Facing rising costs, more and more companies are beginning to rein in their employees’ use of AI, according to a new article from the Wall Street Journal.
An expensive bill: Each use of generative AI tools has an associated token cost—simple word generation, for example, can cost 500 to 1,000 tokens per paragraph. While the cost of tokens has decreased as more tools enter the market, the expense is still high, particularly for newer AI models.
-
While many companies simply associate AI use with productivity and efficiency gains—and therefore cost savings—the reality is often more complicated.
-
Complex problem-solving done through AI can cost more than the solution will save, and where the AI-provided solution misses the mark, human capital will be needed to fill in the blanks, adding more expense.
Tracking token use: After an initial free-for-all where employees were given largely unfettered access to AI tools—and strongly encouraged to make use of them—companies are beginning to track token use to connect specific costs to specific use cases.
-
Some companies are building dashboards to monitor AI usage in general, with the goal of reining in associated costs.
-
Companies may limit certain AI models to certain use cases or tasks.
-
While some companies associate high use of tokens with high performers, others have found that it points to larger inefficiencies in the way their work is done.
Takeaways:
-
As costs begin to mount, companies may want to begin closely mapping employee AI use, including matching certain models to certain use cases to maximize cost efficiency and effectiveness.
-
Usage rates are not necessarily an indicator of employee performance, and different frameworks may need to be applied to different functions, with AI literacy and model or tool capability potentially serving as another differentiating basis.
