Circles, a digital telco operator, has integrated the OpenAI APIs and Codex into its product infrastructure. The published results are significant: +22% ARPU (Average Revenue Per User) and -9% churn. Internal development efficiency has also improved in measurable ways. It is one of the first public case studies where generative AI is applied systematically to retention and personalization in telco.
But the value of this case goes beyond telecommunications. Marketing managers at Italian SMEs and mid-market companies, both B2B and retail, can draw concrete, actionable lessons from it. In particular, the AI-driven personalization logic Circles applies carries over to any context where behavioral segmentation and one-to-one communication are strategic levers. At SHM Studio, we break down the timeline, the mechanisms and the practical implications for the Italian market.
Finally, this case raises an important question for anyone managing a digital marketing budget: is AI applied to retention still a competitive advantage, or is it becoming a baseline requirement? As we'll see, the answer depends heavily on how fast you adopt it.
Circles: timeline of an AI-native integration
Circles is a digital telco operator founded in 2014, with a presence in Asia, the Middle East and Europe. Its business model is asset-light: no proprietary network infrastructure, just a software layer that orchestrates the customer experience. This architecture made AI integration relatively faster than it is for traditional operators.
During 2025, Circles launched a structured partnership with OpenAI. The goal was twofold. First, personalize communications and offers in real time. Second, speed up internal development with Codex, OpenAI's code generation model. The results were published directly on the official OpenAI website.
So this is a documented, verifiable case. These are not projections or internal estimates. They are KPIs measured in real operations.
The numbers behind the case: ARPU, churn and development speed
Three metrics stand out clearly in the case study. First, ARPU grew by 22%. This suggests existing customers are spending more, probably thanks to personalized offers and contextual upselling. In a low-differentiation industry like telco, an increase like this is considered exceptional.
Next, churn dropped by 9%. Retention has historically been the costliest problem for telco operators. According to well-established research by Bain & Company, acquiring a new customer costs on average five to seven times more than keeping an existing one. A 9% reduction therefore has a direct, measurable impact on operating margin.
Finally, development efficiency improved thanks to Codex. Technical teams cut the time needed to implement new features. This is often overlooked in marketing analyses, but it matters: faster development means faster testing and optimization cycles.
Personalization architecture: how the AI layer works
The model Circles adopted follows a logic digital marketers will recognize. The AI analyzes customer behavioral data, such as plan usage, support contact history and browsing patterns, and generates personalized recommendations in real time. The system also adapts the tone and content of communications to each individual profile.
This approach differs from traditional segmentation for a specific reason. Classic segmentation groups customers into static clusters. An AI-native system, by contrast, continuously updates every user's profile. As a result, the offer a customer receives today may differ from the one they'd receive tomorrow, if their behavior has changed.
To dig deeper into the logic of language models applied to marketing, see the Harvard Business Review analysis of generative AI and personalization. Its conceptual framework applies directly to the Circles case.
Winners, losers and who risks falling behind
Who wins in this scenario? Circles, clearly. But also OpenAI, which adds a high-profile case study in telco, an industry historically resistant to rapid innovation. This case also strengthens OpenAI's narrative as an enterprise partner, not just a provider of consumer tools.
Who risks losing? Traditional telco operators that haven't yet started structured AI integration programs. But the risk isn't limited to telecommunications. Any company that competes on retention and personalization, such as insurers, utilities, subscription retailers and B2B SaaS, is in a similar position.
So the real loser isn't a specific company. It's the marketing manager who reads this case as industry news instead of a strategic signal that applies to their own business.
SHM Studio's take: what changes for Italian marketing
At SHM Studio, we closely follow how AI is evolving in marketing and retention. The Circles case confirms a trend we also see in the Italian market: the companies getting concrete results from AI aren't necessarily the biggest, but the ones that have built artificial intelligence into their operating processes, not just their communication tools.
In particular, three elements of the Circles case carry over to Italian SMEs and mid-market companies. First, real-time personalization doesn't require proprietary infrastructure. Second, the impact on retention can be measured in a relatively short time. Third, AI integration speeds up internal processes too, not just customer-facing ones.
So the practical question for an Italian marketing manager is not
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
