Fixing Fibre-to-Flats: The Policy Shift AI Needs

89% of UK premises can now access gigabit-capable broadband, but flats are still dragging the average down. The gap isn’t small either: May 2025 data puts gigabit availability at 86.1% across residential premises, versus 79.6% for flats—and an estimated 1.2 million flats in England and Wales still can’t get a gigabit connection.

That shortfall matters for consumers, but it matters even more for telecom operators trying to modernise networks with AI. Here’s why: AI-driven network planning, rollout optimisation, and fault prediction only work well when the physical rollout environment isn’t fighting you. When a fibre build gets stalled by unclear ownership, non-responsive freeholders, or outright obstruction in multi-dwelling units (MDUs), AI can’t “optimise” its way around missing ducts, blocked risers, or unapproved wayleaves.

The UK government’s December 2025 consultation is a direct attempt to remove one of the most persistent “non-technical” barriers to fibre-to-the-building and fibre-to-the-flat deployment: freeholder friction. It’s a governance move with real technical consequences—especially for telcos betting on AI to manage exploding broadband demand and 5G densification.

The fibre-to-flats problem is a people-and-process bottleneck

The core issue isn’t that fibre is hard to install. It’s that access and permission are hard to secure—at scale—inside MDUs.

If you’ve worked on MDU rollout programs, you know the pattern:

• The operator needs access rights to shared building areas.
• The “right” person to approve access is unclear, changes, or is hidden behind managing agents.
• Even when identified, the freeholder has weak incentives to respond quickly (or at all).

The UK’s existing framework aims to solve this through the Electronic Communications Code—designed to encourage consensual agreements between operators and landowners. And in 2021, the Telecommunications Infrastructure (Leasehold Property) Act (TILPA) added a fast-track court route when landowners repeatedly fail to respond.

But real rollout friction often sits in the messy middle: not quite refusal, not quite silence, not quite resolvable without months of chasing. That’s where the consultation’s proposals target.

Why MDUs are where gigabit goals go to stall

MDUs are operationally different from single-dwelling homes:

• More stakeholders (freeholder, managing agent, residents’ management company, leaseholders, tenants)
• More shared infrastructure constraints (risers, comms cupboards, fire-stopping requirements)
• Higher reputational risk (a botched install affects dozens of residents)
• Higher coordination costs (surveys, appointments, access windows)

When the last-mile is inside a block of flats, the “last 20 metres” often costs more time than the first 20 kilometres.

What the UK is proposing—and why it’s a big deal

The consultation proposes a new lease-based mechanism that gives leaseholders more power to trigger fibre deployment.

The core policy proposal is:

Imply a new right into an existing lease so a residential leaseholder can request a gigabit-capable connection from the relevant freeholder.

Impose a duty on the freeholder not to unreasonably refuse the request.

This is framed as a complementary “nudge measure” that sits alongside the Electronic Communications Code. In plain terms: it’s meant to force engagement, so an operator can get to the negotiation stage without months of deadlock.

Who it applies to (and what counts as “gigabit”)

The proposal would apply to:

• Any residential leaseholder
• In a building with two or more dwellings
• Seeking a connection capable of at least 1,000 Mbps

The consultation runs until February 2026, which makes this a live topic for operator strategy planning in 2026 build seasons.

Why this is directly relevant to AI in telecommunications

Most “AI in telecom” roadmaps assume the network is buildable on schedule. That assumption is often wrong in MDUs.

AI is already used across fibre and 5G programs for things like:

• Network rollout planning (design automation, clustering, route optimisation)
• Workforce and field scheduling (dispatch optimisation, appointment prediction)
• Asset risk scoring (which cabinets, joints, or routes are likely to fail)
• Demand forecasting (where upgrades will be needed first)

But MDUs introduce a category of constraint that AI can’t solve without better inputs: legal access and building permission state.

Here’s a blunt, useful sentence you can reuse internally:

AI can optimise labour, materials, and routing—but it can’t negotiate with a silent freeholder.

The hidden cost: AI models get distorted by “permission noise”

If your rollout data lumps “engineering complexity” and “permission delays” into the same delay bucket, your models learn the wrong lessons.

