In a global context shaped by geopolitical instability and cascading crises, industrial companies face a paradoxical demand: accelerate without ever sacrificing safety — whether in processes, product quality, supply chain, cybersecurity or lead times. At Tech For Industry Show 2026 (Paris Expo Porte de Versailles), a panel drew on the experience of the defense sector, used to operating under maximum constraint, with Patrick Aufort (Defense Innovation Agency), Anne-Pascale Guedon (Destinus France) and Benoît Molard (Boost Aerospace).

Destinus is a private European group that designs and produces low-cost operational combat drones (deep-strike drones and drone interceptors): 1,000 people, several hundred million euros in revenue, a €5 billion valuation. The AID, created in 2018 within the French Ministry of the Armed Forces, works as a "project factory" to detect, capture and mature innovations — including outside the classic defense field — through grants, public contracts and even equity stakes. Boost Aerospace, born from an initiative by Dassault Aviation, Safran and Airbus, supports the digital transformation and cyber-protection of the aerospace, space and defense supply chain.

For the AID, events have validated all its 2018 hypotheses: conflicts extending from the seabed to space, intangible fields (cyber, information warfare, electromagnetic spectrum), the mass use of civilian solutions. The Ukrainian conflict even began with a cyberattack on a space communication station.

Two phenomena weren't anticipated: the accelerating incorporation of innovation in the field — a new building block makes the difference then is countered within weeks — and the cost impact. The emblematic example: using a missile worth around one million euros to shoot down a drone costing a few tens of thousands. The challenge becomes: maintaining, in peacetime, an industrial base able to evolve on a few-week cycle and produce low-cost, while being ready to ramp up.

Accelerating doesn't mean anything goes. In Kyiv, the AID's counterparts cut a certification from 18 months to a few weeks — but reversed course on pyrotechnics after too many incidents: "we can't afford to lose human resources." The lesson: under pressure, you must know exactly where the red lines are, and that requires expertise.

Destinus was "born with Ukraine." Founded in Switzerland for a hydrogen hypersonic project, the company switched to emergency mode at the request of Ukrainian authorities. First need: an autonomous system flying low, under radar, able to strike deep despite jamming — hence a very rustic biplane, "Blériot-style," matched to the real need, produced using the best resources across Europe (plants in Germany and Spain, resources in France).

The two most sensitive innovations are the software (autonomy, inertial navigation) and the payload. For the rest, Destinus claims a "manufacturing secret": no defense-industrial-base dependency issue, because everything rests on open-source components, with value residing in in-house integration. The result: no dependence on a single source validated after 7 years, far better competitiveness, and the ability to produce mass at reasonable cost (with a "kill" performance deliberately non-maximal but sufficient, continuously improved in the field).

For Boost Aerospace, "safety first" doesn't preclude accelerating: the sector has strung together ramp-ups for decades, with growing customization on top (cabins adapted across 5 to 10 aircraft). The main bottleneck is almost always in the supply chain: a major manufacturer assembles parts made by others, including many SMEs. The solution: upskilling the supplier network, tooling processes (non-quality handling, chain management) and bringing all players up to the same speed.

The AID sums up its method in three points. First, trust: manufacturer, operational user and the ministry's engineering interface must know each other before the emergency, not discover each other at the last moment. Second, real-condition experimentation: the DGA created "reference centers" (counter-drone, naval surface drones, ground drones and loitering munitions, electronic warfare) where failure is allowed and the correction loop is ultra-reactive — Destinus tested products there between March and May, with deployment following. Third, expertise: knowing how to judge what is non-negotiable (safety of property and people) and what can be waived.

Destinus confirms: you don't show up at an experimentation center below TRL 6-7, but the real change is cultural. Where civil aerospace targeted nearly 100% (OTD, defects), defense tolerates a "kill" performance of 70-80%, because the product then improves continuously in the field. The condition: teams that understand the priority and urgency, and close interface with end users to pin down the real need and "how far we can go."

