Innovation at Thales: An interview with Patrice Caine

The conflict in Ukraine has highlighted the critical role technological innovation can play in the defense sector. The use of technologies such as drones, cybersecurity, high-speed dissemination of information, and electronic warfare, along with their countermeasures, have proven to play a particularly important role.

In this interview, Patrice Caine talks to Hugues Lavandier, senior partner coleading McKinsey’s Aerospace & Defense Practice worldwide, and Alexandre Ménard, senior partner coleading the McKinsey Industry & Technology Practice in France, about the critical importance of innovation in the defense sector. He discusses the role of two major technologies, in particular, AI and quantum, and how the development of a combination of scientific and digital skills is crucial for technologies tailored to the needs of defense players. Patrice Caine has been the chair and CEO of Thales since December 2014 and chair of the National Association of Research and Technology (ANRT) in France since December 2019.

McKinsey: Thales is a leading technology company. That’s why we would like to start this discussion by talking about a subject that’s right at the heart of the news: AI. It is developing at breakneck speed and has great potential to reshape the economy. What do you think is, and will be, the impact of AI on the defense sector?

Patrice Caine: The deployment of AI in the defense sector is already a reality and is now taking on more of an operational than prospective characteristic for the industry. Let me give you two examples of technologies where AI has already had a significant impact: sensors and decision-support systems.

Let us talk first about sensors, such as radar and sonar. Currently, the level of performance of our radar systems enables us to detect very small, very slow-moving objects, such as prowling munitions or drones. But it is difficult and time-consuming for a human operator to differentiate them from other flying objects, such as birds. The introduction of AI provides real added value in terms of interpretation: It can quickly and reliably determine the nature of these objects and identify those relevant to air surveillance. This is an extremely important advantage when it comes to protecting ourselves against this type of threat, which has increased significantly in recent years. These functions are already present in some of our radars.

A second use case relates to “Command and Control” decision-support systems. The use of these systems includes preparing for missions consisting of highly complex operations, for example, an aerial reconnaissance mission. This phase, in particular, involves making decisions on aspects such as flight trajectory, altitude, speed, the number and configuration of aircraft, and take-off time. It involves resolving an extremely complex combination of factors to be able to arrive at the best possible decision, the one that will meet the objectives while ensuring pilots’ safety and making the best use of resources. AI algorithms have demonstrated that they add real value by quickly finding the best possible compromise—and then leaving it up to the operators to accept or reject what the AI proposes.

But let’s be clear: Whether we’re talking about sensors or support systems for decision-making, AI does not question existing technologies. In some cases, its contribution can be a real game changer, but it works complementary to systems that are, in themselves, already highly advanced and have proven their worth.

The integration of AI-based technologies nevertheless presents a number of challenges, including that of exportability. Obviously, for security reasons, a system that has been trained on operational data from live action would be excluded from being exported, as we do not own such data, and it contains classified defense information. Moreover, it would be inappropriate to use the same training data for all use contexts. The interpretation of a sensor’s results must be adapted, particularly to the circumstances in which it will be deployed. To meet this dual challenge, we pretrain our algorithms using industrial and synthetic data, while fine-tuning is carried out using data from the end customer.

McKinsey: How do you reconcile the length of time it takes to structure the development and implementation of defense projects and platforms with the need for the rapid deployment of architecture and software capable of technological acceleration?

Patrice Caine: We proceed by successive extensions, by modules. Let’s take the example of a platform such as the Rafale. When we introduce AI into a reconnaissance pod, we take care to make as little impact as possible on the Rafale’s weapons system. We integrate this additional capability into the pod to assist the pilot without impacting the rest of the platform. Of course, this imposes very strong design constraints in terms of space and energy consumption. In this respect, it is less complex to integrate AI on a ship, for example, where it is possible to take data hubs on board designed to retrieve onboard information and process it.

McKinsey: What, in your view, is the difference between the incumbent defense companies and large tech players that are increasingly open to hosting defense applications or agile start-ups that have more recent experience with defense issues?

