From the battlefield to smart buildings: what military AI means for digital twins

To map out the environment, intelligence forces reportedly hacked nearly all the traffic cameras in Tehran to monitor exactly who was coming and going, and when. But live video was just one piece of the puzzle. The system simultaneously ingested a “sea of data” from 179 different real-time sources, including satellites, drone surveillance and historical intelligence databases.

Human brains cannot process 179 chaotic data streams at once. Claude did.

Running on dedicated, concentrated data-centre hardware, the AI acted as a massive reasoning engine. It instantly organised the incoming data, identified anomalies and prioritised targets in real-time based on strategic importance. It issued precise location coordinates, allowing a single artillery unit of 20 people to do the analytical work of 2,000 staff. It turned weeks of complex spatial planning into split-second, automated operations.

Strip away the warfare context and look purely at the system architecture, and it is very close to a smart city: a dedicated AI model processing massive, multi-modal spatial data streams to find anomalies, prioritise actions and execute operations in real-time.

When this level of compute becomes commercially mainstream, it will be the missing puzzle piece that turns digital twins and cities from ‘cool 3D dashboards’ into autonomous operational engines. If you swap out military intelligence for BIM data and replace satellite feeds with IoT sensors, the AI’s ‘targets’ simply become carbon emissions, equipment failures and energy waste.

Here is what that looks like in practice:

We cannot talk about this technology without acknowledging the glaring ethical issue. The very capability that made the Maven system so powerful in Iran (live tracking of human movement via hacked traffic cameras) is inherently invasive. If we apply this exact architecture to smart buildings to optimise space utilisation, we are fundamentally talking about the live monitoring of user behaviour.

Using anonymised CO2 sensors to detect if a room is occupied so you can turn off the lights is one thing. But if a commercial digital twin uses high-res camera networks to track exactly who is in the building, how long they spend at their desk, or how often they go to the break room, it crosses a severe ethical line.

As this technology scales into commercial real estate, the FM and operations industries are going to face a massive reckoning. There is a razor-thin line between a highly optimised, sustainable building and a corporate panopticon. We will have to establish strict guardrails, relying on ‘privacy-by-design’ (like using low-res thermal blobs instead of facial recognition) to ensure we are optimising the building, not the surveillance of the employees inside it.

If we can solve the privacy equation, this AI architecture fundamentally shifts the financial reality of our industry.

For years, the BIM business has struggled with a specific problem: BIM is largely treated as an upfront design and construction cost. Once the building is handed over, the rich 3D data is often tossed on a server and forgotten because it is too expensive to manually keep updated.

But think about how Google Maps works. A digital map of roads is somewhat useful on its own. But Google Maps became a multi-billion-dollar asset when it started layering live user data (the speed of smartphones moving in cars) over the static map. Google doesn’t pay cartographers to manually map traffic jams; it monitors user behaviour, and that behaviour dynamically validates and updates the map in real time.

Buildings are about to follow the same path.

An AI engine needs a map to understand its territory. To make sense of thousands of live sensor feeds, it requires a high-fidelity, spatial operating system: it requires BIM. Once that building is live, the continuous monitoring of occupant behaviour, thermal loads, and equipment performance will feed back into the model. The users themselves become the sensors that validate the space.

This completely changes the business case for creating BIM data. Instead of being an upfront handover cost, BIM becomes the foundational grid for a self-updating, highly monetisable operational asset.

But how exactly does a building become a ‘map’ of spatial data? How does this shift rewrite the contracts and deliverables for architecture and engineering firms? And most importantly, when a building becomes a massive data-harvesting platform, who actually owns that goldmine of behavioural data? That is a multi-million dollar paradigm shift.

For more of Léon van Berlo‘s thoughts, head to his Substack page.

Keep up to date with DC+: sign up for the midweek newsletter.