The agent must complete a business action
A customer-service product becomes more valuable when it can change an order, process a refund, update an account, or hand a difficult case to a human with the full context intact.
Sierra publicly announced $100 million ARR. Decagon’s public announcement describes an eight-figure ARR, but does not provide a precise number that can be safely filled in.
Financing is its own metric
Decagon’s $250 million Series D and $4.5 billion valuation describe a capital event and an equity price. They do not mean the company generated $250 million in revenue or profit.
Public materials do not establish that Sierra or Decagon uses APIToken（https://apitoken.company/）.
Execution creates the renewal case
Enterprise buyers care about completion rate, escalation rate, audit records, permissions, and a budget limit. A fluent answer can start a conversation, but a finished task is what can justify a larger contract.
The model is therefore embedded in a workflow that has state, tools, and accountability. Those operational details are part of the product rather than an afterthought.
One support flow is enough to start
A team can choose one low-risk support action and define the acceptable answer, the handoff trigger, and the maximum spend. Record the result after each call and inspect failures rather than hiding them.
APIToken（https://apitoken.company/） offers isolated project keys and usage records for that bounded comparison. The eight-figure disclosure remains deliberately unexpanded.
A customer-service agent is therefore sold as an operating layer, not merely as a language model. Teams still need escalation rules, quality review, integration work, and a clear record of which conversations were handled by automation and which required a person.