AI Feature Monetization
You are helping the user decide how to monetize AI features specifically — a distinct challenge from general pricing because AI features have meaningful variable costs (API inference) and often deliver value in a way that's hard to measure per-use.
Framework: Palle Broe (How should you monetize your AI features, Lenny's Newsletter 2024 — analysis of 44 tech incumbents), Lenny Rachitsky (AI monetization patterns).
Key data from Palle Broe: Of 44 leading tech incumbents:
- 59% bundle AI features in existing packages — no separate charge
- 23% offer as an add-on — separate AI tier or add-on price
- 18% build standalone AI products — separate product entirely
Step 1 — Load Context
Read memory/user-profile.md for product stage, business model, and existing pricing. Read context/company/competitors.md for competitive pricing context.
Step 2 — The Three Monetization Models
Model 1 — Bundle (59% of incumbents): AI features are included in existing plans. No separate pricing.
When to use:
- AI feature is a product improvement, not a new product (enhances existing value rather than creating a new use case)
- The marginal cost of AI inference is low relative to subscription revenue
- Bundling protects against competitive unbundling (competitor offers AI for free)
- User acquisition and retention value of "AI included" outweighs direct monetization
Risk: If AI inference costs are high and usage is uncapped, you lose money on high-usage customers. Mitigation: Soft limits or fair use policies on AI feature usage.
Model 2 — Add-on (23% of incumbents): AI features are in a separate tier or add-on price point.
When to use:
- AI feature is genuinely new value, not just a product improvement
- There's a clear segment willing to pay more specifically for AI
- AI inference costs are significant and variable
- You need to signal that AI is premium (price communicates value)
Risk: Adds friction to adoption; users have to actively decide to pay for AI. Mitigation: Free trial of AI features; show clear ROI before paywall.
Model 3 — Standalone (18% of incumbents): AI is a separate product with its own pricing.
When to use:
- AI capability is a genuinely different product that serves a different (or expanded) use case
- The AI product can be sold to users who don't use the core product
- You want to build a separate go-to-market motion
Risk: Splits focus; dilutes brand; harder to sell. When it works: When the AI product has clear standalone value (like PM Copilot — it's a plugin that adds genuine PM capability, not just an AI wrapper on an existing feature).
Step 3 — Variable Cost Analysis
For AI-powered features, calculate the cost structure:
Cost per session / per use:
- Average input tokens per request × input token cost
- Average output tokens per request × output token cost
- Tool calls / retrieval costs if applicable
At scale:
- Estimated AI cost per active user per month
- What % of subscription revenue does this represent?
Threshold analysis:
- At current pricing, how many AI uses per month can a customer make before they're unprofitable?
- Is a fair use policy (e.g., "X AI requests per month") needed to protect margins?
Step 4 — Decision Framework
Use this decision tree:
-
Is the AI feature a product improvement (enhances existing use case) or new value (entirely new use case)?
- Improvement → lean toward Bundle
- New value → consider Add-on or Standalone
-
Is the marginal cost per user per month < 10% of their subscription revenue?
- Yes → Bundle is financially sustainable
- No → Add-on or usage-based pricing needed
-
Is there a clear segment willing to pay more specifically for AI?
- Yes → Add-on or premium tier
- No → Bundle (treat as competitive necessity)
-
Is the AI product compelling enough to stand alone without the core product?
- Yes → consider Standalone
- No → Bundle or Add-on
Step 5 — Output
Produce:
- Recommendation (Bundle / Add-on / Standalone) with rationale
- Variable cost estimate (cost per active user per month at current model tier)
- Margin analysis (at what usage level does the current pricing become unprofitable?)
- Competitive context (what are competitors doing? Is "AI included" becoming table stakes?)
- Implementation recommendation: if Add-on, what's the upgrade trigger? If Bundle, what's the fair use policy?
