We used our own product to cut our AI tooling bill without cutting AI usage | Harness Blog

Harness describes using its Cost Management Agent to control rapidly growing AI-tooling expenses without reducing AI adoption by consolidating cost and usage data across providers, models, users, and token types. The company first gave developers visibility into their individual spending and emphasized cost per million tokens as an efficiency metric, then introduced centrally managed weekly budgets with alerts at 90% of limits and full access blocks after budgets were exceeded. To avoid disrupting productivity, developers could request budget increases through an in-product workflow with tiered approvals and configurable guardrails. After governance launched in July 2026, average weekly spend per developer reportedly declined from $184 to $115, while cost per million tokens fell from $1.80 to $0.80, which Harness attributes to better model selection, improved cache use, and more appropriate context sizing rather than less AI usage. The company concludes that visibility should precede restrictions, weekly feedback cycles are more effective than monthly ones, individual attribution encourages action, and flexible exception processes make strict spending controls easier to adopt.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.