Organisations looking to expand their use of agentic AI need a cost strategy that can grow alongside adoption. However, this growth may bring higher spending than initially expected. While computing costs might decline, token consumption, which is the main pricing component for many AI tools, is increasing rapidly. If left unchecked, this trend could reduce value over time.
Modern AI adoption is no longer only about innovation. It requires careful financial decision-making. Agentic AI tools depend heavily on processing tokens, the fundamental units used by large language models. These tokens are used in nearly every prompt, task or output, and as your AI systems become more complex, they consume more tokens. As these tools take on greater autonomy and manage intricate workflows, token usage and associated costs can rise quickly.
Although CIOs are already familiar with infrastructure and initial setup expenses, token-based billing represents a major shift from traditional IT models. Unlike legacy software with mostly fixed costs, agentic AI demands ongoing spending similar to electricity or cloud services. While third-party tools with built-in AI can simplify implementation, they also reduce visibility and control over costs. On the other hand, creating in-house tools may prove more efficient in the long run but requires larger upfront investment and stronger strategic oversight.
A key challenge is keeping AI strategies flexible. The best path to scale varies depending on factors like the intended use, technology architecture and whether the organisation decides to buy, build or optimise delivery. Leaders must carefully measure the immediate and future impact of each decision to ensure AI delivers true returns, with a clear understanding of cost factors.
When managed well, agentic AI can dramatically boost enterprise productivity by enabling scalable digital workforces. However, these benefits will only appear if leaders maintain clear control of the core cost drivers, especially token consumption.