Seats, credits and tokens: how AI pricing actually works
By Editorial team ·
Three billing models, three very different ways to get surprised by an invoice.
Per-seat pricing is predictable and punishes partial adoption — you pay for people who log in twice a month. Best when a whole team uses the tool daily.
Credit pricing, common in image and video tools, ties spend to output volume. Watch for credits that expire monthly and for high-quality settings that consume several credits per generation.
Token or usage pricing, used by APIs like ElevenLabs and AssemblyAI, scales cleanly but is hard to forecast before launch. Always model your worst-case month, not your average one.
Whatever the model, set a hard billing cap before you roll anything out to a team.