As defense forces around the world race to scale up, suppliers are feeling the pressure. Requirements are getting more complex. Cybersecurity is a must, and AI is unavoidable. All of this can slow down design cycles, jam up manufacturing and delay delivery — exactly when speed matters most.
But this is where AI — and especially agentic AI — also starts to change the game. It takes the friction out of the entire workflow by sorting through mountains of data to pinpoint requirements, running simulations through digital twins and spotting more efficient logistics paths. But speed alone isn’t the goal. Making AI work in practice also means managing change carefully so that data can be trusted and people can make critical decisions based on clearer, better information.
Breaking Defense discussed these challenges with NTT DATA’s Nick Williams, Head of Manufacturing, Consumer Packaged Goods and Defense, UK and Ireland, and Torsten Albrecht, Managing Director, Head of Industry and Services.
Breaking Defense: What value does AI bring to supply chain management? Why do organizations in the defense industry need it?
Williams: Looking at the geopolitical environment right now, there are challenges everywhere. For our defense clients, demand is high. They’re sitting on significant backlogs, with requirements piling up, and that puts enormous pressure on them. They’re being pushed to increase their speed to market and cut costs while remaining resilient.
What’s unique about the defense industry is the way consortia work. It’s how many of these manufacturers go to market. You already have one company dealing with speed-to-market issues, operational complexity and supply chain logistics. Now multiply that by 10 when you bring in partners across the wider ecosystem.
This is where data, AI and related technologies become hugely relevant. They’re central to helping our clients address the challenges they’re facing right now. Most of the clients we speak to are trying to double or even triple their output over the next few years — with fewer people and in a far more complex world.
Albrecht: The US defense industry has always been strong, largely because demand is so high. The US military is constantly renewing and upgrading its equipment. In Europe, there are also strong players in the defense space, but historically European countries weren’t investing anywhere as much as the US.
That’s changed dramatically. We’re now talking about governments and national militaries spending 10 times more than before. As a result, companies must deliver far more material, faster.
At the same time, battlefield technology is changing fast. New products need to be engineered, time to market has to shrink and production has to scale. And all of this is happening while raw materials are already in short supply. Where do the chips come from? Where do the components and subcomponents come from?
Look at drones. When the war in Ukraine started, the drone market was relatively small. Now, it’s become one of the decisive factors on the battlefield. From what we can see, drone manufacturing is probably the most dynamic part of the industry right now, and it’s facing challenges at every turn.
The interesting thing is that money isn’t a bottleneck. Budgets are high. But even with funding available, many companies are still struggling to keep up with demand.
How does AI help solve the problem of a surge in production demand?
Albrecht: Think about the new products militaries are asking for. In the past, every national military defined its own requirements. Germany would add this requirement, the UK would add that one, and so on. It quickly became very complex.
Now we’re pulling all these requirements together, and they are massive documents. Not 20 pages, but 2,000. An engineer must sit and sift through everything: What’s relevant to me? What do I actually have to deal with? Is anything misleading or redundant? They have to read it all and extract what matters.
This is where GenAI makes a real difference. It cuts that effort down significantly. An engineer can simply say, “I don’t know the mechanics of the laser. Show me all the requirements related to laser technology,” and get straight to what they need.
Williams: In defense manufacturing, AI’s real impact isn’t just about automation. It’s also about enabling trusted, federated decision-making across consortia. Data still needs to stay where it belongs — within different areas of the company, inside a zero trust environment — but AI can surface the most relevant insights under the right policies without breaking those boundaries. That’s the exciting part. It’s what makes real interoperability possible, and it’s how companies across Europe and globally can get to market faster.
There has been plenty of talk about hallucinations in GenAI. Is there a similar problem with agentic AI? What hurdles do companies need to overcome with this technology?
Albrecht: The first and most obvious hurdle is data quality. If your data is bad, the agent will fail, no matter how good the agent itself is. The real risk comes when agentic systems operate autonomously. If something goes wrong, you’d better detect it fast. Otherwise, those agents can start triggering processes you never intended them to.
This isn’t a new topic, but the impact of poor data quality is much bigger than it used to be. In the past, there were more points of human interaction and oversight, so you could stop problems early. Now, as automation kicks in, you might only spot an issue much later. By then, addressing it can be very expensive.
Data is a major hurdle, but there’s also the challenge of operating across multiple companies, especially in a defense context. Just because something is technically possible doesn’t mean it makes sense from a security perspective. Defense companies can’t experiment as freely as others. You can’t risk exposing what you’re doing externally, and that’s a critical constraint.
There’s also a persistent myth that AI is here to replace a lot of people. What we actually see is greater value in supporting people and removing time-consuming work. That creates its own challenges, though. Some of those lower-skilled, repetitive tasks — like data cleansing — will shrink, and people will need to be upskilled to take on more analytical, decision-focused roles.
That’s where change management comes in. The real hurdle is human readiness for AI. Do people actually accept it? This is often overlooked. Whether it’s management, operations or frontline workers, the human side of AI isn’t there yet. And without that, the technology won’t deliver on its promise.
Williams: I’d actually frame change management a bit differently. When you look at AI and the level of investment it needs, you probably need to invest just as much in change management if you want it to succeed.
When we talk to clients, there’s still a lot of mixed messaging. Is AI just a buzzword, or is it real? How do we prove the value? That’s why many people argue for starting small and showing quick wins. But if you talk to most advisors, they’ll tell you the opposite: Do the change management properly, engage stakeholders early and invest up-front.
The real opportunity is to go big but remain focused. Pick one or two projects, implement GenAI properly, and do it end to end. The interesting question is how you drive real change and clearly demonstrate value. Not just in terms of resistance — people worrying about being replaced — but also in building the momentum, governance and trust needed to make AI work inside the organization. AI can be incredibly powerful, but it’s also disruptive.
This is a big shift. Clients have moved from simply observing the benefits of AI to asking, “How do I execute this? How do I make it real?” It’s no longer a maybe. At this point, it’s a necessity.