Closing a telecoms exchange is not a simple act of switching off equipment. It is a multi-year programme involving customer migrations, physical infrastructure build, regulatory compliance, coordination across multiple communication providers, and the careful sequencing of hundreds of interdependent tasks. Do it in the wrong order and customers lose service. Move too slowly and the costs of maintaining a site nobody needs continue to compound.

For most network operators, exchange consolidation programmes have historically been managed through a combination of spreadsheets, manually maintained records, and the institutional knowledge of experienced engineers. That approach is reaching its limits. The scale, complexity, and commercial urgency of current decommissioning programmes demand something more capable. Artificial intelligence is increasingly providing it.

1. The Scale of What is Now Underway

Openreach launched its Exchange Exit Programme formally in April 2025, targeting the closure of approximately 4,600 legacy exchanges. Of its current estate of around 5,600 exchanges, only roughly 1,000 are needed to support modern fibre broadband services. The remainder are economic liabilities: sites consuming energy, requiring maintenance, and serving a customer base that has either already migrated or needs to be moved before closure can proceed.

The first pilot closure, Deddington in Oxfordshire, was completed in November 2025. It was also the first UK location to see the full retirement of the copper-based PSTN. A further 12 exchanges began the exit process in April 2026, with a target closure date of September 2028 for that cohort. The ambition is to close 105 priority exchanges by 2030, with the remaining sites following through the early 2030s.

Each individual exchange takes between 4 and 7 years to move through the full exit process, depending on complexity. The first three years focus on stop-sells and voluntary migration. The final year involves supported migrations and, where necessary, enforced cessation of services.

The commercial stakes are significant. A 2025 survey by Neos found that re-routing networks due to exchange closures will cost impacted businesses an average of £1.4 million each. For communication providers operating across multiple affected exchanges, the cumulative figure is substantially higher. Meanwhile, operators themselves face rising costs if closures are delayed: energy, maintenance, regulatory overhead, and the staff time involved in managing customer migrations all continue to accrue until the site is physically closed.

This is not a UK-specific challenge. Across Europe, the US, and Asia-Pacific, telecoms operators are managing analogous programmes at scale. The underlying economics are consistent: the full fibre transition and 5G rollout have concentrated the network around a smaller number of modern handover points, making the legacy exchange estate increasingly redundant and increasingly expensive to retain.

2. Why Decommissioning is Harder Than it Looks

Dependency Mapping at Scale

Before a single customer line can be migrated, the operator must understand precisely what is connected to the exchange, what product each connection is on, which communication provider is responsible for it, and what the customer's circumstances are. In practice, this data is rarely held in one system. Asset registers, service records, and customer information frequently sit across multiple legacy databases, with gaps, inconsistencies, and records that have not been updated in years.

One of the most commonly cited causes of exchange exit delays is the discovery, during the migration phase, of services that were not captured in initial planning data. Unrecorded lines, orphaned assets, and customers on products with no straightforward modern equivalent all create exceptions that require individual handling.

Regulatory Obligations and Vulnerable Customers

The UK Government's Fixed Telecoms Modernisation Charter sets out specific obligations around the treatment of vulnerable customers during the exchange exit process, particularly those who rely on telecare systems, medical alarms, or other services that run over copper lines. These customers cannot simply be migrated in bulk. Each case requires individual assessment, and the risk of a service interruption for a dependent customer carries serious regulatory and reputational consequences.

This obligation adds a layer of complexity that cannot be resolved through automation alone. But AI can significantly improve the identification, tracking, and case management of at-risk customers, ensuring that human intervention is targeted where it is most needed rather than spread across the full customer population.

Physical Infrastructure Sequencing

The closure of an exchange is not just a network event. It involves the physical removal of equipment, the decommissioning of the building, and decisions about what happens to the site itself. Most of Openreach's exchange estate is owned by TT Group (formerly Telereal Trillium), which acquired the majority of BT's real estate portfolio in 2001. Many closed exchanges are subsequently converted to residential or commercial use. Managing the handoff between network cessation and physical site disposal adds another workstream to an already complex programme.

Sequencing matters enormously. If customer migrations are delayed, physical decommissioning cannot proceed. If physical decommissioning is scheduled before adjacent sites have been upgraded to handle the redirected traffic, network integrity is at risk. Getting this sequencing right, at the scale of thousands of sites, requires a level of planning sophistication that manual processes simply cannot deliver.

3. Where AI Changes the Equation

Intelligent Asset Discovery and Data Reconciliation

The first task AI can meaningfully accelerate is the one that delays most programmes before they start: understanding what is actually in the exchange. AI-powered data reconciliation tools can cross-reference multiple source systems, identify conflicts and gaps, and produce a unified asset view that human analysts would take weeks to produce manually.

This is not a trivial capability. Telecoms operators routinely hold asset data across OSS platforms, field engineering systems, billing records, and service databases that were never designed to be queried together. Natural language processing and machine learning classification models can identify which records relate to the same physical asset, flag anomalies, and surface edge cases for human review. The result is a more reliable planning baseline and a reduction in the mid-programme surprises that cause delays and cost overruns.

Predictive Migration Planning

Once the asset picture is clear, AI can be applied to migration sequencing: working out the most efficient order in which to move customers, taking into account product type, geography, communication provider relationships, and the capacity of the receiving exchange. This is a complex optimisation problem that traditional planning tools handle poorly.

Predictive models can also identify which customers are most likely to require assisted migration and prioritise outreach accordingly. Operators that have applied machine learning to customer segmentation in exchange exit programmes have consistently found that a small proportion of customers account for a disproportionate share of migration complexity. Identifying that cohort early and resourcing it appropriately is one of the highest-leverage interventions available.

