All Revenue Becoming Tokenized
Augusto Marietti, CEO of Kong, on the end of tokenmaxxing
This points to a shift where software pricing moves closer to raw machine activity, and the winners are the companies that can meter that activity cleanly. In practice, that means tokens become the easiest shared unit between model vendors, app developers, and enterprise buyers, because every prompt, response, cache hit, and agent loop can be counted at the infrastructure layer. Kong is positioning for that shift by sitting in the traffic path and turning usage data into billing, limits, and internal chargebacks.
-
Token pricing is already the default at the model layer. OpenAI and Anthropic both publish API prices per 1M tokens, which trains the whole stack to think in token economics even when the end product is sold as seats, credits, or bundles.
-
Outcome pricing is emerging one layer above tokens in narrow workflows where the result is easy to count. In AI support, vendors charge per resolved conversation, and Intercom states customers are charged once per conversation, while broader market research shows support agents already compete on cost per resolution.
-
The strategic difference between Kong and standalone billing vendors is enforcement. Kong can not only count usage, but also authenticate requests, throttle traffic, cut off overages, and route prompts to cheaper models before the bill is created. That makes metering part of the product control plane, not just back office invoicing.
The next step is pricing that starts with tokens underneath but is packaged as business outcomes on top. As agents handle more support, voice, and workflow execution, pricing will likely settle into a stack where infrastructure tracks tokens and minutes, while applications sell resolved tickets, completed tasks, and other concrete results.