Circles squeezed 22% more revenue per user after wiring OpenAI into its telco stack

Here’s a number that makes telco executives pay attention: a 22% ARPU lift in Singapore. That’s what Circles says it achieved after embedding OpenAI’s API into the guts of its operations—not as a surface-level chatbot, but as connective tissue linking support, billing, and marketing in real time.

The company, which runs both a consumer telco in Singapore and a SaaS platform used by operators across 14 countries, built what it calls an AI Concierge. It is a single conversational surface where customers manage accounts, get support, and receive recommendations. Underneath sits CareX, a multi-agent system that routed 65% of customer service interactions to autonomous resolution. One operator hit 55% autonomous resolution within the first week of deployment.

Personalization is where the money shows up. Circles’ Xplore IQ engine pulls customer behavior, account history, and real-time signals to decide what to offer next. When it compared customers who received AI-generated recommendations against a control group, the AI cohort delivered that 22% ARPU increase through higher plan upgrades and add-on subscriptions. Churn dropped 9%. Those are not marginal gains in a commodity business where loyalty is notoriously fragile.

Circle is now pushing CareX into real-time voice support, targeting 95% autonomous resolution across text and voice. That is ambitious—voice introduces latency requirements and failure modes that text does not. But the underlying thesis is worth watching: telcos sit on mountains of usage, billing, and network data that most never operationalize. Circles is betting that stitching it together with an orchestration layer that understands context, not just keywords, changes the economics of running a carrier. The company also claims a 29% boost in internal development efficiency using Codex for design, coding, and testing—a reminder that the same API stack is reshaping how the product itself gets built.