Companies and AI: Towards an Internal Budget Hierarchy

A New Economic Approach to AI

Companies are now adopting a GLP strategy to manage the use of artificial intelligence, an approach that is not without cost. As bills begin to pile up, some executives are asking their employees to reduce their AI consumption. This transition marks a significant shift from the previous mindset of AI being accessible to all.

Stephen Council, Polly Thompson, and Charles Rollet from BI conducted interviews with workers and leaders to understand the shock felt in response to this evolution. AI, once perceived as an all-you-can-eat buffet, is transforming into an à la carte menu where each use must be justified by a limited budget.

Towards Usage-Based Pricing

The concept of "tokenmaxxing," which involved maximizing the use of AI tokens, is losing its appeal as AI giants raise their prices and adopt usage-based pricing. This change could predetermine the winners and losers within companies, depending on the AI budget allocated to each team.

The new dietary regimes regarding AI will not be applied uniformly, which could widen the gap between those who have access to AI and those who do not. Teams with the largest budgets will have the best chances of proving the value of AI, while those with fewer resources may struggle to demonstrate their potential.

The Challenge of Measuring Return on Investment

A persistent problem is the lack of an industry standard for evaluating the return on investment (ROI) of AI projects. Assessing productivity through the use of AI tokens is becoming less popular, and companies are struggling to justify their expenditures without clear metrics.

Ideally, resources should be allocated to the best ideas. However, projects benefiting from the largest number of AI tokens may be perceived as the best ideas simply due to their additional budget. Executives may also be reluctant to abandon costly projects, falling into the trap of sunk cost syndrome.

An Emerging Caste System

The result of these dynamics could be the creation of a quasi-caste system for AI tokens, generating internal tensions that are difficult to reverse once established. This situation could stifle innovation and affect the long-term competitiveness of companies.