Claude Code: The Crucial Balance Between ROI and Innovation in AI

The Importance of Return on Investment in AI According to Boris Cherny

Boris Cherny, a key figure at Anthropic, highlights a crucial question for businesses: how to balance return on investment (ROI) while allowing employees to experiment with artificial intelligence (AI). According to him, while the focus on ROI is justified, it is imperative that employees, regardless of their role, have access to the necessary tokens to test and innovate with AI. This approach would enable companies to effectively manage their costs.

During a recent discussion organized by Scale AI, Cherny emphasized that ROI is the right framework for evaluating investments in AI. He explained that companies should not only focus on expenditures but also on what they are getting in return. This perspective is particularly relevant in the current context where budgets for AI tokens are under pressure.

Budget Concerns and AI Tokens

Jesse Chen, Director of Product Management at Meta, questioned Cherny about the concerns expressed by Andrew Macdonald, COO of Uber. Macdonald recently spoke about the rising costs of AI tokens and the need to justify these expenses. Tokens, which are units of text used by language models to process prompts, are essential for the functioning of chatbots like Claude from Anthropic and other generative coding tools.

Cherny reiterated that ROI is crucial, but he also warned against an overreaction to budgetary concerns. He explained that companies should allow their employees to experiment freely with tokens, without fear of financial repercussions. This freedom to experiment is essential for discovering new internal applications of AI that can then be optimized to reduce costs.

Innovation Arises from Experimentation

According to Cherny, the best ideas for using AI often emerge unexpectedly, for example, from an accountant or a member of the marketing department. These innovative ideas can transform internal processes and lead to the creation of new products. Therefore, it is crucial not to restrict access to tokens, as this could stifle innovation.

Anthropic offers several solutions to help companies manage their costs, including user-based cost controls. Other industry leaders, such as Sam Altman from OpenAI, also share these concerns regarding the ROI of investments in AI.

Economic Challenges for AI Companies

Cherny pointed out that companies like Anthropic, which generate tokens, have an economic interest in selling their generative AI models and tools. This is particularly relevant as highly anticipated IPOs approach. He explained that each token used represents an opportunity cost, as it could be sold to a customer.

The measurement of ROI evolves with the rapid advancements of AI models. Cherny mentioned that assessing the percentage of code written by AI becomes obsolete when AI generates all the code. He suggested focusing on the acceleration of code per engineer and identifying bottlenecks in the development process. The goal is to unleash employees' creative potential to generate ideas more quickly and efficiently.