Businesses Finding It Harder to Predict AI Spending

AI summary of the source article

A report by The Wall Street Journal highlights that only 11% of nearly 400 surveyed companies can accurately project their AI expenditures. Because AI models behave more like human workers, pricing differs from standard software: higher-priced models may complete tasks more efficiently, whereas cheaper models can trigger token run-ups on complex tasks. Researchers found lower-priced models were actually more expensive in 32% of 6,800 tested scenarios. Additionally, a PYMNTS Intelligence report found 60 surveyed U.S. firms deployed AI across an average of seven functions, including payments and finance, with 62% spending over $10 million on new tools in the past year.

Why it matters

Token run-ups and model inefficiencies make AI budgeting unpredictable for enterprises adopting the technology across core operations like payments and finance.

Key facts

• Only 11% of nearly 400 surveyed businesses could accurately project their AI costs, according to a report cited by The Wall Street Journal.
• Researchers found that in 32% of more than 6,800 tested scenarios, lower-priced AI models ended up costing more than higher-priced models.
• A PYMNTS Intelligence survey found 62% of companies spent over $10 million on new AI tools in the last year, deploying them across an average of seven business functions.