Metered Pricing for Per-Message AI
Colin Nederkoorn, CEO and founder of Customer.io, on AI's effect on marketing automation
This pushes customer engagement software toward a two part pricing model, flat subscription for AI that helps marketers build faster, and metered pricing for AI that runs inside every outbound message. The reason is simple, assistant features are occasional and raise retention, but per message personalization fires on every send, so costs rise with campaign volume the same way SMS and email infrastructure costs do. Customer.io already prices with customer growth, via profiles and experiments around MAUs, so charging separately for heavy personalization keeps unit economics aligned with message volume.
-
Customer.io draws a hard line between AI that improves product usage, like translation, segment generation, assistant, and MCP access, and AI that personalizes each message from a user activity stream. The first bucket is absorbed as product investment, the second is treated like a new consumable workload.
-
The scale issue is real. Customer.io was sending 18B messages per year when it was at $50M ARR, and later folded Journeys, Data Pipelines, and design tools into one platform so the model has richer context to personalize against. More context improves output quality, but it also increases the chance that personalization becomes a high frequency inference event.
-
A close analog is Intercom, where AI changed monetization from seat based software toward per resolution charges because the AI does work each time a customer conversation is handled. Marketing automation is moving in the same direction for the narrow slice where the model is executing revenue generating work at send time, not just helping staff operate the software.
The next step is a blended model where every platform includes basic AI assistance in the subscription, then layers on variable charges for high volume inference, whether that is per personalized message, per optimization cycle, or per business outcome. That favors vendors with the deepest customer context and the clearest controls, because they can make expensive AI feel predictable enough for brands to trust at scale.