I recently had the opportunity to sit down with Ravi Belani, CEO and Founder of @AlchemistAcc to discuss how Agentic AI should actually be priced — and where founders are getting it wrong. Ravi has a sharp lens on enterprise AI, and as always, the conversation was grounded in first principles Most teams are still trying to price agents like SaaS. That doesn’t work. Agentic systems don’t behave like seats or simple API calls. They generate variable compute loads, multi-step reasoning chains, orchestration overhead, governance layers, and human-in-the-loop review. The cost structure is nonlinear — and often opaque A few observations from the investing side: • Seat-based pricing breaks quickly due to unpredictability in token usage and consumption • Pure token pricing makes TCO unpredictable which stalls enterprise adoption. • The winners will price against business value — cost saved, revenue generated, cycle time reduced, risk mitigated. If your agent eliminates $5M of operational expense, price it against that. If it materially improves ticket resolution, fraud precision or underwriting velocity, price it against margin expansion. That’s how enterprise buyers think. youtube.com/watch?v=LnTr-A9G…