Cerillion 26.2 adds agentic AI to telecom billing: customer care, package design and churn prediction

Cerillion announced the general availability of Cerillion 26.2 on November 4, 2026, the latest version of its pre-integrated BSS/OSS Suite for communications service providers (CSPs). The release expands the suite’s agentic AI across four areas — customer care, service configuration, product design and churn prediction — rather than adding a single assistant on the side.

The distinction matters because BSS/OSS software runs billing, charging and customer management: the systems a telecom operator cannot afford to get wrong. Cerillion’s pitch is that the new agents work inside the suite’s existing catalogue, promotions and configuration workflows, under central governance, instead of reaching into those systems from outside. The announcement, issued from London via PR Newswire, gives no pricing and no accuracy figures for the churn models — only that they proved “highly effective” in customer pilots — so the release should be read as a product launch, not as measured results.

Four new capabilities, in plain terms

Customer Care Agent. This uses agentic AI to help customer service teams find the relevant customer and service information, resolve enquiries and determine the most appropriate next action. Cerillion says that makes service faster and more consistent, and shortens the training needed to onboard new users.

Service Configuration Agent. This significantly expands Cerillion’s agentic automation to cover a wider range of complex service and system configuration activities. By reducing manual effort and reliance on specialist knowledge, Cerillion says it accelerates the introduction of new services and network integrations.

Competitor-aware Package Creation. This uses details of competing offers to inform agentic package design within Enterprise Product Catalogue. Product teams can develop differentiated propositions and create new packages more quickly in response to market changes, Cerillion says.

Predictive Churn. This uses advanced AI and machine-learning models to identify customers at greatest risk of leaving, giving operators earlier insight into retention risks. The resulting churn scores support targeted campaigns and offers through Cerillion’s Promotions Engine. Cerillion says the models proved highly effective in customer pilots, but published no scores, accuracy figures or pilot details to check that claim against.

The pattern will be familiar to readers following agentic software in other operational roles: Beyond Now applied native agents to telecom revenue operations in October, and PagerDuty’s Paige agent triages incidents alongside human engineers. Cerillion’s version differs in where it sits — inside the billing and catalogue systems themselves, where a wrong action has a direct revenue consequence.

The governance layer underneath

The agents run on three components. The AI Management Centre provides centralised model governance; the MCP Server enables secure and controlled access to Cerillion’s Open APIs using its Progressive Discovery architecture; and the Agent2Agent Server supports coordinated, multi-step workflows across different agents and systems. Operators can adopt Cerillion’s AI capabilities, integrate their existing in-house or third-party AI investments, or combine both, the company says. The MCP pattern is now standard plumbing for agent access — our hands-on guide to the Shim harness shows how MCP servers expose tools to an agent — but Cerillion’s point is control: central governance over models, controlled API access, and coordination across agents, rather than letting each agent improvise its own route into billing data.

Catalogue, charging and open standards

Enterprise Product Catalogue also gains the Dynamic Catalogue Workspace, a flexible, grid-based environment in which product teams can compare and update complex product, package and pricing structures at scale, with changes applied directly within the grid. The Convergent Charging System adds a native Nchf interface for 5G standalone charging, so REST-based and Diameter charging interfaces can coexist in the same deployment — a flexible migration path to 5G standalone while hybrid networks are still supported.

Alongside the product changes, Enterprise Product Catalogue has achieved TM Forum ODA Component Certification for TMFC001 Product Catalog Management. Cerillion has added 15 further TM Forum Real World Open API certifications since Cerillion 26.1, taking its total to 47 across seven customers — more than any other independent BSS/OSS vendor, the company says.

What to make of it

“Cerillion 26.2 demonstrates how advanced AI can deliver practical value across the BSS/OSS, from improving customer care and retention to accelerating product and service innovation,” said Brian Coombs, Cerillion’s chief product officer. Cerillion has a 27-year track record in billing, charging and customer relationship management software, with about 70 customer installations across about 45 countries, according to the announcement. Version 26.1, announced on April 29, 2026, introduced communication and coordination between agents, building on the AI agents and MCP Server from the previous release; 26.2 extends that foundation into customer-facing and retention work, rather than starting from scratch.

What is still unproven is the part operators will care about most: whether churn scores built in pilots hold up across live customer bases, and whether agents working inside billing and configuration systems stay within their guardrails at scale. Cerillion has published the architecture and the certifications; the evidence from live deployments is still to come. For now, 26.2 is generally available — the next test is not another announcement, but the first operators running it on real subscribers.