Tokenmaxxing: Why AI consumption needs control

image via TechRadar

August 17, 2026, 9:47 AM

• •The article cites Anthropic's Claude Fable 5 pricing at $10 per million input tokens and $50 per million output tokens.
• •It says proprietary Pleo data showed Anthropic rising from 12th to 7th among top tech merchants for UK software customers between Q4 2025 and Q1 2026, while average spend per customer rose 43.0%.
• •The author says inefficient AI use includes open-ended prompting, using overly expensive models for routine tasks, and overlapping or unauthorized tools.
• •The piece recommends tracking AI spend by department, team, budget and model alongside performance measures such as drafting speed, code review times and support resolution times.
• •It argues finance teams can add AI-specific reporting, spend policies and approval thresholds to control costs while preserving innovation.

The article argues that business spending on AI is rising quickly as more capable models become more widely used. It says companies often lack visibility into which teams are consuming tokens, which models they use, and whether that spending produces meaningful productivity gains. The author highlights waste from open-ended prompts, model mismatch, duplicate tools and shadow AI subscriptions. The piece concludes that finance teams should monitor AI usage by department, model and outcome so organizations can improve return on investment without slowing adoption.

Read original article