Pricing AI products beyond the user count model

In the rapidly evolving landscape of AI products, pricing strategies remain one of the most challenging yet critical aspects of building a successful business. Kshitij Grover from Orb recently shared invaluable insights on revenue engineering for AI products, particularly focusing on pricing models that transcend traditional approaches. His presentation cuts through the noise to address a fundamental question many AI founders struggle with: how to effectively price products where value creation doesn't necessarily correlate with user count.

Key Points

• AI products often create value in ways that don't scale linearly with user count, requiring more nuanced pricing models that align with actual value delivery
• Usage-based pricing presents compelling advantages for AI companies, including lower barriers to adoption, natural expansion opportunities, and better alignment with customer value
• Implementing effective pricing requires defining the right usage metrics, establishing clear pricing tiers, and developing a comprehensive strategy for handling overages and pricing changes
• Customer feedback loops are essential when testing new pricing models, helping companies refine their approach based on real-world usage patterns and customer responses

Expert Analysis

The most insightful takeaway from Grover's presentation is the concept of "value metrics" versus simple usage metrics. Unlike traditional SaaS applications where user seats made sense, AI applications often deliver value through data processed, insights generated, or outcomes achieved. This distinction is crucial because it fundamentally changes how companies should think about monetization.

This matters tremendously in today's AI landscape where we're witnessing a proliferation of tools that deliver exponential value without corresponding increases in usage volume. When companies like OpenAI shifted from charging by token to charging by capabilities (with different rates for different models), they demonstrated this principle in action. The industry is clearly moving toward pricing models that better reflect the actual value delivered rather than arbitrary usage metrics.

Beyond the Presentation: Real-World Applications

One fascinating example not covered in the presentation is Anthropic's approach with Claude. Rather than simply charging by token count, they've introduced different pricing tiers for different model capabilities, with Claude Opus commanding premium prices for its enhanced reasoning abilities. This illustrates how companies can segment their offerings based on value delivery rather than just consumption metrics.

Another instructive case is Midjourney's evolution from subscription-only to a hybrid model that combines subscription access with usage-base