The Soaring, Unpredictable Cost of AI
Generative AI is no longer an experiment; it's becoming a core part of business operations, and it's expensive. Unlike traditional software with predictable per-user license fees, AI costs are volatile. They hinge on computational power, data processing,
and the complexity of the tasks performed. According to IBM, the average cost of computing is expected to climb significantly, with executives citing generative AI as a primary driver. This has created a strategic problem for CFOs and CTOs: how do you budget for and control spending on something that scales in such a nonlinear and often unpredictable way?
What is a 'Token' in AI?
To understand Deloitte's vision, you first need to understand the 'token.' In the world of Large Language Models (LLMs), a token is the fundamental building block of text or data. Think of it as a piece of a word, a whole word, or even punctuation. When you ask an AI to write an email or analyze data, the model breaks your request down into tokens and generates its response one token at a time. Crucially, almost every commercial AI service prices its usage based on the number of tokens processed—both for the input you provide and the output you receive. A common rule of thumb is that 100 tokens equate to about 75 English words.
Deloitte's Vision: Tokens as the New Currency
Deloitte's analysis argues that as AI becomes more integrated, tokens are becoming the true unit of cost, eclipsing older metrics like software licenses or headcount. The consulting firm suggests that business leaders should treat AI economics with the same rigour as capital allocation, recognizing tokens as the new currency. In this view, the key business calculation is no longer just about whether to invest in AI, but about how efficiently the organization converts tokens into revenue, productivity, or other measurable business value. This creates a direct link between the technical consumption of AI services and tangible financial outcomes.
Why This Matters for Business Strategy
Framing AI costs in terms of tokens provides a new level of transparency and control. It allows a business to measure exactly what it is paying for and how efficiently it is consuming AI resources. By tracking token consumption, a company can analyze the cost-effectiveness of different AI models, applications, or even entire departments. This 'tokenomics' approach enables better forecasting, budgeting, and risk management. For example, a company could determine the precise token cost of processing a customer service inquiry or generating a monthly financial report, allowing for much more granular cost-benefit analysis than ever before.
The Challenges of a Token-Based Economy
While promising, a shift to a token-centric view of work is not without its hurdles. One major risk is mistaking token consumption for productivity. Just as measuring a programmer's output by lines of code proved to be a flawed metric, tracking only the number of tokens used can create bad incentives. Employees might be encouraged to 'burn' tokens on unnecessary tasks simply to appear productive. The real challenge for businesses will be to move beyond simply counting tokens and develop sophisticated metrics that connect token usage to genuine business impact and value creation, a task that requires careful governance and a clear understanding of strategic goals.
