How to price AI features when every user costs you a different amount
CommercialYam8 · reddit · 2026-11-13
Adding an AI feature to a SaaS product breaks normal margin assumptions: every request costs money and heavy users can cost far more than average ones. The author's framework: measure first (typical user cost vs. the heaviest 10% — the gap is usually bigger than expected), decide what's included (a monthly allowance is easiest to explain; heavy features like long-document analysis or bulk generation work better as credits or add-ons), and design the over-limit experience before anyone hits it (clear messaging, a slower fallback model, or paid top-ups, plus visible usage meters). For running open models without your own GPU infra, Mango Inference by MangoBoost is worth evaluating, but no provider choice answers the pricing question on its own.
More from Venture
- Pieter Levels' dashboard shows ~$12M/year revenue at 94.8% profit — solo, no hires — sven_ai · 2026-11-13
- Creator earns $1,000/month using AI dubbing to translate videos for YouTubers — FellMentKE · 2026-11-13
- Character-sheet-first workflow: AI TikTok channels earning $1,000+/month — SarahAnnabels · 2026-11-13
- Founder gets lectured on indie hacking by a newbie — he helped build that era — devenbhooshan · 2026-11-13
- Dropbox Was Born From a Forgotten USB Drive — Build for Your Own Pain — sven_ai · 2026-11-13
- Asked an AI who wins post-Jensen: Broadcom over Google, says Tessara — tengyanAI · 2026-11-13