AI founders: has usage-based pricing ever backfired for you?
Hey IH 👋
I’ve been thinking a lot about pricing for AI products, and something feels a bit off.
Most AI SaaS tools default to usage-based pricing (tokens, API calls, etc.).
It makes sense — you pay for what you use.
But in practice, I keep noticing this pattern:
A customer starts using the product more seriously
They don’t actively track usage
End of the month → bill is higher than expected
Next month → usage drops… or they churn
What’s interesting is:
the product didn’t fail them — the pricing experience did
On the flip side, subscriptions solve that anxiety.
But then you get a different problem:
Light users feel like they’re overpaying
Heavy users extract more value than what they pay
So it becomes a trade-off:
Predictability (subscription)
Fairness (usage-based)
I’m starting to feel like neither model fits AI perfectly.
Some teams are experimenting with hybrid models (base fee + usage), which seems like a middle ground—but it also adds complexity.
I tried to break this down more clearly here (just a structured version of these thoughts):
👉 [Token-Based Pricing vs Subscription Pricing: Which Works Better for AI SaaS? ]
Curious how others here are handling this:
Have you seen churn or drop-offs linked to pricing?
Are users sensitive to unpredictable bills?
What model are you currently using?
