Microsoft wants to put a dollar value on AI’s work – but can you trust the math?
With Agent 365 Business Analytics, Microsoft could reshape enterprise spending, but there’s a critical difference between estimated value and money saved.
Microsoft aims to quantify the business value of AI work with the launch of Agent 365 Business Analytics. This tool attempts to measure AI usage data against labor-market economics to estimate the economic value of AI-assisted tasks. The company leverages data from the US Bureau of Labor Statistics, the O*NET database, and LinkedIn to create a comprehensive picture of AI spend versus value produced.
Instead of traditional software accounting methods, Agent 365 Business Analytics measures AI costs both relative to human labor and in terms of technological expenses. For instance, it compares the time a task would typically take a human to perform against the portion of the task completed by AI, factoring in the estimated value of the contribution made by artificial intelligence. Additionally, it weighs the actual AI processing costs, such as subscriptions, tokens, on-premises infrastructure, and cloud support.
These calculations result in a hybrid labor estimate that combines human contribution, AI-replacement savings, AI costs, and potential AI savings over human costs. However, these estimates are directional and not necessarily reflective of real-world expenses or time savings. Microsoft explicitly states that assisted hours are an estimate and not a measured time saved or financial return.
While Agent 365 Business Analytics can provide a sense of the economic value of AI-assisted work, it may not definitively demonstrate that companies have saved real money. This tool is one source of possible truth for decision-making, but not the absolute truth. Microsoft aims to help companies manage their AI infrastructure, control costs, and integrate various AI offerings without providing the underlying AI models or applications.
Written by urgent.news from ZDNet's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.