AI savings depend on understanding the full cost

For the last two years, AI discussions have focused heavily on jobs, automation and future staffing. TechRadar reports that enterprise leaders are increasingly concerned about another issue: losing visibility into AI spending.

Organizations are adopting tools such as Microsoft Copilot and Google Gemini across business applications, while also building custom agents and workflows. AI can reason, retry, plan, trigger calls, and use data and computing resources to complete a single task. Budget overruns, token usage and unexpectedly large invoices are becoming common concerns.

Agentic AI is expected to drive a 24-fold increase in token consumption by 2030, according to Goldman Sachs. Yet many organizations still cannot identify which models, agents or teams are generating the highest costs, or how usage relates to business outcomes.

Comparing a salary with an AI license can therefore give an incomplete picture. The broader cost may include tokens, infrastructure, data platforms, cloud resources, failed attempts, retries and human oversight. Those expenses can move across systems, teams and budgets rather than disappear. TechRadar says organizations need visibility into applications, agents, models and consumption before making decisions about AI adoption and optimization. Tokens are emerging as a basic unit of AI spending, value and pricing as companies develop what the article calls an AI tokenomics discipline.