JA EN
shiichan

ARPU Up 22%, Churn Down 9%! Circles' OpenAI-Powered Stack Is Turning Heads

OpenAI News ★ View Markdown
#openai#business#ai

Hey everyone, it's Shiichan! Today I've got a fun case study from OpenAI's News page — a look at how a telco company put OpenAI's tools to serious use.

OpenAI News openai.com

What was announced?

According to OpenAI's News, Circles — a Singapore-based SaaS platform for telco operators — built an AI-native stack around the OpenAI API and Codex, and the results are pretty striking. Founded in 2014, Circles provides its platform to telco operators across 14 countries and six continents, and it also runs its own consumer telecom brand, Circles.Life, in Singapore.

Why it matters

Telco support has traditionally been reactive: customers wait for something to go wrong, then repeat the same information across different support channels. Operators sit on rich data — usage patterns, billing, network activity, location — but turning that into timely, personalized help has been genuinely hard. Circles set out to fix that with an AI-native stack built on the OpenAI API.

What changes

Circles built two main systems.

  • AI Concierge / CareX: a proprietary multi-agent orchestration layer called CareX combines customer history with specialist agents for billing, subscriptions, network management, and account services to deliver proactive, context-aware support. It reached a 65% autonomous resolution rate for support interactions (55% even in the first week of early deployment), and Circles is targeting 95% autonomous resolution once voice support launches
  • Xplore IQ: a personalization engine that replaces broad campaigns with hyper-personalized recommendations. In its Singapore deployment, it drove a 22% ARPU increase (through plan upgrades and add-on subscriptions) and a 9% drop in churn thanks to timely, relevant interventions

Internally, Circles' engineering team also uses Codex for design, coding assistance, and testing, reporting a 29% boost in development efficiency.

Dive Deep

Handing this much responsibility to AI agents naturally raises safety questions, and Circles built governance around it:

  • Detecting and encrypting PII before it reaches the model layer
  • Scoping what each agent is allowed to access
  • Keeping a path for human escalation
  • Phased rollouts, rate limiting, and rollback mechanisms

Awais Malik, Circles' Global Chief Growth Officer, put it this way:

AI should empower users, not force-fit them into outdated journeys.

Wrap-up

  • Circles built an AI-native stack on the OpenAI API and Codex
  • Its CareX multi-agent support system hit a 65% autonomous resolution rate (targeting 95% once voice support ships)
  • The Xplore IQ personalization engine drove a 22% ARPU increase and a 9% churn reduction in Singapore
  • Internal use of Codex boosted engineering efficiency by 29%
  • Governance — PII protection, scoped agent access, and more — was built in from the start

If you work anywhere that's sitting on mountains of customer data but hitting the limits of human-only support, or you're curious about how to design permissions and governance for AI agents, this case study is worth a look.

Related posts