Layerpath Studio is seat-based. $X per editor. Simple, predictable, clean.
Now I’m building Path AI — where the AI does the work. And I keep running into the same wall: per-seat doesn’t make sense when the cost isn’t per-person, it’s per-action.
Traditional SaaS: humans do the work. You provide the canvas. Cost per customer is roughly fixed regardless of usage. Seat-based fits.
AI-native: the AI does the work. Every conversation burns tokens. Every action hits an LLM. Customer A might cost $2/month. Customer B might cost $200. Same plan. Completely different cost to serve.
Per-seat assumes delivery cost is fixed. It isn’t anymore.
Salesforce and HubSpot have tried agentic pricing multiple times. Reverted each time. But they’re bolting AI onto existing products. Protecting seat revenue. Different problem than building AI-native from scratch.
For Path AI, the options:
- Credits (buy X, spend on actions)
- Usage (pay per call, per minute)
- Outcome (pay when the AI delivers a result)
None of them are clean. Credits feel like prepaid phone cards. Usage punishes your best customers. Outcome-based is the most honest — you pay when value is created — but defining “outcome” is genuinely hard.
I don’t have this figured out. What I know: if your marginal cost scales with usage, your pricing has to reflect that. Or you eat the variance and pretend it’s fine until it isn’t.