Openreach and Google Cloud bring AI to planning of Britain's fibre network

Openreach has expanded its collaboration with Google Cloud and brought AI tools into the planning of the national full-fibre rollout in the UK. The company already uses Vertex AI for a digital twin of the country's transport and broadband infrastructure, and says Gemini Enterprise cuts engineering overhead by more than half.

How the project works

Openreach is a BT subsidiary and operates the country's largest broadband access network. Scale matters here: the company is responsible not only for communications infrastructure but also runs the second-largest commercial vehicle fleet in the UK. When fibre has to be laid across the whole country, every unnecessary approval, routing error and manual data reconciliation quickly becomes an expensive bottleneck. That is why Openreach is moving part of its planning from spreadsheets and disconnected systems into a single AI environment.

The key element of this setup is a digital twin of Britain's transport and telecom infrastructure, built on Vertex AI. In essence, it is a shared model in which data on roads, crew movements and network condition can be compared and then used to plan construction more precisely. The approach is not meant for attractive visualization, but for getting engineering decisions from hypothesis to field execution faster.

Where the time is saved

The second part of the collaboration is Gemini Enterprise, which Openreach uses to reduce engineering overhead. The company says overhead has already been cut by more than 50%. For a telecom operator of this size, that means not only saved hours but also faster decisions that previously required a long chain of manual steps between the field, engineers and planning. The less administrative friction, the faster the network reaches new addresses.

The source does not disclose the full list of use cases, but it makes clear where AI delivers value in infrastructure programs like this. It is usually not about magical automation of the entire cycle, but about models gathering context faster, highlighting bottlenecks and removing repetitive preparatory work that people used to do by hand. In day-to-day work for design and construction teams, this may look like this:

• Engineering options are prepared faster
• Less manual reconciliation between systems
• Easier coordination of transport and crew dispatch
• Easier prioritization of areas for full-fibre

Why it matters

Moving to full-fibre is not just swapping a cable. It is a long sequence of local decisions: where to dig, how to route, which crew to send and how to avoid extra delays on already congested infrastructure. In a typical setup, such projects run into fragmented data and constant rechecking. So even a modest improvement in planning quickly turns into noticeable savings at national scale. This is where Openreach is trying to gain speed without a proportional increase in headcount.

For Google Cloud, the case is also telling: it is not about an office chatbot, but about embedding AI in the core production loop of a large operator. In this scenario Vertex AI serves as the platform for data and models, and Gemini Enterprise as a layer that takes over part of the routine engineering workload. The more such stories appear in telecom, the more the market will demand measurable ROI, not just promises of automation.

What it means

The Openreach story shows that generative AI at large infrastructure companies is moving from demo mode to the operational core. If the digital twin and enterprise models really do cut overhead by more than half, the next logical step for the market is to use such systems not only to assist employees but as a base layer for network planning, logistics and fieldwork. For Europe this is especially relevant given the high cost of modernizing telecom infrastructure.