Six months ago many firms announced staff cuts, citing artificial intelligence as a route to greater efficiency and cost reduction.

Early automation efforts and reversals

Several large companies that pursued aggressive automation have started rehiring employees after pilot programs proved more expensive or less effective than expected.

Microsoft is reducing some licenses for Claude Code because of high costs, while Uber reportedly exhausted its annual AI budget by April.

Nvidia has acknowledged that compute expenditures for engineering teams are already surpassing payroll costs, creating fresh budgetary pressures.

Starbucks experimented with an AI system for inventory control but found human teams delivered more reliable results in operational practice.

IBM, having planned some AI-driven replacements, has shifted to increasing hires of junior staff rather than continuing broad automation cuts.

Why the economic case weakened

Organizations initially compared a worker’s salary directly to token expenses, but token prices were lower and model usage was smaller at that time.

As models consume tokens increasingly over time, the simple salary-to-token comparison no longer reflects the full, growing operational cost structure.

Hidden expenses, including engineers to maintain AI systems and remediation of model errors, add materially to total cost and reduce expected savings.

Quality degradation also carries monetary effects; consumer-facing products such as Duolingo faced user-impacting issues after premature automation attempts.

Market and sector implications

This sequence of events signals an end to the first corporate wave of AI hype and tempers assertions that automation alone will deliver large-scale headcount savings.

Investors may reassess valuations of AI-branded IPOs, including firms like xAI, OpenAI and Anthropic, as profitability remains elusive for many participants.

Even after investments measured in hundreds of billion dollars on compute, several major AI projects stay unprofitable and continue to require substantial ongoing funding.

Going forward, AI is likely to occupy specialized roles—augmenting code, research, and data tasks—while companies balance token costs and human labor expenses.

Outlook for corporate adoption

Some firms will continue to absorb token costs until economics improve, but many are already recognizing that wholesale replacement of staff is neither immediate nor cost-effective.

As the market adjusts, expectations for quick, universal automation will moderate, and firms will prioritize targeted deployments that demonstrably improve outcomes.