Meta executives say the future or an AI based on an engineer

Meta is reassessing the cost of using internal AI tools. Adam Mosseri, director of Instagram, in an interview with the podcasts, stated that the AI token costs consumed by some engineers could be close to their pay levels within the next year or two, and that the company would then need to set a personal level of use.

Meta started talking about scale management.

The AI token cost, referred to in Mosseri, refers primarily to the costs incurred by the model in processing the hint and generating the response. He indicated that in the future, an enterprise might need to manage its staff ' s mobilization expenditure on the AI tool, like the management of the computing, storage and human budget.

According to him, the allocation criteria could be linked to the company ' s judgement of efficiency in the use of the personnel concerned if the future implementation were to be based on an engineer-set, i.e., whether these expenses would provide a positive return.

Internal compression invalid consumption

Mosseri disclosed that Meta has not yet set an AI token ceiling for any employees, but the company has begun to reduce some of its limited-value internal attempts. Previously, Meta had closed the internal AI token consumption list to reduce related expenditure.

He stated that it was not difficult to create a fast-consuming token-use scenario, but that such practices did not necessarily produce sufficient value.

Technology companies generally tighten their budgets.

The report mentions that Meta is not the only company to adjust AI experimental expenditure. Uber ran out of AI programming budget for the current year before April 2026. Microsoft, on the other hand, cancelled Claude Code’s license because of the increased cost of token, and instead encouraged engineers to make more use of their own Copilot CLI.

At the same time, Mosseri believes that AI token costs may fall in the longer term as model manufacturers compete for prices for users. Until then, however, how to control the scale of internal use is becoming a reality for large technology companies.