Most telcos have spent the last three years adding AI to their BSS stack the same way you add a car sticker: it looks like something changed, but underneath everything is the same. A chatbot in front of the IVR. A churn prediction model that outputs a score into a dashboard nobody checks. An anomaly detection layer that fires alerts into a ticket queue already full of alerts. Point solutions, layered on top of systems that were never designed to be operated by machines.
This approach has a ceiling. It is becoming visible now. The next architectural shift puts autonomous agent crews at the centre, crews that own BSS processes end-to-end, from insight to action, across system boundaries, writes Martin Rueckert, the chief AI officer at Tallence.
The typical BSS environment at a midsize European telco looks roughly like this: an order management system that dates back a decade, a billing platform that has been customised into something unrecognisable, a CRM that talks to both via a middleware layer that one person on the team truly understands, and somewhere between twelve and thirty integration touchpoints that were each added to solve a specific problem at a specific moment in time.
Why point AI hits a wall
When you drop a copilot or a narrow ML model into this environment, it can optimise within its own slice. The churn model fires. The recommendation engine suggests a retention offer. But getting from that insight to an actual action, such as updating the CRM, triggering a workflow in the order system or logging the outcome for the next model to learn from, requires a human to carry the signal across system boundaries. Or it requires yet another integration project.
Martin Rueckert