The MOD Can't Afford to Build Software the Old Way Anymore

The Ministry of Defence's Equipment Plan has a £42.5 billion affordability gap over ten years. The Defence Investment Plan, intended to replace it with a credible funding path, remains unpublished months after it was promised. Westminster insiders point to a £28 billion shortfall that neither the Treasury nor the MOD can resolve.

Meanwhile, defence spending is rising from £62.2 billion in 2025/26 to £73.5 billion by 2028/29. More money is going in than at any point since the Cold War. And yet the gap between what Defence needs and what it can afford keeps widening.

The traditional response to this problem is familiar: defer programmes, hollow out capability, cancel projects, stretch timelines. The Deputy Chair of the Public Accounts Committee has called it a "broken system" of "multi-billion-pound procurement problems, with equipment arriving years late and over-budget."

But there is a structural inefficiency embedded in every software and digital programme across Defence that nobody is seriously addressing: the cost of building software the way we have always built it.

The Old Way Is Too Slow and Too Expensive

Defence software programmes routinely take 18 to 24 months from requirement to delivery. Cost overruns of 10 to 25 percent are considered normal. Scope creep is endemic. By the time capability reaches the user, the requirement has evolved, the threat has changed, and the people who wrote the original specification have moved on.

The NAO has documented this pattern repeatedly. A limited number of specialist suppliers consistently fail to deliver to time and cost. The MOD itself lacks sufficient project management skills internally. And short-term financial planning around the equipment portfolio creates perverse incentives – cancelling one programme to fund another, generating knock-on costs that exceed the original savings.

None of this is new. What is new is that a proven alternative now exists.

The Evidence Is Already In

The government's own data makes the case. DSIT's AI-assisted development trial across 1,000 developers in 50 departments found that AI-accelerated coding saved 28 working days per developer per year. Projected across the civil service, the estimated saving is £45 billion.

The commercial sector is further ahead. A 2026 industry survey found that 91 percent of software development companies have already adopted AI tools for coding, testing, documentation, and project planning. This is not an emerging trend. It is the new baseline.

AI-accelerated software engineering does not simply make developers type faster. It changes the fundamental economics of delivery. Functional prototypes that would take months can be produced in days. Production-ready applications that would take a year can be delivered in weeks. The same capability that a traditional system integrator quotes at £5 million over 18 months can be delivered for a fraction of that cost in a fraction of the time, with full governance, audit trails, and security assurance built in from the start.

The maths is not complicated. If the MOD's software and digital portfolio runs into the billions annually, and AI-accelerated development reduces delivery costs by 60 to 80 percent on suitable programmes, the savings compound rapidly. Not over decades. Over the current spending review period.

Three Barriers Standing in the Way

If the evidence is this clear, why is Defence not already doing this at scale? Three structural barriers.

Procurement structures that favour incumbents. Multi-year framework contracts with established system integrators do not accommodate rapid, iterative AI-accelerated delivery by smaller, more agile suppliers. The Procurement Act 2023 explicitly enables smaller lots and SME reservation. The MOD has set a £2.5 billion annual target for SME procurement spend. The mechanisms exist. Procurement teams have not yet operationalised them for AI-accelerated capability.

Security concerns used as a brake rather than a design constraint. "AI cannot be trusted for defence systems" is the reflexive objection. It sounds responsible. It is not. The Five Eyes cybersecurity agencies published joint guidance on 1 May 2026 setting out exactly how to adopt agentic AI carefully, with layered security controls, least-privilege access, behavioural monitoring, and kill switches. The framework for secure AI adoption exists at the highest levels of allied intelligence. The question is not whether AI-generated code can be made secure. It is whether the governance frameworks are in place to ensure it. For those who have built them, they are.

A culture of "big programme" delivery. Defence procurement is designed around multi-year, multi-billion programmes with long approval cycles, extensive governance boards, and sequential delivery milestones. AI-accelerated development works best on rapid, iterative, smaller engagements: proofs of concept in weeks, minimum viable products in months, production capability in a quarter. The delivery model does not fit the procurement model. Until one of them changes, the cost savings remain theoretical.

What Needs to Change

Four things would unlock material savings within the current Defence Investment Plan period.

Ring-fence funding for AI-accelerated delivery. Even 5 percent of the software and digital budget directed toward rapid, governed AI development would demonstrate the cost model at scale. Not as a pilot. Not as an innovation experiment. As a parallel delivery track with real programmes, real users, and real accountability for outcomes.

Use existing procurement routes. G-Cloud, DASA, the Defence Sourcing Portal, and the new £80 million Sovereign AI procurement programme are all available now. The £80 million programme specifically targets cybersecurity and national security AI capabilities, with contracts up to £5 million per project. Buyers do not need new frameworks. They need to use the ones they have with the urgency the funding gap demands.

Demand governance, not prohibition. The answer to AI security concerns is not to ban AI from defence software development. It is to mandate governance: human-in-the-loop assurance at every stage, security-cleared engineers reviewing every output, compliance gates embedded in the development lifecycle, full audit trails from requirement to deployment. Blanket prohibition does not make Defence safer. It makes Defence slower and more expensive while adversaries adopt AI without hesitation.

Engage the UK SME sector. Small and medium enterprises have been building governed AI development practices for the past two years while large system integrators run internal pilots and update their corporate strategies. The capability exists in the UK supplier base today. The SME procurement target exists for exactly this reason. The Procurement Act enables it. The only missing ingredient is the willingness to engage at pace.

The Strategic Reality

The Defence Investment Plan is not just a financial document. It is a statement about whether the United Kingdom can deliver defence capability at the pace the threat environment demands.

Russia is modernising its military capability and testing British defences through cyberattacks, sabotage, and airspace incursions. China is investing in AI-enabled military systems at scale. Both are using AI to accelerate their own capability development. Neither is waiting for the UK to resolve its Treasury negotiations.

AI-accelerated software engineering is not a technology curiosity or a nice-to-have efficiency saving. With a £28 billion funding gap, a threat environment that is not waiting for Treasury sign-off, and proven evidence that AI development reduces costs by the majority of the original budget, it is a strategic necessity.

The old way of building defence software is a luxury the MOD can no longer afford. The question is not whether AI-accelerated development will become the norm in Defence. It is whether the UK adopts it fast enough to matter.

---

Signal Coding is a UK-owned AI-accelerated software engineering company serving Government, Defence, and National Security. We deliver governed AI development with defence-grade security assurance: human-in-the-loop oversight, SC/DV cleared engineers, and full audit trails from requirement to deployment. Contact us to discuss how AI-accelerated delivery can close the gap between what your programme needs and what it can afford.