use AI more effectively. Coinbase disclosed how it cut AI costs by nearly half while increasing token usage, focusing on better default model settings, routing and caching rather than lowering usage caps. CEO Brian Armstrong described analyzing prompts to allocate requests to suitable models and expanding cache reuse. Microsoft’s Nicolas Bustamante highlighted background agents and a possible shift to delay-tolerant token pricing.

AI & Enterprise

Cost-effectiveness becomes a competitive edge as AI cost-cutting techniques evolve

Companies are stepping up efforts to use AI more cost-effectively as usage grows and expenses weigh. Some are cutting AI costs by mixing cheaper models with high-performance systems from OpenAI and Anthropic, or by using routing software that selects models automatically. Others set limits on employee usage or restrict access to expensive models. Firms are also refining prompts to reduce “thinking” time. Open-source models, including Chinese ones, are gaining share.