Example:

• Building A is delayed 12 weeks due to freeholder non-response.
• Building B is delayed 12 weeks due to asbestos discovery and rework.

If your data doesn’t clearly separate these, your predictive models may “conclude” that certain building types are technically difficult when the true root cause is governance.

The proposed leaseholder right helps because it reduces the likelihood of long, ambiguous stalls—meaning:

• Cleaner operational datasets
• Better schedule prediction
• More reliable supplier commitments
• More accurate capex phasing

That’s not a small benefit. It changes how confidently an operator can run AI-assisted rollout at national scale.

Regulatory reform as an enabler of AI-driven network optimisation

Policy is infrastructure. Not in a philosophical way—literally.

When regulation reduces transaction costs and delays, it unlocks compounding benefits:

• Faster premises passed
• Faster take-up revenue
• Lower “cost to connect” per flat
• More stable deployment pipelines (which matters for contractors and supply chain)

And crucially for this series on AI in Telecommunications:

AI thrives on stable, repeatable processes. Regulation that reduces MDU variance makes AI more valuable.

What operators should do now (while the consultation is open)

If you’re responsible for network strategy, rollout operations, or digital transformation, the consultation window is not just a policy event—it’s a planning opportunity.

Practical moves that pay off regardless of the final legislative details:

1. Tag permission states as first-class data

   • Create explicit categories: “freeholder unknown”, “agent unresponsive”, “refused”, “approved”, “in negotiation”, “court route”, etc.
   • Don’t bury these in notes fields. AI can’t learn from data it can’t see.

2. Build an MDU-specific risk model

   • Treat MDUs as a different product, not a harder version of SDUs.
   • Include variables like ownership complexity, managing agent responsiveness history, and building access constraints.

3. Operationalise “fast engagement” playbooks

   • Standardise the first 10 days of contact attempts.
   • Use templated communications, clear proof packs, and documented escalation.
   • Make it easy for leaseholders to support the request if the law shifts in their favour.

4. Design your AI scheduling around uncertainty bands

   • Assume some MDUs won’t be buildable until permission is secured.
   • Use probabilistic schedules that don’t over-commit crews and contractors.

What leaseholders, freeholders, and ISPs need to align on

The winning outcome is boring: consistent engagement, predictable installs, and fewer legal escalations.

For leaseholders

If this proposal becomes law, leaseholders gain a clearer route to request gigabit access. The most important behavioural change will be coordination:

• One resident asking helps.
• A residents’ group asking with documented demand helps more.

For freeholders and managing agents

Freeholder resistance often stems from risk concerns (disruption, liability, building integrity) or simple admin inertia. The policy direction signals that “ignore it” will become less viable.

The sensible stance for freeholders is to standardise how they handle deployment requests:

• Pre-approved building access processes
• Standard terms for comms cupboard access
• Clear points of contact
• Documented rules for reinstatement and fire safety

For ISPs and altnets

Operators should be honest with themselves: MDU success is not just about rights; it’s about trust and execution. If installs are messy, noisy, or poorly communicated, you’ll generate resistance even when the law is on your side.

In my experience, the operators who win MDUs treat the building like a long-term customer, not a one-time construction site.

The bigger picture: gigabit coverage is the floor, not the ceiling

The UK’s national gigabit ambition (with 2032 as the stated horizon) is often discussed as a coverage target. But for telecom leaders focused on AI, it’s better understood as a capability baseline.

More fibre to more premises means:

• Higher and more variable traffic patterns
• More stringent latency expectations (especially as Wi‑Fi 7 and edge services spread)
• More pressure to automate assurance and operations

If flats remain under-served, network planners get forced into inefficient patchwork solutions. If flats get connected at scale, AI-enabled operations become more viable:

• Better end-to-end observability
• Cleaner service assurance loops
• More predictable churn and customer experience

The consultation’s intent is straightforward: reduce obstruction and accelerate engagement. The strategic implication is bigger: a smoother MDU rollout environment makes AI in telecommunications more effective, because it removes a major source of operational randomness.

If you’re building an AI roadmap for network rollout or network optimisation in 2026, track this policy closely—and treat “permission friction” as a problem you can model, not just a headache you tolerate.

What would your fibre build plan look like if “unresponsive freeholder” stopped being a top-three delay reason?