Hit enter to search or Esc to close

`
Take Me Back

Rethinking defence for the age of AI

By Ben Legg & Sam Lombard

UK MOD © Crown copyright 2025

Over the past month, we at Newton have been exploring one question from multiple angles: what does AI really mean for the way complex organisations work? Across our recent Enterprise AI Summit at Bletchley Park and our discussion gatherings with leaders in individual sectors, the challenge has shifted from “What can AI do?” to “What does AI allow us to do differently?”

Nowhere is that question more urgent than in defence. AI is removing historic constraints on productivity, capability and pace, while fundamentally changing the threat landscape. Choosing not to move carries its own risk to our international standing and national security. The challenge now is how radically defence is prepared to rethink the way its people, organisations and wider enterprise work.

Sam Lombard (Director, Defence & Aerospace) caught up with AI-advocate and former COO of Google Europe, Ben Legg. His perspective is shaped by four very different career chapters: starting out as an engineer in the Army; moving into major businesses including Coca-Cola India; helping scale Google Europe; and, today, working with global businesses and investors on how AI and agentic organisations will reshape productivity.

Across those worlds, the technology has changed dramatically. But one challenge has remained remarkably consistent: how do large organisations move faster when the environment around them is moving faster still?

Sam spoke to Ben about what defence can learn from technology, why leaders shouldn’t delegate AI, and what the relationship between human expertise and machines could mean for the future of the defence enterprise.

Sam: We hear constantly in public discourse as well as across a huge range of sectors that AI represents a profound shift in the way we – as humans and as organisations – work. Is this really different from previous waves of technological change?

Ben: I think it is. Historically, major technologies have taken decades to embed; with AI, the cycle is becoming dramatically shorter. One of the biggest constraints on developing technology used to be the number of engineers available to build it. That constraint is disappearing because increasingly, if you can conceive of something, AI can help you prototype, build and iterate it at a speed that simply wasn’t possible before.

I’ve seen teams triple their productivity in ten months. At an individual level, simply becoming competent with the tools available today can make someone materially more productive almost immediately. And this is going to accelerate, which means leaders cannot afford to treat AI as a technology project that sits somewhere else in the organisation.

Sam: We see a massive challenge across all our defence partners to increase workforce capability, skills development, and retain talent. Does that demand mean that AI could eventually replace large numbers of jobs?

Ben: I don’t think the evidence necessarily points to a simple jobs apocalypse.

Take radiology. Geoffrey Hinton famously argued that we should stop training radiologists because AI was becoming better at elements of the job. Yet the number of radiologists has continued to rise, as have salaries. Why? Because technology solves one set of problems and humans move on to the next set.

We’re never going to decide that healthcare is good enough, education is finished, transportation cannot improve or entertainment has met every possible need.
In the longer term, AI and robotics may do much more of the actual ‘doing’ of work. That makes the human role different: identifying unmet needs, framing the right problem, exercising judgement and empathy, and working out what ought to happen next.

Sam: So expertise becomes more important, rather than less?

Ben: Absolutely: you need to know enough to challenge the machine. AI is extraordinarily powerful at looking backwards and increasingly good at understanding the present. I’m less convinced by its ability to imagine genuinely new futures. That’s where human expertise, curiosity and critical thinking matter.

A knowledgeable person using AI can be dramatically more capable than either the person or the AI operating alone. But someone who simply accepts whatever the model says risks amplifying poor judgement.

The combination we’re aiming for is expert humans plus AI, and that has implications for the workforce we develop. Young people entering employment need to be power-users of AI, but they also need genuine expertise in something. Knowing how to ask a machine a question isn’t a substitute for knowing whether its answer makes sense.

“A human plus AI will remain more powerful than AI alone.”

Ben Legg

Sam: There’s a really interesting question here on what AI means for the future of defence capability. It’s an uncomfortable premise, but should we be moving away from ‘traditional’ materiel to invest in a more diverse capability portfolio?

Ben: For me, the national security argument changes the equation. Defence cannot afford not to use AI, because our adversaries are using it.

The threats we need to defend against are broader than conventional military platforms. An adversary might attack elections, energy, water, communications, financial infrastructure or the systems that allow society to function. So, we have to ask a broader question: what are we actually defending, and are we organising our resources around the threats we will face rather than those we have traditionally planned for? Cyber, AI with sovereign capabilities, autonomous systems, intelligence and national resilience can potentially create completely different forms of defensive capability.

