PUBLISHED DATE: 23 October 2026
As organisations increasingly integrate artificial intelligence (AI) into business operations, managing AI token consumption has become a critical component of cost control, performance optimisation, and operational resilience. AI tokens, which represent the units of text processed by large language models, directly influence service costs, system capacity, and user experience. This document examines the challenges associated with escalating token usage and outlines practical strategies for monitoring, governing, and optimising consumption across AI-enabled environments.
The paper explores key drivers of token expenditure, including prompt design, model selection, automation workflows, and user behaviour. It presents governance frameworks and technical approaches that enable organisations to balance innovation with sustainable resource utilisation. Particular attention is given to the relationship between token efficiency and organisational resilience, highlighting how effective token management can reduce costs, improve service availability, and mitigate operational risks associated with excessive or uncontrolled AI usage.
By providing actionable recommendations, performance metrics, and best-practice controls, this document supports decision makers, technology leaders, and operational teams in developing robust AI consumption strategies. The findings demonstrate that proactive token management is not only a financial consideration but also a strategic capability that strengthens an organisation’s ability to scale AI solutions securely, efficiently, and sustainably.
Managing AI token consumption for operational resilience [PDF, 1.4 MB]
This guidance was co-authored by:
- U.S. Cybersecurity and Infrastructure Security Agency
- Australian Cyber Security Centre
- Canadian Centre for Cyber Security
- U.K. National Cyber Security Centre