McKinsey research found that telecom operators without automated operations workflows spend up to 30% more on network management than peers who have implemented intelligent automation. In the context of a decommissioning programme spanning thousands of sites over a decade, that differential compounds significantly.

Real-Time Programme Monitoring and Risk Flagging

Exchange exit programmes run over years. The challenge of maintaining an accurate picture of programme status, across hundreds of concurrent migration workstreams, is significant. Manual reporting creates lag: by the time a delay is visible in a programme dashboard, it has often already had downstream effects on the broader schedule.

AI-powered programme monitoring tools can ingest data from field systems, service records, and communications with providers in near real-time, flagging deviations from plan as they emerge rather than after the fact. This capability shifts programme management from reactive to predictive: teams can intervene at the point where a delay is forming rather than after it has materialised.

Energy and Operational Cost Modelling

Each exchange that remains open carries an ongoing cost: energy, maintenance contracts, security, staff access, and the regulatory overhead of managing services on a site that is in transition. AI can model these costs at a site-by-site level, enabling operators to prioritise closures based on the financial return from decommissioning rather than simply on geographic or operational convenience.

This matters more than it might appear. Global telecoms CAPEX is projected to decline by 1.5% in 2026 to $320 billion, with operators under sustained pressure to reduce operational expenditure. In this environment, every month a legacy exchange remains open is a month of avoidable cost. AI-driven cost modelling gives programme sponsors the commercial visibility needed to maintain pace and justify investment in the migration resources required.

Repurposing Decisions: Close or Convert

Not every decommissioned exchange should simply be handed back to its landlord. Many operators are beginning to assess whether legacy central office space can be repurposed as edge computing or AI inferencing infrastructure, given the combination of existing power capacity, physical security, and strategic location that exchange buildings often provide. In an environment where new data centres take years to permit and build, the repurposing of existing exchange buildings is attracting serious commercial interest.

AI-assisted site assessment tools can evaluate which exchange buildings have the highest repurposing potential based on power availability, structural condition, geographic positioning relative to population centres, and proximity to fibre routes. This turns what has traditionally been a purely cost-driven decision into a genuine strategic opportunity.

4. The Data Foundation That Makes It Possible

The AI capabilities described above depend entirely on the quality and accessibility of the underlying data. Exchange exit programmes that have struggled with AI adoption have consistently identified the same root cause: data that is too fragmented, too inconsistent, or too poorly governed to support reliable modelling.

Building the data foundation for an AI-enabled decommissioning programme is therefore not a technical afterthought. It is a strategic prerequisite. This means investing in data integration, quality, and lineage before AI tools are deployed, and treating the data estate as a programme deliverable in its own right.

The operators making the most progress are those that have established governed data domains for their exchange exit programmes, with clearly defined ownership of customer, asset, service, and physical infrastructure records. These operators are also instrumenting their migration workflows to generate the real-time event data that AI monitoring tools need to function, rather than relying on periodic manual reporting.

Telstra generated over $130 million from copper asset sales over two years. BT secured more than $140 million in advance payments for forward copper asset sales. The commercial case for accelerating decommissioning is clear. The data foundation to do it intelligently is where investment attention now needs to follow.

5. The Delivery Model Question

Exchange consolidation programmes of this scale are not run by a single internal team. They involve field engineering partners, communication providers, local authorities, real estate managers, and regulatory bodies. Coordinating this ecosystem, and applying AI tools consistently across it, requires a delivery model that goes beyond conventional programme management.

The most effective approach treats AI not as a technology bolt-on to an existing programme structure, but as an embedded capability within the programme itself. This means specialist data and AI resources working alongside delivery teams, with shared access to programme data and clear accountability for the insights AI tools generate. It also means building the governance structures that allow AI-flagged risks and anomalies to be acted on quickly, without creating new bottlenecks in the decision-making chain.

Many operators are finding that the combination of internal programme resource with external specialist capability, particularly for AI integration and data engineering, delivers better outcomes than either approach alone. Internal teams bring the operational and regulatory knowledge that cannot easily be acquired from outside. External specialists bring the technical depth and prior programme experience that internal teams cannot build at pace.

Conclusion: Decommissioning as a Strategic Programme

Exchange consolidation and legacy infrastructure rationalisation have historically been treated as maintenance activities: necessary, costly, and unglamorous. The scale of what is now required changes that framing entirely. Closing 4,600 exchanges over a decade, while maintaining service continuity and meeting regulatory obligations, is a strategic programme of the first order.

AI does not make this programme simple. The underlying complexity, the customer migration challenge, the physical infrastructure dependencies, and the multi-party coordination required will remain demanding regardless of what technology is applied. What AI does is make the programme more predictable, more data-driven, and more capable of operating at the speed and scale required.

Operators that treat exchange consolidation as a purely operational challenge, managed through spreadsheets and manual reporting, will find themselves consistently behind schedule and above budget. Those that invest in the data foundations and AI capabilities needed to run these programmes intelligently will deliver faster, cheaper, and with fewer of the service disruptions that damage customer trust and attract regulatory scrutiny.

The first exchanges have been closed. Several thousand more remain. The difference between operators that get this right and those that do not will be visible in their cost base and their programme delivery record for the rest of this decade.

About VE3 Global

VE3 is a UK-headquartered technology and enterprise AI consultancy with offices in London and Pune. We support infrastructure operators and large enterprises with data strategy, AI integration, legacy modernisation, and programme delivery. Our experience spans telecoms, utilities, and public sector transformation programmes.

To discuss how VE3 can support your exchange consolidation or infrastructure rationalisation programme, visit ve3.global