I'm writing a guest chapter on AI pricing for another authors upcoming book. So I'm thinking about this a lot lately, and wanted to give you a sneak-preview of a linkedin post that will come out later next week. - I've done maybe 10 pricing projects involving significant AI functionality this year. Here is how I think about it. AI should be considered a 2-layer stack: 1️⃣ The AI compute 'fuel' (i.e. token pricing at OpenAI) 2️⃣ The AI solution (i.e. the value you add on top of AI) The dilemma with AI pricing is that currently: ◾ Fuel is expensive - at least way more so than traditional SaaS. ◾ Solutions are immature and early stage, not yet adding a lot of value. So any AI pricing model needs to both work today AND tomorrow. AI PRICING TODAY: 🔹 Charges usage-based on fuel consumption to ensure costs are covered, as usage patterns of customers is often unpredictable. 🔹 Is mostly focused on low barriers to entry to get users onboarded in order to develop the solution layer and get data on behaviour and cost patterns. 👉 This is unsustainable as fuel costs will drop and 2025 customer will refuse to pay a price-per-token (or token equivalent) that is based on 2024 token costs. This is especially true for enterprise. AI PRICING TOMORROW: 🔹 Charges based on the outcome created by the solution layer and just factors in fuel costs in the use case. 🔹 Protects against cost-downside of over-usage with 'fair usage limits'. HOW I SOLVE FOR THE TODAY-TOMORROW PROBLEM IN AI PRICING 🔸 Focus on speed: just get usage and adoption as fast as possible. You likely have a core non-AI product that monetizes just fine. Consider AI a 'development budget' and focus on profitability later. 🔸 Tell customers you are BOTH charging for fuel and for a solution outcome. Educate them. But keep it 90% fuel and 10% solution early on. 🔸 Over time: shift $$ from fuel to solution pricing. Cut Fuel pricing aggressively, even anticipating future cost reductions. Consider solution pricing separately and from a value perspective.