OpenAI-Powered Telco Concierge Lifts ARPU 22% at Circles

The Numbers Behind the Personalization Push

Circles, the Singapore-founded telco SaaS company, says AI-driven personalization built on OpenAI's API lifted average revenue per user by 22% and reduced churn by 9% in its Singapore market, according to a case study OpenAI published August 3, 2026.

Circles built the result around its AI Concierge, a conversational interface that pulls usage, billing, network, and location signals into one experience. The personalization engine behind it, called Xplore IQ, compared customers who received AI-generated recommendations against those who didn't, and the group getting personalized nudges toward plan upgrades and add-ons drove the ARPU gain.

On the support side, Circles' multi-agent architecture, CareX, now resolves 65% of customer service interactions without a human agent, up from 55% in an early one-week deployment at one operator.

Why This Matters for Marketers

Most vendor claims about AI personalization arrive without a control group. Circles' 22% ARPU figure is notable because it was measured against customers who didn't get the recommendations, giving CMOs in telecom and adjacent subscription businesses an actual comparison point rather than a marketing promise. If you're weighing where AI agents fit in your own stack, it's worth reviewing what these systems can realistically do versus what's still capability versus hype.

What's Next, and What's Still Unproven

Circles says it's expanding CareX into real-time voice support and targeting 95% autonomous resolution across text and voice combined, up from 65% today. That's a company target, not a delivered result, and it's worth watching whether voice resolution holds up at the same rate as chat, especially for billing disputes that need escalation. OpenAI's broader push into support agents, seen in its enterprise agent platform for customer workflows, suggests Circles' architecture is part of a wider pattern rather than a one-off.

Circles also credits Codex with a 29% increase in development efficiency across design, coding, and unit testing, though the company hasn't detailed how that figure was measured. Circles works with telco operators in 14 countries, so the Singapore numbers are a single-market data point, not a global average, and marketers should treat the 95% resolution target as a stated goal until Circles reports actual voice performance.