Patrice Caine: In the defense sector, even more than in other sectors of the economy, the need to be anchored in operational reality is crucial. In my view, there are four key elements that need to be mastered in order to be credible.

First, to harness the potential of AI, one must have a high level of mastery in both the digital and physical fields. It is vital to understand the underlying physics of defense technologies. This is particularly true where sensors are concerned. Here is a concrete example: To me, it seems extremely complicated to develop algorithms for underwater acoustics without knowing exactly how sonar works. How does an acoustic wave make a ceramic resonate? How does this ceramic react to produce an electrical signal from this acoustic wave? How do you convert an electrical signal into an analog signal to deduce the nature of the object detected? To enhance the performance of a sensor by using AI, knowledge of digital technology alone is not enough. It’s the combination of highly specialized expertise in digital technology and the physical sciences that makes it possible.

Second, it is necessary to have a good understanding of the concepts behind the use of the technology. In the same example of submarine sonar, it is essential to know what operators are trying to detect and what they are trying to understand when they are in the field. In-depth knowledge of these needs is required to successfully combine theory with diverse and complex operational realities. A company such as Thales benefits from decades of close collaboration with defense forces. This is a considerable asset when it comes to developing AI that is truly useful to them.

Next, it is essential to be able to understand issues related to embeddability. On a combat aircraft, the computing power available is often limited due to the physical space available and the electrical power capacity. And, for reasons of security and stealth, it would be unthinkable to have a constant connection to the cloud. Add to this the constraints of electromagnetic radiation or even thermal resistance (military equipment must be able to operate between below 40°C and over 90°C). These are major technical challenges for new entrants and pure tech players.

The fourth point concerns cybersecurity. AI provides an additional capability with high added value, but it also represents a proven source of vulnerability, which we must be able to guard against. A few pixels altered in an image during a cyberattack can lead to errors in the interpretation by an AI algorithm. Imagine if this led to a tank being identified as a civilian vehicle and the consequences that could have. Thanks to our 6,000 cybersecurity experts, a company such as Thales has the capacity to protect its AI against these attacks. But few companies in the world have this type of capability at their disposal.

Combining physical and digital sciences, having a good understanding of the concepts of deploying the technology, mastering the constraints of embeddability, and being able to protect against cyberattacks—the number of credible players able to meet these four challenges simultaneously is very small. In this respect, the large historic tech companies such as Thales have a head start.

What I’m saying is slightly less true regarding decision-support systems, which are more about software than the hardware. The barriers to entry are, therefore, a little lower. Even so, this type of system has between 10 million and 20 million lines of code. You must first be able to master the operation of a gigantic piece of software before you can envisage integrating AI into it.

McKinsey: The topic of leadership often comes up when talking about defense matters. What is the role, not just of AI but of innovation in general, in terms of leadership for defense players? More specifically, what will allow European defense technology players to maintain their leading positions? And what will make a difference?

Patrice Caine: I’m convinced that, after several years in which a great deal of attention has been focused on digital technology, disruptive innovation in the coming decades will be driven by the physical sciences. Coming back to AI: We mustn’t forget that large language models (LLMs) can only be trained using existing data and knowledge. How can we create something truly revolutionary on the basis of what already exists? On the other hand, in the physical field, we are identifying major disruptions ahead. I see the physical sciences getting even with the digital sciences, and in this field, quantum technology is the next big revolution.

The first quantum revolution was an upheaval, even though it is not necessarily widely known. It led to the development of atomic clocks that gave rise to GPS and even lasers, MRIs, and transistors. The second revolution will take us a step further by taming the extremely puzzling characteristics of matter at the elementary level.

McKinsey: You talk of a quantum revolution. How will it materialize in the defense sector? What are the most promising use cases, and when do you see them becoming available?

