Companies struggle to measure AI's ROI

2026-06-05 · Source: Semafor · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Corporate Strategy & Leadership, Public Policy & Governance · Depth: Fundamental Awareness, extended

Summary

Microsoft launched Scout, a new AI assistant powered by OpenClaw, designed for enterprise tasks like scheduling and expense filing, with costs tied to GitHub Copilot subscriptions. Companies are grappling with AI's return on investment, with IBM's Gary Cohn noting "massive over-investment" and JPMorgan observing token costs exceeding some employee salaries. This struggle is partly due to difficulties in measuring AI's value, especially in software engineering workflows, a market that jumped sevenfold to \$4 billion from 2024 to 2025. AI safety concerns are escalating: Anthropic called for a development slowdown due to models like Claude writing 80% of its code, while top AI CEOs jointly warned about AI's potential for bioweapon creation. Data privacy concerns are driving a bet on on-device AI, with Nvidia embedding more capabilities into laptops. The AI infrastructure buildout faces public opposition, leading to a California city banning data center construction, and ballooning energy demands. Chinese AI firm DeepSeek is nearing a \$7.4 billion funding round, challenging Silicon Valley with aggressive pricing.

Key takeaway

For technology executives navigating AI investments, the current landscape demands a critical re-evaluation of strategies. You should prioritize developing robust ROI metrics for AI initiatives, especially for software engineering workflows, and explore hybrid deployment models that balance performance, cost, and data privacy. Be prepared for increased regulatory scrutiny and public pushback on infrastructure, while also monitoring the competitive landscape from aggressively priced international AI solutions.

Key insights

AI's enterprise adoption faces ROI measurement hurdles, high token costs, and significant safety/infrastructure challenges, even as new agents and on-device solutions emerge.

Principles

• AI ROI is difficult to quantify, especially for non-measurable tasks.
• Frontier AI models often yield better performance despite higher token costs.
• AI's autonomous capabilities necessitate development slowdowns for safety.

Method

Perplexity's "Computer" autonomously selects local versus cloud workloads based on data sensitivity to enhance privacy. Google finances a "virtual power plant" to curb household/business energy consumption, freeing 100 MW for data centers.

In practice

• Monitor AI token usage to manage escalating operational costs.
• Prioritize AI automation for measurable tasks like sales and CX.
• Explore on-device AI solutions for enhanced data privacy.

Topics

• AI Return on Investment
• AI Governance & Safety
• AI Infrastructure
• On-device AI
• Enterprise AI
• Token Costs

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

Related on AIssential

• How AI Is Driving Revenue, Cutting Costs and Boosting Productivity for Every Industry in 2026 — AI is becoming essential infrastructure, driving industry-wide intelligence and ROI focus.
• Why enterprise AI ROI starts with observability — AI observability links model behavior to business outcomes, enabling precise ROI measurement and risk mitigation.
• How AI is changing Software Engineering: A Conversation with Gergely Orosz, @pragmaticengineer — Measuring AI token output can lead to counterproductive "token maxing" behavior among engineers.
• TAI #203: OpenAI, Anthropic, and Wall Street Race to Build the AI Deployment Layer — Enterprise AI adoption requires significant hands-on integration and workflow redesign, not just model access.
• AI token costs are exceeding some employees’ salaries — AI adoption faces high token costs, ROI measurement challenges, and increasing regulatory and data privacy scrutiny.
• Uber's AI Spending, the Pope on AI, TikTok and Universal's AI Music Deal — AI adoption faces ROI measurement, ethical governance, infrastructure gaps, and content licensing challenges across industries.

Open in AIssential →

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