Per-Message LLM Personalization Costs
Colin Nederkoorn, CEO and founder of Customer.io, on AI's effect on marketing automation
This reveals that LLM personalization is not a feature you simply switch on inside a messaging platform, it is a new delivery cost line that can be large enough to force a pricing change. Customer.io already sends messaging at massive scale, 18B messages per year as of 2023, so running an inference on every send would push AI from a lightweight assistive feature into core infrastructure spend. That is why it makes sense to absorb AI in low cost workflow features like translation or segment building, but meter per message generation where usage can spike instantly with a single 10M person campaign.
-
Customer.io has long sold into technical growth teams that wire product events into messaging. A marketer can trigger emails off actions like abandoned carts, failed onboarding, or feature inactivity. LLM personalization adds one more step to that flow, but unlike rules based templating, every send now carries a variable compute bill.
-
Scale is the key reason the cost number matters. Customer.io was at $50M ARR and 18B annual messages in 2023, and later estimates put it at $78M ARR in 2024 and $100M ARR by September 2025. At that scale, even a tiny model cost per message compounds into a material share of revenue if the company eats it wholesale.
-
There is a useful comparison with Intercom. Intercom moved AI into outcome based pricing, charging per resolved conversation instead of only per seat, because LLM usage changes the unit economics of the product itself. Customer.io is reaching the same conclusion in marketing automation, where AI that drafts or rewrites each outbound message behaves more like metered infrastructure than bundled SaaS.
The next step is a split market. AI that helps teams build faster in the background will get bundled into the base platform, while AI that runs on every send will become a usage priced layer. As model costs keep falling, that layer can expand from premium personalization for big campaigns into a standard part of how marketing automation products are sold.