AI Monetization Is a Metering Problem
Augusto Marietti, CEO of Kong, on the end of tokenmaxxing
The key shift is that AI turns monetization into a real time infrastructure job, not just an invoicing job. In AI products, value is created and cost is incurred inside each request, where tokens, latency, model choice, and even power use can change second by second. A company sitting in the traffic path can measure those units at the moment they happen, then use that data for pricing, internal chargebacks, cost controls, and billing later.
-
Kong already sits in front of API and LLM traffic, so it can observe each call before it reaches OpenAI, Anthropic, or an internal model. That makes metering concrete. Count prompt and completion tokens, measure response time, apply routing rules, and attach a price or budget policy at the gateway instead of reconstructing usage after the fact.
-
This is why metering vendors matter more in AI than classic SaaS billing vendors. Metronome and Orb were built around usage records and flexible pricing, while Kong bought OpenMeter to add an engineering first metering layer directly into its gateway stack. OpenMeter now powers Kong Metering & Billing, and Kong charges a 0.4% take rate on billing volume processed through it.
-
The practical buyer is often not finance, it is the platform or infrastructure team. They need to answer simple operational questions in real time, which business unit used which model, which prompts should be routed to a cheaper model, and whether an agent workflow is profitable after token and latency costs. Billing systems consume that data, but they do not create it.
The next layer of AI infrastructure will bundle gateway, policy, metering, and payments into one control plane. As agent workflows create many more model calls per task, the winner will be the platform that can measure usage precisely at request level, turn that into pricing and margin controls, and make monetization automatic for every API and AI call moving through the stack.