The way we use and pay for AI is changing

The way businesses utilize and pay for artificial intelligence (AI) is undergoing significant changes. Traditionally, companies purchased AI services like Microsoft 365 on a flat-fee basis, with invoices received alongside other essential subscriptions like Zoom. However, this method is giving way to more granular pricing models.

Recently, major players such as Anthropic and GitHub have announced changes to their AI services. Anthropic will separate agentic usage from its Claude subscription and charge it separately starting June 15, while GitHub Copilot will transition to a token-based pricing model on June 1. These changes come without any adjustments on the published pricing pages, leaving enterprises uncertain about the implications.

This shift in AI pricing has caught finance teams off guard, as it requires a new approach to monitoring and managing costs. Companies like Uber have already faced the consequences, burning through their AI budgets by April. Even larger firms like KPMG are projecting average AI spend of $207 million over the next year, nearly double compared to the previous year.

In response to these challenges, businesses are reevaluating their AI usage. Companies are limiting access to powerful AI models, canceling licenses, and restricting employee access to agentic tools. While these actions aren't due to a lack of AI effectiveness, they are driven by the unpredictability of AI costs and the lack of contractual leverage for businesses.

To navigate this new landscape, businesses should insist on specific terms in their contracts, such as price-change notice provisions, tokenizer stability clauses, audit rights over consumption data, and exit and portability terms. These provisions would provide clearer guidance and control when AI usage patterns change unexpectedly.

As the expiration dates for trials and upgrades approach in July, the impact of these AI pricing changes will become more apparent in August. Businesses that have not yet addressed their AI spending may face significant challenges, highlighting the need for improved supplier relationships, FinOps functions, and legally sound contracts that align with the evolving nature of AI consumption.