Patrice Caine: First of all, I’d like to make it clear that I’m not talking about the development of the quantum computer. Thales is not involved in that race, as building computers is not one of our specialties. As for the rest, no one knows when this quest will succeed. However, we are working in areas which we are more certain will give rise to viable technologies that can be industrialized in the medium term.

In particular, the second quantum revolution will give us access to radically higher performance in the field of sensors. We’re not talking about incremental improvements but improvements by a factor of between 100 and 1,000. The technologies involved already exist and have proven their worth, but the challenge now is to take them from prototype to the industrial stage and then to build the economic models.

For example, quantum sensors could have a major impact on inertial navigation systems. Today, without recalibration, the most efficient of these drift by around a kilometer on a journey between Paris and the East Coast of the United States. With a cold atom inertial unit, based on quantum technologies, it shifts from a difference of kilometers to a difference of meters, an improvement of a factor of 1,000. This is of interest to civil aviation, but it is especially of great interest to the defense sector. In the case of a nuclear submarine, for example, such a system will create the ability to avoid having to surface to carry out recalibrations—a major operational advantage! So, I’m deeply convinced that the initial use cases will more likely come from the defense sector.

Let’s take another example, that of SQUIDs and SQIFs. These technologies make it possible to dissociate the size of antennas from the frequency of the signal to be transmitted or received. The potential impact on the defense industry is enormous. Currently, to communicate at very low frequencies (particularly with submarines), antennas several hundred meters long are required. With these new devices, antennas the size of a fingernail will be able to perform the same function. You can imagine the advantages, particularly in terms of stealth.

McKinsey: In this new era, talent is essential, whether it’s engineers or tech talent. What are the implications of these technological innovations in terms of talent acquisition? In one of our recent articles, we highlighted the challenges faced by European players when they encounter a shortage of skilled workers in a highly competitive labor market, making attracting and retaining talent a major challenge. How is Thales competing with the big tech players in terms of attracting talent?

Patrice Caine: When it comes to talent, Thales has the undeniable advantage of being a strong brand with exciting projects for young engineers across all verticals. What’s more, we combine a high level of expertise with concrete social benefits. Aeronautics enthusiasts who join us, for example, have the possibility of contributing to improving the environmental performance of aircraft. Others join us because they want to help protect their country’s sovereignty or help combat cybercrime.

Another one of our strengths is the opportunity offered to our talent to change their discipline or area of expertise during the course of their career, for example, from the world of defense to that of aeronautics, from space to cyber, et cetera. Our employees can also change their technical discipline from physical to digital sciences or a combination of both. However, in actual fact, many of them become so passionate about their field that they often don’t wish to change.

Our positioning also gives us the advantage of working in long cycles and balancing volatility caused by an increase or decrease in activity, which helps us build loyalty and avoid losing skills. However, we remain focused on this issue. Continuing to attract qualified talent over the long term is a big challenge, as our employees are our main resource.

McKinsey: When it comes to attracting and retaining talent, we see that new generations are paying more attention to environmental and social aspects, placing meaning at the heart of their career plans. How are you responding to this trend?

Patrice Caine: When I joined Thales 22 years ago, environmental or ethical issues were less central to the recruitment process. Today’s young talent is much more focused on the social purpose of our business. This has led us to think and formulate our role differently. Today, we express it in terms of a triple ambition: to help make the world safer, greener, and more inclusive. Everything we do relates to at least one of these objectives.

Our defense, security, and cybersecurity activities enable our clients to protect their populations, their institutions, and their physical and digital infrastructure. It’s not always well-known, but Thales is currently one of the biggest global players in terms of the security of applications and digital data. In aeronautics, we are contributing to the efforts to reduce the sector’s carbon emissions, in particular by optimizing flight paths.

Finally, by playing an active part in the fight against the digital divide, we are helping to build a more inclusive world. For example, we have developed the SATRIA satellite, which connects Indonesia’s 13,000 islands to the internet at an affordable cost. We also contribute to providing all human beings on the planet with a secure legal identity, a prerequisite to being able to vote, travel, or access social services.