Openreach brings Google Cloud AI into fleet and fibre planning

Openreach, a subsidiary of BT Group, is leveraging AI and Google Cloud to optimize its fleet operations and support fiber expansion. The initiative involves analyzing data from 24,000 vans using BigQuery analytics platform to reduce emissions and improve efficiency, while also building a digital twin of UK transport networks to speed up fiber deployment.

• Openreach expanded its Google Cloud collaboration on 25 March 2026, applying data and AI tools to a fleet then described as 24,000 vans and to full-fibre planning.
• BigQuery supports fleet geoanalytics, Vertex AI underpins a transport-corridor digital twin, and Gemini Enterprise is being used in data-engineering work.
• The emissions, savings and productivity figures are company-reported; public coverage does not disclose an audited baseline or calculation method.

Operational AI, not an autonomous network

Openreach and Google Cloud announced an expanded collaboration on 25 March 2026. The work applies cloud analytics and AI to two operational problems: running a large engineering fleet and deciding where full-fibre infrastructure may be extended. This is not a claim that AI drives the vans or makes construction decisions on its own.

At the announcement, Openreach described a fleet of more than 24,000 vans travelling almost 200 million miles a year. It had moved fleet telematics data to Google Cloud. BigQuery geoanalytics is being used to compare real-world vehicle use, routes and charging availability, helping managers identify where an electric van could replace a diesel vehicle. The same data can expose idling and excessive travel, flag potential faults and reduce the time vehicles are unavailable.

ITPro reported Openreach’s claim that the cloud-based analysis had helped it deploy electric vehicles faster and that the additional EVs avoided about 10,000 tonnes of CO2e a year. The announcement also attributed millions of pounds in annual savings to the programme. Those numbers should be read as company-reported outcomes: neither the announcement nor the independent reports publish the baseline fleet mix, attribution method or audited calculation behind them.

The EV programme is broader than the Google Cloud project. Electrive reported in June 2025 that BT Group had begun receiving an order of more than 3,500 electric vans, intended mainly for Openreach, and expected the group fleet to approach 8,000 EVs when delivery was complete. That was a delivery forecast, not evidence that every ordered vehicle was already in service when the AI collaboration was announced.

A digital twin for planning

Openreach also says it built a digital replica of the United Kingdom’s transport corridors with Vertex AI. The model combines information on 35 million homes and businesses with road, rail and waterway networks and existing broadband assets. Its stated purpose is to help planners see where full fibre might be extended sooner and which premises may be eligible.

The system is a planning aid, not proof of a faster nationwide build. Public material does not give a comparison against the previous planning process, the number of premises brought forward, a reduction in build cost or a measured change in completion time. Telco Magazine’s account confirms the platform and data scope while attributing the claimed operational benefits to Openreach.

A third strand concerns data engineering. Openreach says Gemini Enterprise converts complex legacy queries into production-ready BigQuery code and has cut time-to-insight by more than 50 per cent. The percentage is again self-reported; no workload definition, error rate or external benchmark has been published.

Why the control boundary matters

Openreach controls routing, vehicle replacement, maintenance and fibre-build decisions. Google Cloud supplies the data warehouse, AI platform and agent tools used to inform those decisions. The distinction matters because better analytics can change priorities and reduce waste, but physical deployment still depends on engineers, street access, charging infrastructure, vehicle supply and the quality of network and location data.

The collaboration is worth tracking because it joins fleet operations, capital planning and data engineering in one cloud environment. If the reported gains hold, it offers a practical example of AI affecting a telecom operator’s cost base and emissions rather than only its customer-facing services. It also increases Openreach’s operational dependence on the availability, governance and evolution of Google Cloud services.

What remains unproven

The public record establishes the tools, data domains and intended operating uses. It does not independently verify the annual savings, the CO2e attribution, the 50 per cent productivity figure or the amount of fibre deployment accelerated by the digital twin. It also does not disclose the model-governance arrangements, data-retention controls or the extent to which planners can challenge automated recommendations.

What to watch

Useful evidence would include the emissions methodology, electric-vehicle deployment by region, reductions in idling and unavailable-vehicle time, measured build-time changes, and a before-and-after assessment of query conversion quality. Until those appear, the programme should be described as a deployed decision-support system with promising company-reported results, not as independently proven automation of the network rollout.