Bouygues Telecom & Deepomatic : Automating Fiber Connection Quality Control

To make the actual subscriber connections, Bouygues uses the services of a large number of contractors who perform several thousand work orders daily.

With the cadence accelerating, it was difficult to ensure consistent quality. It was often necessary to organize repeat visits to connect with a customer. Furthermore, inspecting all work orders was impossible, and only 5% to 10% of installations were audited. Likewise, the surge in demand has been piling pressure on technicians who, due to lack of time to manage the most difficult situations, often failed to follow procedure. As a result, the network integrity was deteriorating due to non-compliant installations. Facing fierce competition, Bouygues Telecom wanted to differentiate itself by offering a superior quality of service and quality of experience.

Using a standard smartphone, technicians in the field take pictures of the installation where the physical connection is made between the subscriber and the access network. Deepomatic uses image recognition neural networks to analyze the images in real-time and report anomalies such as unconnected fibers, wrong color coding, or non-compliant cable paths. With analysis and recommended actions available instantly, technicians can fix defects immediately instead of planning to rework them later. In case of doubt, the application raises an alert so a technician can make the final decision. This makes concentrating their skills only on the most complex cases possible.

Deepomatic successfully identifies over 98% of defects and is improving over time. As the rollout progresses and the nature of equipment, processes, and defects it encounters changes, Deepomatic’s algorithms adapt. The project is currently centered on the work performed on street cabinets and equipment at the edge of the access network. It is being used for over 250,000 work orders per month. However, plans are underway to extend its application area to other work orders.