That isn’t an argument for abandoning traditional capability, I think there’s a need to continually challenge the balance.

Sam: So does technology require defence to rethink how it makes decisions as well as what it buys?

Ben: I think that’s fundamental. Technology is accelerating the world around defence, yet many of the processes governing decisions were designed for an era of much slower change. If a capability takes years to specify and build, but the underlying technology changes every few months, there is an obvious tension.

Should we be designing more capabilities to evolve continuously? Should we revisit assumptions every quarter rather than committing to requirements that could be overtaken before they are delivered? Should we place multiple smaller bets rather than waiting until we are certain about one answer?

Those aren’t primarily technology questions; they’re about operating models, regulation, risk and leadership – issues which all sectors grapple with.

Sam: But defence operates in a highly regulated environment for good reason. How do you reconcile speed with the right assessment of risk?

Ben: Highly regulated environments naturally find innovation harder. You see the same tension in financial services – organisations cannot simply deploy models they don’t understand, there have to be safeguards. But regulation can also create cultures where avoiding risk becomes almost synonymous with doing a good job.

But we know that behaviour can change when the situation demands it. Covid demonstrated that organisations can place several bets, iterate quickly and accept a level of uncertainty when the cost of moving slowly or not at all becomes unacceptable. I’m firmly in the camp of ‘why do we need to wait for a crisis before behaving with that urgency?’

UK MOD © Crown copyright 2025

Sam: You’ve worked in organisations operating at very different speeds. What did Google teach you about that?

Ben: Part of the Google mentality was deliberately trying to achieve in a quarter what many companies would expect to achieve in a year. That requires a different relationship with failure. If everything you attempt succeeds, you probably aren’t stretching yourself enough. A 70% success rate may actually tell you more than 100%, because 30% failure creates learning.

But that only works with the right culture: open and honest, engineering-led, prepared to challenge assumptions, and ultimately willing to let the right answer win regardless of where it comes from.

Sam: Defence is an incredibly complex enterprise with hundreds of moving parts which all have to coalesce in the right way for the right things to be delivered. Could AI help that enterprise make better collective decisions to enable greater progress?

Ben: Potentially, this is one of the most interesting opportunities to align all the organisations around the same problems and making decisions faster. AI could help organisations interrogate enormous amounts of information, test assumptions and surface options in ways that humans simply cannot do at the same speed.

But the critical word is help. AI should enable people to make better decisions, not remove responsibility for making them. A machine can give you another brain in the room. It can challenge a hypothesis, find evidence, analyse thousands of documents or offer an alternative interpretation. But people still need the expertise to challenge what comes back. A human plus AI will remain more powerful than AI alone.

“For defence leaders, the starting point should be practical: identify the three to five biggest barriers to delivery – slow decisions, skilled people tied up in routine work, information trapped between organisations or delays in mobilising resources. Choose a problem, explore where technology could remove the constraint, and start experimenting. Progress comes from using the tools, challenging the results and building on what works, rather than waiting for a definitive AI strategy. In a world where technology is getting faster, threats are changing and our adversaries are learning too, ‘no’ cannot be the default answer. It has to be ‘There must be a way’ and not ‘Can we do this?’”

Phil Sunley, Global Head of Defence & Aerospace Newton

Ben Legg

Entrepreneur, AI advocate & former COO of Google Europe

Ben Legg

Entrepreneur, AI advocate & former COO of Google Europe

Ben helped relieve the Siege of Sarajevo, turned around Coca-Cola India, wrote the hyperscalers playbook as COO of Google Europe, and ran three private equity-backed tech businesses. Today, he helps CEOs design and build their AI-powered future, and helps investors see where value is being created and destroyed in the fast-changing agentic economy.

Sam Lombard

Director, Defence & Aerospace

Sam Lombard

Director, Defence & Aerospace

Newton in defence & aerospace

Find out more

Select country

Visit the US site en-us
Visit the UK site en
Back to the Australian site en-au

It seems like you're in XXXX

Would you like to visit our XXXX website?

Visit the US site en-us
Visit the UK site en
Continue to the Australian site en-au