AI pilot era is ending. Now worry about the bill

As AI technology transitions from experimental pilots to widespread production deployments, Indian businesses are grappling with a crucial question: "Kitna deti hai?" - how much value they are receiving for every rupee spent on AI. Snowflake India's Managing Director, Vijayant Rai, highlights that cost considerations have become a top priority as companies adopt AI across more employees, customers, and business processes.

Rai explains that customers are now more conscious about the costs associated with AI, such as token consumption and returns on investments. The shift from sporadic use cases to larger-scale deployments has made it essential for enterprises to address the cost side of AI. Snowflake has around 13,900 customers globally, with approximately 13,600 already utilizing AI features on its platform.

To tackle the cost concerns, Snowflake has introduced dynamic model routing within its Cortex AI Gateway and integrated it into Snowflake CoCo and CoWork. This feature enables the selection of AI models based on factors like cost and performance, which can help reduce expenses without compromising on quality. Rai emphasizes that enterprises should consider using lower-cost models for simple, repetitive tasks, while reserving more powerful models for complex scenarios.

Snowflake is also expanding access to open models from various providers, including DeepSeek-V4-Flash 0731, GLM-5.3, and others. According to Srirangam, Senior Regional Vice President of Solution Engineering, model routing allows customers to let the system make better decisions rather than restricting them to a single model choice.

However, he cautions that while dynamic model routing can improve token efficiency by up to three times in certain workloads, the actual cost savings may vary depending on specific enterprise use cases.