Companies across various industries are encouraging their workforce to leverage artificial intelligence (AI) for tasks ranging from writing and analysis to automation. However, a critical question is emerging: are employees learning to use these AI tools effectively, or are they simply being pushed to use them more often without proper guidance?
The Scrutiny of AI Spending and Usage
Recent discussions among major corporations highlight growing concerns over workplace AI costs and token consumption. Amazon, for instance, has cautioned against “tokenmaxxing”—using AI tokens in high volumes without generating meaningful value, emphasizing that increased usage does not automatically translate into better work.
This issue extends beyond Amazon. Several prominent companies are re-evaluating their employees' AI tool usage and associated expenditures:
- Accenture reportedly introduced a monthly allocation of $29 per employee for Claude usage, with additional costs charged to project budgets, clarifying it's not a hard limit but a cost management strategy.
- Microsoft has urged employees to manage token spending more carefully, reducing its internal expenditure on Anthropic's Claude and making OpenAI models the default for internal AI coding.
- Meta encouraged more efficient AI token use after Google restricted its access to Gemini models due to computing capacity limits.
- Uber implemented a $1,500 monthly spending cap per employee for specific AI coding tools, including Cursor and Claude Code, after exhausting its annual AI budget early in 2026.
The Critical Gap in Effective AI Training
A July 2026 report by The Conference Board, based on a global survey of nearly 1,300 workers, revealed that while 55% of employees regularly use AI, only 33% had received employer-provided AI training in the preceding six months. This significant gap suggests that access to tools does not equate to effective utilization.
Employees need more than just prompts; they require an understanding of which tasks are best suited for AI, how to formulate precise instructions, and the importance of verifying AI-generated responses. Ankur Dhingra, CEO of ProHance, noted that companies often encourage AI consumption without adequate focus on token efficiency, cost-effectiveness, and reducing rework.
A study published in Nature Human Behaviour further highlighted this challenge: management consultants given ChatGPT access and training performed better on data-science tasks but showed no retention of technical knowledge when tested without AI. This implies that completing a task with AI does not necessarily build underlying employee skills.
Optimizing AI Use and Cost Management
Rashmi Nair, VP-HR at OpenText India, advises against using the most advanced AI model for every task, as this can unnecessarily inflate costs. Simpler models are often sufficient for tasks like data extraction, sorting, and summary generation, reserving more complex models for advanced analysis and coding. Employees should also be vigilant about not inputting sensitive client or personal information into unapproved AI tools and always verifying critical outputs against original sources.
Companies can implement strategies to manage costs by routing simpler requests to lower-cost models and deploying advanced ones only when necessary. Improving data retrieval systems and minimizing duplicate files can also reduce unnecessary AI usage.
Measuring Outcomes, Not Just Usage
Dhingra cautioned against using token consumption as a measure of employee performance. High usage doesn't indicate accuracy, successful task completion, or time saved. Instead, he advocates for assessing improvements in work quality, turnaround time, reduced rework, and the complexity of tasks employees can handle.
A February 2026 study in Management Science, involving 4,867 software developers, found that AI coding assistants increased completed tasks by 26.08%. However, the gains varied, with less-experienced developers seeing greater productivity boosts. This reinforces that workplace AI productivity should be measured by tangible outcomes.
Ultimately, effective AI use for employees means knowing when and how to apply the right tool and critically evaluating its output. For employers, it entails providing targeted, task-specific training, clear data-use guidelines, and sensible cost controls. The true measure of workplace AI adoption lies in whether it helps produce better, higher-quality work, not just in the volume of prompts or tokens consumed.