Britain Bets GBP2 Billion AI Can Fix NATO Training

Britain is spending GBP2 billion on something NATO armies usually pretend they already have: a training system that can keep up with the war they are studying every night on Ukrainian drone footage.

The Ministry of Defence announced on 10 July that Omnia Training, a Raytheon UK-led consortium, has won a 15-year contract to build the British Army’s new Collective Training Service. The centrepiece is a Combat Laboratory using AI, advanced analytics and virtual environments to train up to 60,000 soldiers a year. The official language is polished. The underlying admission is not. The British Army thinks its old training machinery is too slow for the battlefield now taking shape in Europe.

That matters well beyond Britain. London says the system will strengthen support for NATO and allies, and it is right to frame the programme that way. If the alliance has learned anything from Ukraine, it is that readiness cannot mean a brigade completing a periodic exercise, filing its lessons learned, and waiting months for doctrine to catch up. The battlefield changes faster than that.

The Contract Is About Speed, Not Software

The headline numbers are straightforward. According to the UK MOD, the Army Collective Training Service sits inside the Collective Training Transformation Programme. It will support around 400 jobs across the UK, including 270 skilled roles, and draw on a supply chain of 44 British businesses. The consortium includes Raytheon UK, Capita, Cervus, Rheinmetall UK and Skyral. The contract also includes 100 apprenticeships.

Those figures will please ministers. But they are not the strategic point.

The useful part is the Combat Laboratory itself. The MOD says it will integrate simulation, live systems and analytics, assess operations, spot patterns and monitor performance with data and AI. It is supposed to support training from teams of about 100 soldiers up to formations of 50,000. That range is important. Ukraine has shown that tactical adaptation and operational coordination are now inseparable. A platoon’s drone discipline, a battalion’s electronic warfare plan and a corps-level logistics decision can all collide within the same afternoon.

Traditional training often separates those problems because it has to. Range space is limited. Opposing forces are expensive. Data capture is uneven. After-action reviews depend on what observers saw, what commanders admit and what the exercise design allowed. An AI-enabled training system does not magically solve those constraints, but it can make them less crippling. It can preserve more data, run more repetitions and expose commanders to patterns they would otherwise miss.

Ukraine Is The Unwritten Requirement

The MOD was unusually explicit that the programme is learning lessons from Ukraine. That is the right benchmark. It is also a brutal one.

On 10 July, the BBC reported from Zaporizhzhia that local authorities were responding to intensified Russian drone and missile pressure with more shelters, anti-drone nets and anti-shatter film on public buildings. The acting mayor said Russian attacks were hitting municipal transport, private buses, cars, residential buildings and children. Local authorities said 884 Russian drones were intercepted in the last week of June alone. That is one city’s defensive rhythm, not a NATO exercise inject.

This is where the usual Western training vocabulary starts to sound thin. Interoperability is necessary. Combined arms is necessary. Mission command is necessary. None of it is enough if units are not trained against mass cheap drones, degraded communications, fast target cycles, contested logistics and the political pressure created by strikes on civilians.

The US Department of War made a similar point from another angle on 8 July, when Joint Interagency Task Force 401 published its counter-drone handbook. The guide stresses layered defence, common vocabulary and permissions for warfighters to respond to drone threats in real time. Its most important line is almost mundane: there is no silver bullet. That is exactly why training matters. If there is no single technical answer, armies need soldiers and commanders who can combine detection, electronic warfare, kinetic defeat, deception, dispersion and local decision-making under pressure.

The Risk Is A Beautiful Simulator With Old Assumptions

Here is the uncomfortable part. NATO has a habit of buying sophisticated training systems and then feeding them comfortable scenarios.

A Combat Laboratory will only be as useful as the assumptions placed inside it. If the British Army uses the system to rehearse tidy manoeuvre with occasional drone interruptions, it will have wasted a rare opportunity. If it models enemy adaptation seriously, measures failure honestly and forces commanders to operate when networks break, it could become one of the more important readiness investments in Europe.

The danger is not that AI will replace judgement. The danger is that AI will produce beautifully quantified reassurance. A dashboard can tell a commander that a formation improved its decision cycle by 18 percent. It cannot decide whether the exercise was hard enough. It cannot force senior officers to accept politically ugly lessons. It cannot make a peacetime army enjoy being embarrassed.

That is where Ukraine’s lesson is sharpest. Adaptation there has not been elegant. It has been driven by punishment. Units learn because the alternative is death, failed attacks and cities under nightly strike. NATO cannot and should not replicate that cost in training. But it has to replicate more of the friction.

What NATO Should Watch

Britain’s GBP2 billion bet should be judged by three tests.

• First, realism: does the Combat Laboratory simulate dense drones, electronic warfare, degraded GPS, ammunition limits and civilian infrastructure pressure as normal conditions rather than special events?
• Second, alliance utility: can NATO partners plug into the system, share lessons and compare performance, or does this become another national capability with polite interoperability language attached?
• Third, institutional honesty: will the Army publish hard lessons and change doctrine quickly, or will uncomfortable findings disappear into internal briefings?

The programme also sits inside a broader British promise to make the Army ten times more lethal by 2035, backed by a GBP298 billion defence investment plan over four years. That slogan invites scepticism because all such slogans do. But the training contract is at least pointed at a real weakness. Europe does not just need more tanks, shells and air defence interceptors. It needs formations that can learn faster than Russia can adapt.

My view is simple: this is a good investment if Britain uses it to make commanders uncomfortable. It is a bad one if it becomes a procurement success story with virtual explosions and tidy metrics.

The next thing to watch is whether allies are allowed into the laboratory in a meaningful way. NATO’s training problem is collective. Russia will not care which national budget line paid for the simulation. The alliance either learns at wartime speed, or it keeps discovering, exercise by exercise, that Ukraine has been teaching the class all along.