We grew AI usage 52x, then cut our costs 90% (www.prosus.com)

Prosus has dramatically scaled its AI operations, increasing token usage by 52 times to 8.2 trillion tokens per month while effectively reducing costs by 90%. The company operates AI agents alongside 40,000 employees and 5 million partners to serve 1 billion customers. This transformation highlights the economic challenges of AI at scale, necessitating a strategic approach that balances model sophistication with cost efficiency. Key strategies included routing tasks to appropriate models based on complexity, reducing unnecessary request data, enhancing GPU utilization, and creating specialized models for repetitive tasks. These initiatives not only slashed operating expenses but also optimized performance without compromising quality. The implications for the AI/ML community are significant: as AI technology continues to evolve, companies can leverage these insights to sustainably integrate advanced AI into their operations. By continually evaluating model performance, streamlining data needs, and optimizing resource use, businesses can make AI an integral, cost-effective part of their daily functions. This proactive approach positions organizations to harness AI’s full potential, transitioning from sporadic use to continuous, efficient deployment, fundamentally shifting the economic landscape of AI applications.