On Monday 16 March I sat on a panel called Reimagining Modern Warfare at DAIC Connect 2026 in the City of London, in the Defence Rewired stream hosted by Commodore Marcus Hember of the Royal Navy. He was on the panel too, with Colonel (Retired) Bob Christopher MBE of Intellium AI, Dan Tadross of Scale AI, Shefali Sharma of Oxford Dynamics, Lucinda Kirk of Nominal and me for Actica.
Seven months on I can't give you a transcript. I remember what I went in to argue and two things I said, and the rest is a bit of a blur, so this is my side of it and nobody else's.
The first thing I said was that automation in the military isn't new. Since 1980 ships have carried guns that detect, track and fire on incoming missiles without a person approving each shot, because nobody can react in the few seconds available. I wrote that up afterwards in What does "human in the loop" actually mean?. It mattered for that panel because, if machines acting on their own is nothing new, reimagining can't be about that. It has to be about where the person sits.
Where the person sits today is in front of a growing number of screens. Most of what I see called AI in defence is added to a force designed before it existed: a tool for analysing intelligence, another for logistics, another for targeting, each useful and each with its own interface. A commander ends up with an app zoo, and the joining up is still done by people.
The second thing I said is the argument I still hold. The place for AI here is as an abstraction layer: the layer a commander states intent through. The commander says what they want to happen and the limits to keep, and the system works out which sensors, data, models and units that involves and comes back with options and the reasoning behind them. It's an idea I keep coming back to, and not only in defence, because the same shape appears wherever people have started telling software what they want instead of operating it.
If that is where AI belongs, the design question changes. You don't add the layer to a ship, a brigade or a headquarters after they're built. You design them around it, instead of fitting it to structures that assumed people would do all the integration.
I still think it because of the programmes I've worked on, across nuclear, critical national infrastructure, national security and defence. In each of them the technology existed before the organisation was ready for it, and the time went on changing how people worked, who they trusted with data and what they were willing to hand to a machine. I expect the same here.
So when I hear "reimagining modern warfare", I don't hear better models. I hear a force redesigned around the layer its commanders talk to: who commands what, how a headquarters is staffed, which decisions stay with a person and which are handed over inside limits that person has set. Bolting AI onto the force we have gives us today's force, better equipped. Designing the force around the layer is the reimagining, and most of that work is organisational.
Sources