AI token costs are exceeding some employees’ salaries

2026-06-03 · Source: Semafor · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Cloud Computing & IT Infrastructure · Depth: Novice, medium

Summary

Enterprise AI is grappling with escalating token costs and challenges in measuring return on investment. JPMorgan's Payments division reports some employees' token spending exceeds their salaries, leading corporations to monitor and potentially ration AI tool access. IBM Vice Chairman Gary Cohn notes "massive over-investment," while startup CEOs attribute spending cuts to difficulties in quantifying AI's value, particularly in software engineering. Microsoft is launching Scout, an AI assistant powered by OpenClaw, to bring agentic AI capabilities like scheduling and expense filing to broader business roles, with costs tied to GitHub Copilot credits. Concurrently, tech firms like Nvidia and Perplexity are investing in local AI processing on PCs, driven by rising data privacy concerns. OpenAI CEO Sam Altman is also engaging with US lawmakers and developing a policy framework, following President Trump's executive order requiring 30 days' notice for new powerful AI models, signaling impending guardrails for advanced AI.

Key takeaway

For Directors of AI/ML overseeing enterprise deployments, carefully evaluate token costs against tangible ROI, as current spending often exceeds measurable value. Prioritize AI applications with clear, quantifiable outcomes, such as sales or customer experience, and explore local processing solutions for sensitive data to address growing privacy concerns. Additionally, prepare for evolving regulatory landscapes, as new guardrails and policy frameworks are actively being developed by governments and leading AI labs.

Key insights

AI adoption faces high token costs, ROI measurement challenges, and increasing regulatory and data privacy scrutiny.

Principles

• Frontier AI models offer a 10-20% performance edge over smaller, fine-tuned alternatives.
• Increasing data privacy concerns drive demand for local AI processing solutions.
• User trust in AI customer service remains low, often preferring human interaction.

Method

Perplexity's "Computer" autonomously selects local vs. cloud workloads based on data sensitivity, while Microsoft Scout requires user authorization and pre-approval for security.

In practice

• Monitor AI token usage to manage escalating operational costs.
• Prioritize AI automation for tasks with clearly measurable ROI, like sales.
• Implement local AI processing for sensitive data to address privacy concerns.

Topics

• AI Token Costs
• AI ROI
• AI Agents
• Data Privacy
• On-device AI
• AI Policy

Best for: CTO, VP of Engineering/Data, AI Product Manager, Director of AI/ML, Executive, Policy Maker

Related on AIssential

• AI startup offers free home cleaning to train robots — AI development faces a critical juncture balancing rapid innovation, economic viability, and urgent safety and societal concerns.
• Why companies are penny-pinching on tokens — Frontier AI models offer performance gains but incur significant token costs, challenging ROI measurement for enterprises.
• What you'll pay for AI agents will be wildly variable and unpredictable — AI agents incur vastly higher and unpredictable token costs, primarily driven by input tokens and cache reads, without guaranteeing improved performance.
• I think Anthropic and OpenAI have found product-market fit — Frontier AI labs have found product-market fit for coding agents, driving significant enterprise revenue through aggressive pricing.
• 📈 Why AI bills rise as costs fall — AI costs escalate despite falling token prices due to demand elasticity and agents' hidden, amplified token consumption.
• Your AI Pricing Problems Are Actually Business Model Problems — AI's usage-based costs fundamentally break traditional SaaS unit economics, demanding a business model transformation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Semafor.