Case Stories

How Circles boosted ARPU by 22% using a multi-agent OpenAI stack

Circles abandoned the single-prompt chatbot for a multi-agent architecture. The result? A 22% increase in average revenue per user and a 65% autonomous resolution rate.

KytoAI & Automation Firm
·
August 5, 2026
0

Key Takeaways

  • 1Stop building one massive AI bot and start building specialized agents.
  • 2Circles used OpenAI to boost ARPU by 22% and dev efficiency by 29%.
  • 3Predictive intent engines can turn support tickets into autonomous upsells.

The breaking point: A 65% drop-off rate when users asked complex billing questions, leading directly to a 4.5% monthly churn.

[@portabletext/react] Unknown block type "tableBlock", specify a component for it in the `components.types` prop

Worth trying this week: Open your Zendesk or Intercom data. Identify the top 3 customer support intents that immediately precede an account upgrade. Build your first specialist AI agent around just one of those intents.

Frequently Asked Questions

What is a multi-agent architecture?

Instead of using one massive prompt to handle all queries, a multi-agent architecture routes user requests to specialized AI models designed for specific tasks, like billing or technical support.

How did AI improve revenue for Circles?

By analyzing user behavior and intent before they asked a question, the AI autonomously offered relevant upsells, directly increasing the average revenue per user by 22%.

AI AgentsOpenAITelecomCase StudyRevenue
Share this article

Kyto

AI & Automation Firm

We design and build AI automations and business operating systems. Agency results + Academy sovereignty.

Related Articles

Ready to automate?

Let's Build Your Operating System.

Book a free discovery call to see how AI automation can transform your operations.

Book Discovery Call