Boost Aerospace's advice is clear: manufacturers too often stack point tools without designing an overall transformation. With data as the company's "gold mine," you must build a global digital transformation project that integrates from the outset both the use and protection of data. Safety — cyber and product quality alike — isn't a layer added at the end: it's designed intrinsically, in the product and the processes. And a posture inherited from defense: experiment fast, fail fast, correct fast.

For the AID, the main human barrier is the relationship with risk-taking. Taking a risk isn't offloading it onto others: on a project, each must own their share. The contracting authority by not demanding unattainable performance and using all the public-procurement code allows; the manufacturer by launching before every technological lock is lifted; the operational user by accepting a non-perfect product, provided safety is guaranteed. So you must build awareness and culture at all levels, especially intermediate ones.

The AID raises a point set to become major: the interpretation of ethics. Facing adversaries who don't share the same values, the "human in the loop" principle for autonomous systems and AI must not be read as needing someone "behind a button" at every moment — at the risk of losing competitiveness — but as upstream human validation, or a field commander's decision after securing the situation. A cross-cutting issue for AI and the augmented human.

At Destinus (1,000 people, ~20 hires a week), it's a (measured) "love of risk" that motivates: experienced profiles from large groups (Airbus…) or special forces, who love adventure but know the limits and respect values. "With us, it's not resistance to change — it's more the change of resistance."

For Boost Aerospace, AI offers enormous productivity potential, but the central stake is governance: what do we want to do with it, what performance to seek, what to protect? — a trade-off to strike sector by sector (the example of Dassault Aviation, protecting its IP while adopting AI through dedicated protocols, is cited). For the AID, many AI use cases are already in the field, with sovereignty stakes (mastering the language model, choosing in-house vs cloud compute) and a need for explainable AI — a pilot will accept a recommendation only if they understand its reasoning.

Against received wisdom, the AID points to ground robotics as still immature: a ground robot operates in a heterogeneous environment (varied soils, obstacles, "negative obstacles" like holes and puddles) and hasn't yet reached the hoped-for autonomy. Conversely, some technologies accelerate development itself: AI, additive manufacturing, simulation/modeling capacity, low-code / no-code tools make it possible to do, faster and with fewer people, what once took months or years.

Destinus acquired a Zurich company (150+ developers) for a Vision Navigation System software, transposed from civilian to military for two critical applications: navigation in contested environments (GPS-denied, against jamming and spoofing) and terminal guidance (autonomous target confirmation and destruction, human intervention being too slow), plus video/sensor processing and embedded radars.

Regulators, technologies or markets? Boost Aerospace doesn't decide: the three complement each other in cycles, and the Formula 1 analogy illustrates it — constant performance and innovation (technology), the expectation of a show (market), and regulation without which the sector would have disappeared after accidents. The AID adds a fourth player, investors, and concludes: the winner will be whoever creates the best ecosystem (balancing regulation / market access / speed / financial support), targeting at least the European market, since most defense technologies are dual-use. Destinus decides, as a "good trader": it's the market — customer need — that shapes things, with necessary regulation (standards, STANAG) against the explosion in the number of players.

Can you reconcile speed and safety in manufacturing? Yes, according to the panelists: they aren't contradictory imperatives but two sides of the same competitiveness. The condition is knowing precisely where the safety "red lines" are that you don't cross, which requires expertise.

What is the Defense Innovation Agency (AID)? Created in 2018 within the French Ministry of the Armed Forces, it works as a "project factory" to detect, capture and mature innovations (including outside defense), through grants, public contracts and equity stakes, preparing the technology building blocks for future military capabilities.

What is TRL (Technology Readiness Level)? It's a technology maturity scale. In the cited example, an innovation must reach TRL 6-7 to be tested in an experimentation center, while accepting non-maximal operational performance, then continuously improved.

Why is explainable AI important? Because a user — a pilot, for instance — will accept an AI recommendation only if they understand its justification. Explainability conditions trust and therefore field adoption.

This panel was recorded at Tech For Industry Show 2026, the B2B trade show for industrial transformation (Smart Manufacturing, AI, Supply Chain) at Paris Expo Porte de Versailles. Watch the full discussion above, and join us on May 19–20, 2027, Hall 7, for the next edition.