The quest to monetize artificial intelligence (AI) is fraught with unpredictability. Free versions of AI like ChatGPT offer a tantalizing glimpse at what’s possible, but as firms seek to recoup their hefty investments, paid-for features are on the rise.

However, setting prices for these services is surprisingly complex. Simon Gooch from Saviynt points out that trying to predict costs over months or years makes no sense due to the volatile nature of tokens and agentic AI systems.

The issue boils down to unpredictability: subtle variations in prompts can yield different answers, and the number of tokens consumed by businesses and consumers is skyrocketing. Microsoft and Uber have both faced unexpected token usage challenges, highlighting how difficult it is to manage costs accurately.

Companies are finding creative ways around this, such as using flat-fee personal accounts, but these may not be sustainable in the long term. As AI platforms face pressure from shareholders, stricter cost controls are inevitable, predicts Will Venters.

The situation becomes even more complex when integrating AI into products for widespread use. Managers must consider not just core software development costs but also testing, security, and guardrails, which can lead to unexpected token expenditures.