As AI coding becomes part of everyday software development, companies are discovering a new productivity constraint: employees can have plenty of working hours left while the AI resources increasingly used to get the work done come with limits and costs.
WHAT’S HAPPENING
For years, companies managed technology workers around familiar resources: salaries, working hours, headcount, computers and cloud infrastructure.
AI is adding another one to the list:
Tokens.
Every interaction with a generative AI system consumes computational resources, and at enterprise scale those costs can add up quickly.
Uber provides one of the clearest examples. After aggressively encouraging engineers to use AI coding tools, the company burned through its entire 2026 AI coding budget by April. The company subsequently began focusing more closely on costs and the value generated from that AI usage.
Meta is confronting the same question. Instagram head Adam Mosseri said companies may eventually need to impose AI-token spending limits on individual engineers, potentially allocating larger budgets to workers who demonstrate that their AI usage produces sufficient returns.
WHY IT MATTERS
This creates a new workplace equation.
A company can pay an employee for an eight-hour workday while simultaneously limiting access to a tool that may now handle a meaningful portion of that employee’s coding, research or analysis.
At the same time, simply giving workers unlimited AI resources creates another problem.
More token consumption does not automatically mean more productivity.
Uber executives have publicly questioned how directly increased AI coding usage translates into additional useful products and features.
Google Cloud is even advising software engineers to think deliberately about token efficiency, arguing that excessive context can increase cost and latency while potentially reducing the quality of AI output.
The challenge is therefore becoming less about whether companies should use AI and more about how much AI should be allocated to each job — and what return companies should expect from it.
WHO BENEFITS
Highly productive AI users could gain access to larger AI budgets if companies begin allocating tokens according to demonstrated results.
Employers may gain better control over rapidly growing AI expenses while learning which workflows actually produce measurable value.
AI infrastructure and management companies could benefit from demand for systems that monitor token consumption, costs and employee usage.
Workers overall could benefit if AI budgets are designed around outcomes rather than simply forcing employees to maximize activity during every hour of the workday.
WHO LOSES
Employees with rigid AI limits could see productivity fall if they lose access to tools in the middle of important work.
Companies that equate token consumption with productivity could spend heavily without receiving corresponding business benefits.
Companies that restrict AI too aggressively face the opposite risk: paying skilled employees while limiting the tools that make them more productive.
And traditional workplace measurement systems based largely on hours worked may become increasingly difficult to apply when AI can dramatically change how much work gets accomplished during those hours.
WHAT HAPPENS NEXT
Expect AI-token management to become another layer of corporate budgeting.
Companies may establish departmental allowances, per-employee limits, approval thresholds and different levels of AI access depending on job responsibilities.
The trend is already spreading beyond private companies. The U.S. Army reportedly had to restore AI usage limits after a shared pool advertised as offering effectively unlimited token access was exhausted.
The larger shift is easy to miss.
For decades, companies primarily asked:
How many people do we need to do the work?
The AI workplace may increasingly require a second question:
How much machine intelligence should each person be allowed to use?
